From hsinnrei at adobe.com  Wed Nov  1 08:52:26 2006
From: hsinnrei at adobe.com (Henry Sinnreich)
Date: Wed, 1 Nov 2006 05:52:26 -0800
Subject: [p2p-sip] What consensus? (was: What controls the lock icon?
	(Was RE:[Sip]	RE: TLS meaning))
In-Reply-To: <OF1988AD72.4203B00A-ONC2257219.00359487-C2257219.00413E78@il.ibm.com>
Message-ID: <24CCCC428EFEA2469BF046DB3C7A8D221412A8@namail5.corp.adobe.com>

Avshalom Houri writes:

 

>Other then a secure channel or S/MIME I do not think that it will be 

>possible to send a SIP message and get real assurance that the message 
>will be sent securely in every hop.

 

As you correctly conclude, P2P SIP has the opportunity to finally make
SIP secure.

 

Thanks, Henry

 

________________________________

From: Avshalom Houri [mailto:AVSHALOM at il.ibm.com] 
Sent: Wednesday, November 01, 2006 5:55 AM
To: sip at ietf.org
Subject: Re: What consensus? (was: What controls the lock icon? (Was
RE:[Sip] RE: TLS meaning))

 


Seems that we need a mechanism to create a trusted "channel" between 
endpoints that goes through all the intermediaries. The channel should 
verifiable as secure between the endpoints. After the channel is created

then the SIP signalling can go securely over the channel. 

The protocol for creating such a channel may be SIP based on 
external to SIP and after we have such a mechanism we can decide if we 
will use SIPS for messages that need to go over the secure channel 
or we will use new URI type (e.g. SSIP) or just use some 
header in the SIP message. 

The other alternative is the usage of S/MIME which has its drawbacks as 
we can see in another thread. 

Other then a secure channel or S/MIME I do not think that it will be 
possible to send a SIP message and get real assurance that the message 
will be sent securely in every hop. 

--Avshalom 




Dean Willis <dean.willis at softarmor.com> 

31/10/2006 18:51 

To

Michael Thomas <mat at cisco.com> 

cc

"Joel M. Halpern" <joel at stevecrocker.com>, "'IETF SIP List'"
<sip at ietf.org>, Juha Heinanen <jh at tutpro.com>, "Elwell, John"
<john.elwell at siemens.com> 

Subject

Re: What consensus? (was: What controls the lock icon? (Was RE: [Sip]
RE: TLS meaning))

 

 

 





On Oct 31, 2006, at 8:47 AM, Michael Thomas wrote:

> My advise to the working group is to just start over from first  
> principles as if
> SIPS didn't even exist -- which in reality it doesn't. We need to  
> be thinking
> about these things for the long term, and if it's not baked it's  
> not baked. There
> have been many drafts that have lingered on and completely  
> overhauled over
> the years. Why SIPS is treated so differently in that respect is  
> somewhat
> mystifying because dealing with security for SIP is a genuinely  
> hard problem.
> It shouldn't be rushed, and it deserves better than a kludgey  
> security patina
> airbrushed in to get past the IESG.

I'm very tempted to endorse this, but I think you're missing one thing.

Flimsy as it is, I believe that sips: is as strong as the rest of the  
current SIP security model since it depends on the same trust  
assumptions.

We can't just restart on sips: without restarting from those  
fundamental trust assumptions.

This is not trivial. Many people would argue that it would turn SIP  
into a fundamentally different protocol. I think it needs to be done,  
but I also think we need to clearly document how the current trust  
model (including sips:) works, because that's what's going to be  
deployed in the real world for a relatively long time.

The P2PSIP work is having to re-examine those fundamental trust  
assumptions. My current hope is we can do some work there that can be  
"ported back" into SIP.

--
Dean

_______________________________________________
Sip mailing list  https://www1.ietf.org/mailman/listinfo/sip
This list is for NEW development of the core SIP Protocol
Use sip-implementors at cs.columbia.edu for questions on current sip
Use sipping at ietf.org for new developments on the application of sip

-------------- next part --------------
An HTML attachment was scrubbed...
URL: https://lists.cs.columbia.edu/pipermail/p2p-sip/attachments/20061101/46595181/attachment-0001.html

From huang-ming.pan at comcast.net  Wed Nov  1 09:40:07 2006
From: huang-ming.pan at comcast.net (Peter Pan)
Date: Wed, 1 Nov 2006 06:40:07 -0800
Subject: [p2p-sip] What consensus? (was: What controls the lock
	icon?(Was RE:[Sip]	RE: TLS meaning))
References: <24CCCC428EFEA2469BF046DB3C7A8D221412A8@namail5.corp.adobe.com>
Message-ID: <000f01c6fdc3$a1069110$030aa8c0@comcast.net>

The disentrance for a secure channel is sweet. It seems to me the first of
the keypoints that will make P2PSIP win. SIP itself seems never one :(

Peter
----- Original Message -----
From: Henry Sinnreich
To: Avshalom Houri ; sip at ietf.org
Cc: p2p-sip at cs.columbia.edu
Sent: Wednesday, November 01, 2006 5:52 AM
Subject: Re: [p2p-sip] What consensus? (was: What controls the lock
icon?(Was RE:[Sip] RE: TLS meaning))


Avshalom Houri writes:

>Other then a secure channel or S/MIME I do not think that it will be
>possible to send a SIP message and get real assurance that the message
>will be sent securely in every hop.

As you correctly conclude, P2P SIP has the opportunity to finally make SIP
secure.

Thanks, Henry




From: Avshalom Houri [mailto:AVSHALOM at il.ibm.com]
Sent: Wednesday, November 01, 2006 5:55 AM
To: sip at ietf.org
Subject: Re: What consensus? (was: What controls the lock icon? (Was
RE:[Sip] RE: TLS meaning))


Seems that we need a mechanism to create a trusted "channel" between
endpoints that goes through all the intermediaries. The channel should
verifiable as secure between the endpoints. After the channel is created
then the SIP signalling can go securely over the channel.

The protocol for creating such a channel may be SIP based on
external to SIP and after we have such a mechanism we can decide if we
will use SIPS for messages that need to go over the secure channel
or we will use new URI type (e.g. SSIP) or just use some
header in the SIP message.

The other alternative is the usage of S/MIME which has its drawbacks as
we can see in another thread.

Other then a secure channel or S/MIME I do not think that it will be
possible to send a SIP message and get real assurance that the message
will be sent securely in every hop.

--Avshalom



Dean Willis <dean.willis at softarmor.com>
31/10/2006 18:51 ToMichael Thomas <mat at cisco.com>
cc"Joel M. Halpern" <joel at stevecrocker.com>, "'IETF SIP List'"
<sip at ietf.org>, Juha Heinanen <jh at tutpro.com>, "Elwell, John"
<john.elwell at siemens.com>
SubjectRe: What consensus? (was: What controls the lock icon? (Was RE: [Sip]
RE: TLS meaning))









On Oct 31, 2006, at 8:47 AM, Michael Thomas wrote:

> My advise to the working group is to just start over from first
> principles as if
> SIPS didn't even exist -- which in reality it doesn't. We need to
> be thinking
> about these things for the long term, and if it's not baked it's
> not baked. There
> have been many drafts that have lingered on and completely
> overhauled over
> the years. Why SIPS is treated so differently in that respect is
> somewhat
> mystifying because dealing with security for SIP is a genuinely
> hard problem.
> It shouldn't be rushed, and it deserves better than a kludgey
> security patina
> airbrushed in to get past the IESG.

I'm very tempted to endorse this, but I think you're missing one thing.

Flimsy as it is, I believe that sips: is as strong as the rest of the
current SIP security model since it depends on the same trust
assumptions.

We can't just restart on sips: without restarting from those
fundamental trust assumptions.

This is not trivial. Many people would argue that it would turn SIP
into a fundamentally different protocol. I think it needs to be done,
but I also think we need to clearly document how the current trust
model (including sips:) works, because that's what's going to be
deployed in the real world for a relatively long time.

The P2PSIP work is having to re-examine those fundamental trust
assumptions. My current hope is we can do some work there that can be
"ported back" into SIP.

--
Dean

_______________________________________________
Sip mailing list  https://www1.ietf.org/mailman/listinfo/sip
This list is for NEW development of the core SIP Protocol
Use sip-implementors at cs.columbia.edu for questions on current sip
Use sipping at ietf.org for new developments on the application of sip



_______________________________________________
p2p-sip mailing list
p2p-sip at cs.columbia.edu
https://lists.cs.columbia.edu/cucslists/listinfo/p2p-sip


From edpimentl at gmail.com  Wed Nov  1 09:58:51 2006
From: edpimentl at gmail.com (Ed Pimentel)
Date: Wed, 1 Nov 2006 09:58:51 -0500
Subject: [p2p-sip] What consensus? (was: What controls the lock
	icon?(Was RE:[Sip]	RE: TLS meaning))
In-Reply-To: <000f01c6fdc3$a1069110$030aa8c0@comcast.net>
References: <24CCCC428EFEA2469BF046DB3C7A8D221412A8@namail5.corp.adobe.com>
	<000f01c6fdc3$a1069110$030aa8c0@comcast.net>
Message-ID: <6B6A6EF9-6B71-4200-B349-B4951CE6F07D@gmail.com>

What is missing is the concept of a federated SAML / OpenID / Yadis   
"iD Provider"
Or even an adaptation of what was with H.350 (Internet2.0) but  for  
P2P-SIP
Sincerely,
-EdPimentl

On Nov 1, 2006, at 9:40 AM, Peter Pan wrote:

> The disentrance for a secure channel is sweet. It seems to me the  
> first of
> the keypoints that will make P2PSIP win. SIP itself seems never one :(
>
> Peter
> ----- Original Message -----
> From: Henry Sinnreich
> To: Avshalom Houri ; sip at ietf.org
> Cc: p2p-sip at cs.columbia.edu
> Sent: Wednesday, November 01, 2006 5:52 AM
> Subject: Re: [p2p-sip] What consensus? (was: What controls the lock
> icon?(Was RE:[Sip] RE: TLS meaning))
>
>
> Avshalom Houri writes:
>
>> Other then a secure channel or S/MIME I do not think that it will be
>> possible to send a SIP message and get real assurance that the  
>> message
>> will be sent securely in every hop.
>
> As you correctly conclude, P2P SIP has the opportunity to finally  
> make SIP
> secure.
>
> Thanks, Henry
>
>
>
>
> From: Avshalom Houri [mailto:AVSHALOM at il.ibm.com]
> Sent: Wednesday, November 01, 2006 5:55 AM
> To: sip at ietf.org
> Subject: Re: What consensus? (was: What controls the lock icon? (Was
> RE:[Sip] RE: TLS meaning))
>
>
> Seems that we need a mechanism to create a trusted "channel" between
> endpoints that goes through all the intermediaries. The channel should
> verifiable as secure between the endpoints. After the channel is  
> created
> then the SIP signalling can go securely over the channel.
>
> The protocol for creating such a channel may be SIP based on
> external to SIP and after we have such a mechanism we can decide if we
> will use SIPS for messages that need to go over the secure channel
> or we will use new URI type (e.g. SSIP) or just use some
> header in the SIP message.
>
> The other alternative is the usage of S/MIME which has its  
> drawbacks as
> we can see in another thread.
>
> Other then a secure channel or S/MIME I do not think that it will be
> possible to send a SIP message and get real assurance that the message
> will be sent securely in every hop.
>
> --Avshalom
>
>
>
> Dean Willis <dean.willis at softarmor.com>
> 31/10/2006 18:51 ToMichael Thomas <mat at cisco.com>
> cc"Joel M. Halpern" <joel at stevecrocker.com>, "'IETF SIP List'"
> <sip at ietf.org>, Juha Heinanen <jh at tutpro.com>, "Elwell, John"
> <john.elwell at siemens.com>
> SubjectRe: What consensus? (was: What controls the lock icon? (Was  
> RE: [Sip]
> RE: TLS meaning))
>
>
>
>
>
>
>
>
>
> On Oct 31, 2006, at 8:47 AM, Michael Thomas wrote:
>
>> My advise to the working group is to just start over from first
>> principles as if
>> SIPS didn't even exist -- which in reality it doesn't. We need to
>> be thinking
>> about these things for the long term, and if it's not baked it's
>> not baked. There
>> have been many drafts that have lingered on and completely
>> overhauled over
>> the years. Why SIPS is treated so differently in that respect is
>> somewhat
>> mystifying because dealing with security for SIP is a genuinely
>> hard problem.
>> It shouldn't be rushed, and it deserves better than a kludgey
>> security patina
>> airbrushed in to get past the IESG.
>
> I'm very tempted to endorse this, but I think you're missing one  
> thing.
>
> Flimsy as it is, I believe that sips: is as strong as the rest of the
> current SIP security model since it depends on the same trust
> assumptions.
>
> We can't just restart on sips: without restarting from those
> fundamental trust assumptions.
>
> This is not trivial. Many people would argue that it would turn SIP
> into a fundamentally different protocol. I think it needs to be done,
> but I also think we need to clearly document how the current trust
> model (including sips:) works, because that's what's going to be
> deployed in the real world for a relatively long time.
>
> The P2PSIP work is having to re-examine those fundamental trust
> assumptions. My current hope is we can do some work there that can be
> "ported back" into SIP.
>
> --
> Dean
>
> _______________________________________________
> Sip mailing list  https://www1.ietf.org/mailman/listinfo/sip
> This list is for NEW development of the core SIP Protocol
> Use sip-implementors at cs.columbia.edu for questions on current sip
> Use sipping at ietf.org for new developments on the application of sip
>
>
>
> _______________________________________________
> p2p-sip mailing list
> p2p-sip at cs.columbia.edu
> https://lists.cs.columbia.edu/cucslists/listinfo/p2p-sip
>
> _______________________________________________
> p2p-sip mailing list
> p2p-sip at cs.columbia.edu
> https://lists.cs.columbia.edu/cucslists/listinfo/p2p-sip


From bryan at ethernot.org  Wed Nov  1 12:14:28 2006
From: bryan at ethernot.org (David A. Bryan)
Date: Wed, 1 Nov 2006 12:14:28 -0500
Subject: [p2p-sip] IMPORTANT!!! Updated Charter
In-Reply-To: <454378A5.8040404@research.panasonic.com>
References: <8b2769930610271420n7162e480w7fbac933bc34e5d4@mail.gmail.com>
	<20061028061203.GE6564@sbrim-wxp01>
	<454378A5.8040404@research.panasonic.com>
Message-ID: <8b2769930611010914kad12dedid442565361c061ff@mail.gmail.com>

Any further comments? I'd like to make the final "pre-BOF" round of
changes, so if there are other issues, lets discuss them now.
Otherwise I am hoping I can iterate this with the small changes below,
and take this version to the BOF.

Currently I plan to fix the following:

1) fix the "opeer" typo

2) Change bullet 3 under "excluded topics" to read:

"3. Locating resources based on something other than URIs. In other
  words, arbitrary search of attributes is out of scope, but locating
  resources based on their URIs is in scope. Using URIs need not imply
  using the DNS or having a record in the DNS for the URI."

3) The third issue raised by Scott, with proposed wording by Eunsoo
and Scott, is a bit trickier. In the very beginning, while explaining
what P2PSIP is, we say:

"The Peer-to-Peer (P2P) Session Initiation Protocol working group
(P2PSIP WG) is chartered to develop protocols and mechanisms for the
use of the Session Initiation Protocol (SIP) in settings where the
service of establishing and managing sessions is principally handled
by a collection of intelligent endpoints, or P2PSIP peers,
rather than centralized servers as in client-server SIP as currently
deployed."

Later we say:

"P2PSIP clients and peers use the resolution
services of the peers as an alternative to the SIP correspondent
discovery process of RFC 3263. Session management, messaging, and
presence functions are performed using traditional SIP."

The intent in the first one is that the end points are the things
doing the work, rather than servers -- it is a very broad statement
(and one I think is true) asserting that we don't need servers for
managing the calls, so I'd like to leave that.

The second is a more specific point, saying that we will reuse SIP as
much possible and that we are really looking at replacing the
resolution portion of SIP. Apparently the wording is a bit confusing
as a consequence.

I would like to propose changing the second sentence under discussion to read:

"P2PSIP clients and peers use the resolution
services of the peers as an alternative to the SIP correspondent
discovery process of RFC 3263. [The remaining aspects of] session
management, messaging, and
presence functions are performed using traditional SIP."

(new text in [])

Scott and Eunsoo: does that fix it?

David

On 10/28/06, Eunsoo Shim <eunsoo at research.panasonic.com> wrote:
> Scott W Brim wrote:
>
> >First, you missed an "o" -- search for "opeer".
> >
> >Second,
> >
> >
> >
> >>service of establishing and managing sessions is principally handled
> >>by a collection of intelligent endpoints, or P2PSIP peers,
> >>
> >>
> >
> >
> >
> Since 'P2PSIP Peer' has a certain meaning, narrower than 'intelligent
> endpoint',
> For example, P2PSIP Clients are also such intelligent endpoints.
> One remedy may be just removing 'or P2PSIP peers' from the sentence.
>
> >I think you want to take "and managing" out of the first line.
> >Compare you have below ...
> >
> >
> >
> >>P2PSIP overlay. P2PSIP clients and peers use the resolution
> >>services of the peers as an alternative to the SIP correspondent
> >>discovery process of RFC 3263. Session management, messaging, and
> >>presence functions are performed using traditional SIP.
> >>
> >>
> >
> >
> >
> Thanks.
>
> Eunsoo
> _______________________________________________
> p2p-sip mailing list
> p2p-sip at cs.columbia.edu
> https://lists.cs.columbia.edu/cucslists/listinfo/p2p-sip
>

From slavitch at gmail.com  Wed Nov  1 12:47:55 2006
From: slavitch at gmail.com (Michael Slavitch)
Date: Wed, 1 Nov 2006 12:47:55 -0500
Subject: [p2p-sip] IMPORTANT!!! Updated Charter
In-Reply-To: <8b2769930611010914kad12dedid442565361c061ff@mail.gmail.com>
References: <8b2769930610271420n7162e480w7fbac933bc34e5d4@mail.gmail.com>
	<20061028061203.GE6564@sbrim-wxp01>
	<454378A5.8040404@research.panasonic.com>
	<8b2769930611010914kad12dedid442565361c061ff@mail.gmail.com>
Message-ID: <1fc32c8f0611010947kcd80298m9fe5198f930ead2d@mail.gmail.com>

Looks like soup is almost done.

Good work.

M

From hsinnrei at adobe.com  Wed Nov  1 14:42:32 2006
From: hsinnrei at adobe.com (Henry Sinnreich)
Date: Wed, 1 Nov 2006 11:42:32 -0800
Subject: [p2p-sip] IMPORTANT!!! Updated Charter
In-Reply-To: <8b2769930611010914kad12dedid442565361c061ff@mail.gmail.com>
Message-ID: <24CCCC428EFEA2469BF046DB3C7A8D221412B1@namail5.corp.adobe.com>

The updated charter is excellent!

Folks: There is only on action item now, to get the P2PSIP WG formed.

We must postpone any technical issues for later, after having the WG
chartered.

Henry

-----Original Message-----
From: p2p-sip-bounces at cs.columbia.edu
[mailto:p2p-sip-bounces at cs.columbia.edu] On Behalf Of David A. Bryan
Sent: Wednesday, November 01, 2006 11:14 AM
To: Eunsoo Shim
Cc: P2P-SIP
Subject: Re: [p2p-sip] IMPORTANT!!! Updated Charter

Any further comments? I'd like to make the final "pre-BOF" round of
changes, so if there are other issues, lets discuss them now.
Otherwise I am hoping I can iterate this with the small changes below,
and take this version to the BOF.

Currently I plan to fix the following:

1) fix the "opeer" typo

2) Change bullet 3 under "excluded topics" to read:

"3. Locating resources based on something other than URIs. In other
  words, arbitrary search of attributes is out of scope, but locating
  resources based on their URIs is in scope. Using URIs need not imply
  using the DNS or having a record in the DNS for the URI."

3) The third issue raised by Scott, with proposed wording by Eunsoo
and Scott, is a bit trickier. In the very beginning, while explaining
what P2PSIP is, we say:

"The Peer-to-Peer (P2P) Session Initiation Protocol working group
(P2PSIP WG) is chartered to develop protocols and mechanisms for the
use of the Session Initiation Protocol (SIP) in settings where the
service of establishing and managing sessions is principally handled
by a collection of intelligent endpoints, or P2PSIP peers,
rather than centralized servers as in client-server SIP as currently
deployed."

Later we say:

"P2PSIP clients and peers use the resolution
services of the peers as an alternative to the SIP correspondent
discovery process of RFC 3263. Session management, messaging, and
presence functions are performed using traditional SIP."

The intent in the first one is that the end points are the things
doing the work, rather than servers -- it is a very broad statement
(and one I think is true) asserting that we don't need servers for
managing the calls, so I'd like to leave that.

The second is a more specific point, saying that we will reuse SIP as
much possible and that we are really looking at replacing the
resolution portion of SIP. Apparently the wording is a bit confusing
as a consequence.

I would like to propose changing the second sentence under discussion to
read:

"P2PSIP clients and peers use the resolution
services of the peers as an alternative to the SIP correspondent
discovery process of RFC 3263. [The remaining aspects of] session
management, messaging, and
presence functions are performed using traditional SIP."

(new text in [])

Scott and Eunsoo: does that fix it?

David

On 10/28/06, Eunsoo Shim <eunsoo at research.panasonic.com> wrote:
> Scott W Brim wrote:
>
> >First, you missed an "o" -- search for "opeer".
> >
> >Second,
> >
> >
> >
> >>service of establishing and managing sessions is principally handled
> >>by a collection of intelligent endpoints, or P2PSIP peers,
> >>
> >>
> >
> >
> >
> Since 'P2PSIP Peer' has a certain meaning, narrower than 'intelligent
> endpoint',
> For example, P2PSIP Clients are also such intelligent endpoints.
> One remedy may be just removing 'or P2PSIP peers' from the sentence.
>
> >I think you want to take "and managing" out of the first line.
> >Compare you have below ...
> >
> >
> >
> >>P2PSIP overlay. P2PSIP clients and peers use the resolution
> >>services of the peers as an alternative to the SIP correspondent
> >>discovery process of RFC 3263. Session management, messaging, and
> >>presence functions are performed using traditional SIP.
> >>
> >>
> >
> >
> >
> Thanks.
>
> Eunsoo
> _______________________________________________
> p2p-sip mailing list
> p2p-sip at cs.columbia.edu
> https://lists.cs.columbia.edu/cucslists/listinfo/p2p-sip
>
_______________________________________________
p2p-sip mailing list
p2p-sip at cs.columbia.edu
https://lists.cs.columbia.edu/cucslists/listinfo/p2p-sip


From eric_d_cooper at sympatico.ca  Wed Nov  1 15:04:34 2006
From: eric_d_cooper at sympatico.ca (Eric Cooper)
Date: Wed, 1 Nov 2006 15:04:34 -0500
Subject: [p2p-sip] Comments on draft-willis-p2psip-concepts-03
References: <20061031015443.7CEB45C01E@laser.networkresonance.com><6F914E17-0377-4F4A-B794-9DFF64E1A2BD@softarmor.com><868xiw35kg.fsf@raman.networkresonance.com><13F6728B-6CD3-4F23-BBD0-DB5E01067E53@softarmor.com><86lkmw1h7p.fsf@raman.networkresonance.com><8b2769930610311145w4089f2datac58a32384e8291c@mail.gmail.com><86lkmwz2fy.fsf@raman.networkresonance.com><8b2769930610311217h29bd1a37wb1c0f1566f271414@mail.gmail.com>
	<4F21983F-69D1-4311-8D3F-492646181083@softarmor.com>
Message-ID: <BAYC1-PASMTP108AFDF599E303646E37C1CCF80@CEZ.ICE>

> 
> I actually envision an "outbound peer" that has the same kind of  
> relationship to a client (or to another peer) as an "outbound proxy"  
> does to a UA. It acts as the NAT-traversal anchor for requests from  
> outside the NAT to reach the peer or client behind a NAT.
> 
> Dean
> 

I understand why an 'outbound peer' would be a useful notion for a client, but I  don't understand why a peer would need one.  
I thought that NAT traversal was going to be included as part of the peer protocol.  Is there a special distinction between peers with public IP addresses and peers that are behind NATs?

Eric Cooper.

_______________________________________________
> p2p-sip mailing list
> p2p-sip at cs.columbia.edu
> https://lists.cs.columbia.edu/cucslists/listinfo/p2p-sip


From hsinnrei at adobe.com  Wed Nov  1 08:52:26 2006
From: hsinnrei at adobe.com (Henry Sinnreich)
Date: Wed, 1 Nov 2006 05:52:26 -0800
Subject: [p2p-sip] What consensus? (was: What controls the lock icon?
	(Was	RE:[Sip]	RE: TLS meaning))
In-Reply-To: <OF1988AD72.4203B00A-ONC2257219.00359487-C2257219.00413E78@il.ibm.com>
Message-ID: <24CCCC428EFEA2469BF046DB3C7A8D221412A8@namail5.corp.adobe.com>

Avshalom Houri writes:

 

>Other then a secure channel or S/MIME I do not think that it will be 

>possible to send a SIP message and get real assurance that the message 
>will be sent securely in every hop.

 

As you correctly conclude, P2P SIP has the opportunity to finally make
SIP secure.

 

Thanks, Henry

 

________________________________

From: Avshalom Houri [mailto:AVSHALOM at il.ibm.com] 
Sent: Wednesday, November 01, 2006 5:55 AM
To: sip at ietf.org
Subject: Re: What consensus? (was: What controls the lock icon? (Was
RE:[Sip] RE: TLS meaning))

 


Seems that we need a mechanism to create a trusted "channel" between 
endpoints that goes through all the intermediaries. The channel should 
verifiable as secure between the endpoints. After the channel is created

then the SIP signalling can go securely over the channel. 

The protocol for creating such a channel may be SIP based on 
external to SIP and after we have such a mechanism we can decide if we 
will use SIPS for messages that need to go over the secure channel 
or we will use new URI type (e.g. SSIP) or just use some 
header in the SIP message. 

The other alternative is the usage of S/MIME which has its drawbacks as 
we can see in another thread. 

Other then a secure channel or S/MIME I do not think that it will be 
possible to send a SIP message and get real assurance that the message 
will be sent securely in every hop. 

--Avshalom 




Dean Willis <dean.willis at softarmor.com> 

31/10/2006 18:51 

To

Michael Thomas <mat at cisco.com> 

cc

"Joel M. Halpern" <joel at stevecrocker.com>, "'IETF SIP List'"
<sip at ietf.org>, Juha Heinanen <jh at tutpro.com>, "Elwell, John"
<john.elwell at siemens.com> 

Subject

Re: What consensus? (was: What controls the lock icon? (Was RE: [Sip]
RE: TLS meaning))

 

 

 





On Oct 31, 2006, at 8:47 AM, Michael Thomas wrote:

> My advise to the working group is to just start over from first  
> principles as if
> SIPS didn't even exist -- which in reality it doesn't. We need to  
> be thinking
> about these things for the long term, and if it's not baked it's  
> not baked. There
> have been many drafts that have lingered on and completely  
> overhauled over
> the years. Why SIPS is treated so differently in that respect is  
> somewhat
> mystifying because dealing with security for SIP is a genuinely  
> hard problem.
> It shouldn't be rushed, and it deserves better than a kludgey  
> security patina
> airbrushed in to get past the IESG.

I'm very tempted to endorse this, but I think you're missing one thing.

Flimsy as it is, I believe that sips: is as strong as the rest of the  
current SIP security model since it depends on the same trust  
assumptions.

We can't just restart on sips: without restarting from those  
fundamental trust assumptions.

This is not trivial. Many people would argue that it would turn SIP  
into a fundamentally different protocol. I think it needs to be done,  
but I also think we need to clearly document how the current trust  
model (including sips:) works, because that's what's going to be  
deployed in the real world for a relatively long time.

The P2PSIP work is having to re-examine those fundamental trust  
assumptions. My current hope is we can do some work there that can be  
"ported back" into SIP.

--
Dean

_______________________________________________
Sip mailing list  https://www1.ietf.org/mailman/listinfo/sip
This list is for NEW development of the core SIP Protocol
Use sip-implementors at cs.columbia.edu for questions on current sip
Use sipping at ietf.org for new developments on the application of sip

-------------- next part --------------
An HTML attachment was scrubbed...
URL: https://lists.cs.columbia.edu/pipermail/p2p-sip/attachments/20061101/46595181/attachment-0002.html
-------------- next part --------------
_______________________________________________
Sip mailing list  https://www1.ietf.org/mailman/listinfo/sip
This list is for NEW development of the core SIP Protocol
Use sip-implementors at cs.columbia.edu for questions on current sip
Use sipping at ietf.org for new developments on the application of sip

From RADHIKA.R.ROY at saic.com  Wed Nov  1 14:56:50 2006
From: RADHIKA.R.ROY at saic.com (Roy, Radhika R.)
Date: Wed, 1 Nov 2006 14:56:50 -0500 
Subject: [p2p-sip] IMPORTANT!!! Updated Charter
Message-ID: <62D92A9A02BCC845B202323D49A48426230EE5@0591-ITS-EXMP02.us.saic.com>

Hi, all:

Like Henry, I also think that the forming of the P2PSIP WG will be the
only objective from now on. All hard works of all members need to have a
fruit.

Best regards,
Radhika

-----Original Message-----
From: p2p-sip-bounces at cs.columbia.edu
[mailto:p2p-sip-bounces at cs.columbia.edu] On Behalf Of Henry Sinnreich
Sent: Wednesday, November 01, 2006 2:43 PM
To: David A. Bryan; Eunsoo Shim
Cc: P2P-SIP
Subject: Re: [p2p-sip] IMPORTANT!!! Updated Charter

The updated charter is excellent!

Folks: There is only on action item now, to get the P2PSIP WG formed.

We must postpone any technical issues for later, after having the WG
chartered.

Henry


From dean.willis at softarmor.com  Wed Nov  1 22:28:05 2006
From: dean.willis at softarmor.com (Dean Willis)
Date: Wed, 1 Nov 2006 21:28:05 -0600
Subject: [p2p-sip] Comments on draft-willis-p2psip-concepts-03
In-Reply-To: <BAYC1-PASMTP108AFDF599E303646E37C1CCF80@CEZ.ICE>
References: <20061031015443.7CEB45C01E@laser.networkresonance.com><6F914E17-0377-4F4A-B794-9DFF64E1A2BD@softarmor.com><868xiw35kg.fsf@raman.networkresonance.com><13F6728B-6CD3-4F23-BBD0-DB5E01067E53@softarmor.com><86lkmw1h7p.fsf@raman.networkresonance.com><8b2769930610311145w4089f2datac58a32384e8291c@mail.gmail.com><86lkmwz2fy.fsf@raman.networkresonance.com><8b2769930610311217h29bd1a37wb1c0f1566f271414@mail.gmail.com>
	<4F21983F-69D1-4311-8D3F-492646181083@softarmor.com>
	<BAYC1-PASMTP108AFDF599E303646E37C1CCF80@CEZ.ICE>
Message-ID: <57A90420-7FDE-4E69-9B24-7F26DB12E786@softarmor.com>


On Nov 1, 2006, at 2:04 PM, Eric Cooper wrote:

>>
>> I actually envision an "outbound peer" that has the same kind of
>> relationship to a client (or to another peer) as an "outbound proxy"
>> does to a UA. It acts as the NAT-traversal anchor for requests from
>> outside the NAT to reach the peer or client behind a NAT.
>>
>> Dean
>>
>
> I understand why an 'outbound peer' would be a useful notion for a  
> client, but I  don't understand why a peer would need one.
> I thought that NAT traversal was going to be included as part of  
> the peer protocol.  Is there a special distinction between peers  
> with public IP addresses and peers that are behind NATs?
>

Well, what part of the peer protocol is it that allows peers behind  
NATs to function?

It's not like we can just turn on a magic "NAT Traversal Bit" in the  
protocol.

The only way I know of to reliably talk to anything behind a NAT is  
to enlist the help of somebody that is already talking to the node- 
behind-the-nat via a connection established from behind-the-nat.

So for something "outside" to talk to a peer that is behind a NAT, we  
need to find another peer (or some other helper function) that is  
outside the NAT and already has a connection to the peer-behind the  
NAT. If this helper thing is a peer, then I'd say that there's an  
"outbound peer" relationship between the peer-beind-the-nat and the  
peer-outside-the-nat-that-is-helping.

--
Dean


From eric_d_cooper at sympatico.ca  Thu Nov  2 09:32:08 2006
From: eric_d_cooper at sympatico.ca (Eric Cooper)
Date: Thu, 2 Nov 2006 09:32:08 -0500
Subject: [p2p-sip] Comments on draft-willis-p2psip-concepts-03
References: <20061031015443.7CEB45C01E@laser.networkresonance.com><6F914E17-0377-4F4A-B794-9DFF64E1A2BD@softarmor.com><868xiw35kg.fsf@raman.networkresonance.com><13F6728B-6CD3-4F23-BBD0-DB5E01067E53@softarmor.com><86lkmw1h7p.fsf@raman.networkresonance.com><8b2769930610311145w4089f2datac58a32384e8291c@mail.gmail.com><86lkmwz2fy.fsf@raman.networkresonance.com><8b2769930610311217h29bd1a37wb1c0f1566f271414@mail.gmail.com>
	<4F21983F-69D1-4311-8D3F-492646181083@softarmor.com>
	<BAYC1-PASMTP108AFDF599E303646E37C1CCF80@CEZ.ICE>
	<57A90420-7FDE-4E69-9B24-7F26DB12E786@softarmor.com>
Message-ID: <BAYC1-PASMTP0260F9D9DF718C9309C0B4CCFF0@CEZ.ICE>

>>>
>>> I actually envision an "outbound peer" that has the same kind of
>>> relationship to a client (or to another peer) as an "outbound proxy"
>>> does to a UA. It acts as the NAT-traversal anchor for requests from
>>> outside the NAT to reach the peer or client behind a NAT.
>>>
>>> Dean
>>>
>>
>> I understand why an 'outbound peer' would be a useful notion for a  
>> client, but I  don't understand why a peer would need one.
>> I thought that NAT traversal was going to be included as part of  
>> the peer protocol.  Is there a special distinction between peers  
>> with public IP addresses and peers that are behind NATs?
>>
> 
> Well, what part of the peer protocol is it that allows peers behind  
> NATs to function?
> 

As an example, there could be some messaging involving a bootstrap server that would request that a bootstrap peer initiate UDP hole punching with a new peer (one that was performing it's insertion into the overlay).

> It's not like we can just turn on a magic "NAT Traversal Bit" in the  
> protocol.
> 
> The only way I know of to reliably talk to anything behind a NAT is  
> to enlist the help of somebody that is already talking to the node- 
> behind-the-nat via a connection established from behind-the-nat.
> 

In the scenario I'm thinking of, this would be a connection from a (new) peer, talking to a bootstrap server.
Suppose there's also a separate connection between a bootstrap peer that has already joined the overlay and that same bootstrap server.  The existing peer might even be behind a NAT.  The bootstrap server could coordinate UDP hole punching between the existing peer and the new peer.

> So for something "outside" to talk to a peer that is behind a NAT, we  
> need to find another peer (or some other helper function) that is  
> outside the NAT and already has a connection to the peer-behind the  
> NAT. If this helper thing is a peer, then I'd say that there's an  
> "outbound peer" relationship between the peer-beind-the-nat and the  
> peer-outside-the-nat-that-is-helping.
> 

If the UDP hole punching succeeds, then a 'link' between the two peers (both behind NATs) is established.  Would they both be 'outbound peers'?

I was just wondering if you thought the 'outbound peer' behaviour was optional, or just part of the regular responsibilities of a peer, like storage and routing.

Best Regards,

Eric.


From eunsoo at research.panasonic.com  Thu Nov  2 11:00:57 2006
From: eunsoo at research.panasonic.com (Eunsoo Shim)
Date: Thu, 02 Nov 2006 11:00:57 -0500
Subject: [p2p-sip] IMPORTANT!!! Updated Charter
In-Reply-To: <20061102055850.GJ5340@sbrim-wxp01>
References: <8b2769930610271420n7162e480w7fbac933bc34e5d4@mail.gmail.com>
	<20061028061203.GE6564@sbrim-wxp01>
	<454378A5.8040404@research.panasonic.com>
	<8b2769930611010914kad12dedid442565361c061ff@mail.gmail.com>
	<20061102055850.GJ5340@sbrim-wxp01>
Message-ID: <454A1639.4010301@research.panasonic.com>

Scott W Brim wrote:

>Excerpts from David A. Bryan on Wed, Nov 01, 2006 12:14:28PM -0500:
>  
>
>>I would like to propose changing the second sentence under discussion to read:
>>
>>"P2PSIP clients and peers use the resolution
>>services of the peers as an alternative to the SIP correspondent
>>discovery process of RFC 3263. [The remaining aspects of] session
>>management, messaging, and
>>presence functions are performed using traditional SIP."
>>
>>(new text in [])
>>
>>Scott and Eunsoo: does that fix it?
>>    
>>
>
>I still wonder a little about "and managing sessions" in the first
>paragraph (P2PSIP peers "manage" SIP sessions?) but it's a nit in the
>long run.  We can deal with it in 30 seconds in San Diego.  See you.  
>
>Scott (arriving Saturday 6pm via SFO)
>  
>
I guess Scott and I are talking about the same thing.
Simply dropping 'P2PSIP peers' from the first sentence should work for me.
The current sentence is:

"The Peer-to-Peer (P2P) Session Initiation Protocol working group
(P2PSIP WG) is chartered to develop protocols and mechanisms for the
use of the Session Initiation Protocol (SIP) in settings where the
service of establishing and managing sessions is principally handled
by a collection of intelligent endpoints, or P2PSIP peers,
rather than centralized servers as in client-server SIP as currently
deployed."

My proposal is:

"The Peer-to-Peer (P2P) Session Initiation Protocol working group
(P2PSIP WG) is chartered to develop protocols and mechanisms for the
use of the Session Initiation Protocol (SIP) in settings where the
service of establishing and managing sessions is principally handled
by a collection of intelligent endpoints,
rather than centralized servers as in client-server SIP as currently
deployed."

If necessary, we will be able to discuss and fix this easily in San
Diego as Scott said.
Thanks.

Eunsoo

From swb at employees.org  Thu Nov  2 11:10:16 2006
From: swb at employees.org (Scott W Brim)
Date: Thu, 2 Nov 2006 17:10:16 +0100
Subject: [p2p-sip] IMPORTANT!!! Updated Charter
In-Reply-To: <454A1639.4010301@research.panasonic.com>
References: <8b2769930610271420n7162e480w7fbac933bc34e5d4@mail.gmail.com>
	<20061028061203.GE6564@sbrim-wxp01>
	<454378A5.8040404@research.panasonic.com>
	<8b2769930611010914kad12dedid442565361c061ff@mail.gmail.com>
	<20061102055850.GJ5340@sbrim-wxp01>
	<454A1639.4010301@research.panasonic.com>
Message-ID: <20061102161014.GO2068@sbrim-wxp01>

Excerpts from Eunsoo Shim on Thu, Nov 02, 2006 11:00:57AM -0500:
> I guess Scott and I are talking about the same thing.  Simply
> dropping 'P2PSIP peers' from the first sentence should work for me.
> The current sentence is:
> 
> "The Peer-to-Peer (P2P) Session Initiation Protocol working group
> (P2PSIP WG) is chartered to develop protocols and mechanisms for the
> use of the Session Initiation Protocol (SIP) in settings where the
> service of establishing and managing sessions is principally handled
> by a collection of intelligent endpoints, or P2PSIP peers, rather
> than centralized servers as in client-server SIP as currently
> deployed."
> 
> My proposal is:
> 
> "The Peer-to-Peer (P2P) Session Initiation Protocol working group
> (P2PSIP WG) is chartered to develop protocols and mechanisms for the
> use of the Session Initiation Protocol (SIP) in settings where the
> service of establishing and managing sessions is principally handled
> by a collection of intelligent endpoints, rather than centralized
> servers as in client-server SIP as currently deployed."

Works well for me.  Thanks.

From bryan at ethernot.org  Thu Nov  2 11:29:33 2006
From: bryan at ethernot.org (David A. Bryan)
Date: Thu, 2 Nov 2006 11:29:33 -0500
Subject: [p2p-sip] IMPORTANT!!! Updated Charter
In-Reply-To: <20061102161014.GO2068@sbrim-wxp01>
References: <8b2769930610271420n7162e480w7fbac933bc34e5d4@mail.gmail.com>
	<20061028061203.GE6564@sbrim-wxp01>
	<454378A5.8040404@research.panasonic.com>
	<8b2769930611010914kad12dedid442565361c061ff@mail.gmail.com>
	<20061102055850.GJ5340@sbrim-wxp01>
	<454A1639.4010301@research.panasonic.com>
	<20061102161014.GO2068@sbrim-wxp01>
Message-ID: <8b2769930611020829y38a7ce31q2159fb61b30d0a0c@mail.gmail.com>

Good enough for me then. I'll drop it.

Expect an iterated version (the one we will go to BOF with) shortly.

David

On 11/2/06, Scott W Brim <swb at employees.org> wrote:
> Excerpts from Eunsoo Shim on Thu, Nov 02, 2006 11:00:57AM -0500:
> > I guess Scott and I are talking about the same thing.  Simply
> > dropping 'P2PSIP peers' from the first sentence should work for me.
> > The current sentence is:
> >
> > "The Peer-to-Peer (P2P) Session Initiation Protocol working group
> > (P2PSIP WG) is chartered to develop protocols and mechanisms for the
> > use of the Session Initiation Protocol (SIP) in settings where the
> > service of establishing and managing sessions is principally handled
> > by a collection of intelligent endpoints, or P2PSIP peers, rather
> > than centralized servers as in client-server SIP as currently
> > deployed."
> >
> > My proposal is:
> >
> > "The Peer-to-Peer (P2P) Session Initiation Protocol working group
> > (P2PSIP WG) is chartered to develop protocols and mechanisms for the
> > use of the Session Initiation Protocol (SIP) in settings where the
> > service of establishing and managing sessions is principally handled
> > by a collection of intelligent endpoints, rather than centralized
> > servers as in client-server SIP as currently deployed."
>
> Works well for me.  Thanks.
> _______________________________________________
> p2p-sip mailing list
> p2p-sip at cs.columbia.edu
> https://lists.cs.columbia.edu/cucslists/listinfo/p2p-sip
>

From swb at employees.org  Thu Nov  2 00:58:52 2006
From: swb at employees.org (Scott W Brim)
Date: Thu, 2 Nov 2006 06:58:52 +0100
Subject: [p2p-sip] IMPORTANT!!! Updated Charter
In-Reply-To: <8b2769930611010914kad12dedid442565361c061ff@mail.gmail.com>
References: <8b2769930610271420n7162e480w7fbac933bc34e5d4@mail.gmail.com>
	<20061028061203.GE6564@sbrim-wxp01>
	<454378A5.8040404@research.panasonic.com>
	<8b2769930611010914kad12dedid442565361c061ff@mail.gmail.com>
Message-ID: <20061102055850.GJ5340@sbrim-wxp01>

Excerpts from David A. Bryan on Wed, Nov 01, 2006 12:14:28PM -0500:
> I would like to propose changing the second sentence under discussion to read:
> 
> "P2PSIP clients and peers use the resolution
> services of the peers as an alternative to the SIP correspondent
> discovery process of RFC 3263. [The remaining aspects of] session
> management, messaging, and
> presence functions are performed using traditional SIP."
> 
> (new text in [])
> 
> Scott and Eunsoo: does that fix it?

I still wonder a little about "and managing sessions" in the first
paragraph (P2PSIP peers "manage" SIP sessions?) but it's a nit in the
long run.  We can deal with it in 30 seconds in San Diego.  See you.  

Scott (arriving Saturday 6pm via SFO)

From bryan at ethernot.org  Fri Nov  3 16:50:36 2006
From: bryan at ethernot.org (David A. Bryan)
Date: Fri, 3 Nov 2006 16:50:36 -0500
Subject: [p2p-sip] Revised charter for BOF posted
Message-ID: <8b2769930611031350j790951b7m13ba709f94e6333f@mail.gmail.com>

A revised charter has been posted, incorporating the changes discussed
on list, as well as a couple of small typo corrections or word choice
changes pointed out to me.

As before there are diffs in pdf so folks can see exactly what I changed.

This is the version I'll bring to the BOF, so come prepared to discuss
this version and also to hum for it!

(diffs and charter are at http://www.p2psip.org/ietf.php, shortcut to
just the charter is http://www.p2psip.org/charter_061103.php)

Thanks!

David

From jdrosen at cisco.com  Fri Nov  3 17:38:31 2006
From: jdrosen at cisco.com (Jonathan Rosenberg)
Date: Fri, 03 Nov 2006 17:38:31 -0500
Subject: [p2p-sip] comments on draft-willis-p2psip-concepts-03
Message-ID: <454BC4E7.5070408@cisco.com>

Overall, this document looks pretty decent to me. My comments below come
mostly from an 'outsider' to p2psip, in that I have not been actively
following the list.



> This document defines concepts and terminology for use of the Session
>  Initiation Protocol in a peer-to-peer environment where the
> 
> 
> 
> Bryan, et al.            Expires April 25, 2007                 [Page
> 1]  Internet-Draft       P2PSIP Concepts and Terminology
> October 2006
> 
> 
> traditional proxy-registrar function is replaced by a distributed 
> mechanism that might be implemented using a distributed hash table or
>  other distributed data mechanism with similar external properties.

THe abstract makes it pretty clear that p2psip is meant to just handle a
distributed registrar, rather than a mechanism for finding other SIP
resources (most notably, media relays/TURN servers). Generally I found
the document unclear about whether discovery of those kinds of resources
is in scope.

> The authors believe that this goal is roughly consistent with that of
> a Protocol Model as defined in [13].

This is perhaps a nit; but I've viewed rfc4101 as asking for short
overview sections as part of the spec. This doc is not short and since
there is not even a protocol, not really an overview.

> P2PSIP:  A suite of communications protocols related to the Session 
> Initiation Protocol (SIP) [3] that extends SIP by using peer-to- peer
> techniques for resolving the targets of SIP requests.  The

OK - so not discovering relays then?

> A notable property of P2PSIP Peers is that they can be fully
> functional peers while residing behind NATs.  Contrast this with some
> other
> 
> 
> 
> Bryan, et al.            Expires April 25, 2007                 [Page
> 6]  Internet-Draft       P2PSIP Concepts and Terminology
> October 2006
> 
> 
> P2P systems which make a distinction between "peer" and "super- peer"
> (or "node" and "super-node") where the latter must have a public IP
> address.

This is a bold claim. Seems like you are dismissing the notion of
supernodes as part of the design. Is that purposeful? Getting nat
traversal to work without any kind of bootstramp element is very hard.

> P2PSIP Resource-ID:  A non-human friendly value that uniquely 
> determines which P2PSIP Peer is responsible for storing information
> about this resource (user).  In a DHT approach, this is a numeric
> value in the hash space resulting from hashing the P2PSIP Resource
> Name.  Since Resource-ID is in the same space as the P2PSIP Peer-ID,
> it allows for a mapping between the values, used to map a P2PSIP
> Resource to the P2PSIP Peer that stores it.



Its association with a node in the overlay is a consequence of the DHT
algorithm, not an inherent property of the identifier itself. The
identifier identifies the resource, not the node on which it is stored.


>  P2PSIP Bootstrap Server:  A network node used by other nodes which
>       are attempting to contact a P2PSIP Bootstrap Peer.  It may return
>       the address of a P2PSIP Bootstrap Peer, or act as a proxy to relay
>       messages to and from the Bootstrap Peer, or act as a Bootstrap
>       Peer itself.  The P2PSIP Bootstrap Peer itself should be a fairly
>       stable and well-known host.  

I think you mean Bootstrap Server in this last sentence.

>  P2PSIP Peer Insertion:  The act of inserting a P2PSIP Peer into the
>       current routing structure (presumably a distributed database or
>       hash table) of a P2PSIP Overlay.  For example, the routing
>       structure map the peer's IP address or DNS name to the peer's
>       P2PSIP Peer-ID.

This last sentence doesn't parse.

>   2.  Media relay: a P2PSIP peer or client capable of relaying media
>        sessions.  For example, an RTP relay as described in [14]

This is the bit that made me think you could use the mesh to find TURN 
servers. If not, this needs to be clarified.

> Gateway: a P2PSIP peer or client that converts from P2PSIP to
>        some other protocol, such as PSTN (Public Switched Telephone
>        Network).

How would this work? Would you use the p2p mesh to find gateway resources?

> In a solution where the P2PSIP Client
>    Protocol is SIP, such as is proposed in [7], there may no difference
>    between UA client "C" and standard SIP UA "A".

This contradicts your terminology, where you say that a P2PSIP UA is 
distinguished in that it knows about p2p operations. It seems to me that 
even if the client protocol is sip, you'd still need to do something 
different with regard to bootstrapping...

> Both the "Proxy Peers" and "Redirect Peers" can serve as adapters
>    between ordinary SIP devices and the the P2PSIP Overlay.  Each
>    accepts standard SIP requests and resolves the next-hop by using the
>    P2PSIP overlay Peer Protocol to interact with the routing knowledge
>    of the P2PSIP Overlay, then processes the SIP requests as appropriate
>    (proxying or redirecting towards the next-hop).  Note that proxy
>    operation is bidirectional - the proxy may be forwarding a request
>    from an ordinary SIP device to the P2PSIP overlay, or from the P2PSIP
>    overlay to an ordinary SIP device.

What about requests between p2psip peers? What if two peers can't 
directly reach each other due to nat. Would a proxy peer serve the role 
of connecting them together?

> The Gateway Peer provides a similar sort of adaptation to and from
>    the public switched telephone network (PSTN).  However, there is a
>    subtle distinction.  The gateway function itself can be viewed as a
>    "user" or within the P2PSIP overlay, and is addressed using a P2PSIP
>    Overlay User Identifier.  

The latter is never defined by your terminology. Do you mean P2PSIP 
Resource-ID? I have a hard time understanding how this is going to work. 
PSTN gateways don't have a (useful) unique identifier. Rather, when I 
want to mkae a call to a number, any gateway might potentially work. 
What resource name would I use when trying to find the gateway? The 
problem is that gateway selection is not easily modeled as a hash table 
lookup operation, so I have a hard time understanding how it fits in.

I also think it would help a lot to spend some time on how SIP 
interworks with p2psip. Can a regular sip client call a p2p-sip client? 
Would this require that a overlay ID map 1-1 with a domain name (I think 
so)?

>  o  Enrolling a Peer in the Overlay
> 
>    o  Inserting a P2PSIP Peer into the Overlay
> 
>    o  Enrolling a Resource or User in the Overlay
> 
>    o  Inserting a Resource or User into the Overlay
> 
>    o  Retrieve information about a Resource or User from the Overlay

A notable omission here is "Reach a resource or user from the overlay". 
Retrieving information (namely its IP) would seem to imply that it is 
always directly reachable and we know that is not going to be true. What 
operation is used in that case? Or if it is a retrieval, would you 
retrieve a p2psip proxy to which that user has a connection/flow?

> Finally, it should be re-iterated that the Client Protocol is a
>    strict subset of the Peer Protocol (at least logically, the two may
>    differ syntactically).  This has two consequences:

Actually its been reiterated like three times already....

> 3.4.3.  Conventional SIP Device using a Proxy Peer

This flow here makes it seem like a proxy peer is only used by normal 
sip UA and not peers. Per above I don't see how this is working on the 
presence of nat.

> 3.5.2.  More on The Difference Between a Peer, Client, and User Agent
> 
>    P2PSIP Peers directly interact with and contain the routing and
>    storage fabric of the overlay.  P2PSIP Clients simply use the routing
>    and storage facilities provided by the peers to get and put
>    information.  The peers speak the P2PSIP Peer Protocol, which can
>    express all the necessary routing and storage operations required for
>    the overlay.  Clients speak the P2PSIP Client protocol, which is
>    presumably a subset of the peer protocol, and is limited to storage
>    insertion (put) and storage retrieval (get).  Some designs do not
>    require a separate client protocol.

very repetitive. You've said this several times now.

>  Presumably, the P2PSIP UA connects to a
>    peer or client and uses the user's credentials to authenticate a
>    route-set (Contact: plus Path:) to itself, and the peer or client
>    stores the route-set into the overlay, using a key derived from the
>    user's identity.

Authenticate a route set or insert a route set? This is confusing. 
Insertion of a route-set would seem to imply that you ARE using peers as 
proxies when going from one peer to another. This contradicts text and 
figures above.

> In the typical case, the peer or client coming up is also a P2PSIP
>    User Agent with one or more associated P2PSIP Resource (User)
>    Identifiers.  The next step then is to insert a P2PSIP Resource
>    Record (a Contact:) into the P2PSIP Overlay.

Now, here its a Contact and not a Path/Contact.

> 4.3.  How Do We Find Gateways?
> 
>    This needs to be not only netpath efficient, but also embodies
>    elements of the TRIP and SPEERMINT problems.

Right - this is my point. I don't see how this works.

> We presume that the question of how to send media between two peers
>    or clients is already solved by the STUN/TURN/ICE suite of protocols,
>    assuming that the necessary STUN servers and media relays can be
>    located.

:)

> 4.13.  Admissions Control
> 
>    What do we need to say about admissions control with respect to the
>    enrollment of peers and users?  Do we need to discuss per-call
>    admissions control in a P2P environment?

terminology wise, I'd suggest not to use 'admission control' since it 
has well-defined meaning and a LOT of baggage already.

>  This model presumes that all addressable elements, aka "users", are
>    unique.  Are their other classes of resources that need some sort of
>    class-addressable identifier that does not refer to a unique user?

This is exactly the problem with gateways and turn servers.


Thanks,
Jonathan R.
-- 
Jonathan D. Rosenberg, Ph.D.                   600 Lanidex Plaza
Cisco Fellow                                   Parsippany, NJ 07054-2711
Cisco Systems
jdrosen at cisco.com                              FAX:   (973) 952-5050
http://www.jdrosen.net                         PHONE: (973) 952-5000
http://www.cisco.com


From dean.willis at softarmor.com  Sun Nov  5 11:27:58 2006
From: dean.willis at softarmor.com (Dean Willis)
Date: Sun, 5 Nov 2006 10:27:58 -0600
Subject: [p2p-sip] Comments on draft-willis-p2psip-concepts-03
In-Reply-To: <BAYC1-PASMTP0260F9D9DF718C9309C0B4CCFF0@CEZ.ICE>
References: <20061031015443.7CEB45C01E@laser.networkresonance.com><6F914E17-0377-4F4A-B794-9DFF64E1A2BD@softarmor.com><868xiw35kg.fsf@raman.networkresonance.com><13F6728B-6CD3-4F23-BBD0-DB5E01067E53@softarmor.com><86lkmw1h7p.fsf@raman.networkresonance.com><8b2769930610311145w4089f2datac58a32384e8291c@mail.gmail.com><86lkmwz2fy.fsf@raman.networkresonance.com><8b2769930610311217h29bd1a37wb1c0f1566f271414@mail.gmail.com>
	<4F21983F-69D1-4311-8D3F-492646181083@softarmor.com>
	<BAYC1-PASMTP108AFDF599E303646E37C1CCF80@CEZ.ICE>
	<57A90420-7FDE-4E69-9B24-7F26DB12E786@softarmor.com>
	<BAYC1-PASMTP0260F9D9DF718C9309C0B4CCFF0@CEZ.ICE>
Message-ID: <07831F27-918D-463C-8659-BBDB50032DF5@softarmor.com>


On Nov 2, 2006, at 8:32 AM, Eric Cooper wrote:

>>>
>>
>> Well, what part of the peer protocol is it that allows peers behind
>> NATs to function?
>>
>
> As an example, there could be some messaging involving a bootstrap  
> server that would request that a bootstrap peer initiate UDP hole  
> punching with a new peer (one that was performing it's insertion  
> into the overlay).
>
>> It's not like we can just turn on a magic "NAT Traversal Bit" in the
>> protocol.
>>
>> The only way I know of to reliably talk to anything behind a NAT is
>> to enlist the help of somebody that is already talking to the node-
>> behind-the-nat via a connection established from behind-the-nat.
>>
>
> In the scenario I'm thinking of, this would be a connection from a  
> (new) peer, talking to a bootstrap server.
> Suppose there's also a separate connection between a bootstrap peer  
> that has already joined the overlay and that same bootstrap  
> server.  The existing peer might even be behind a NAT.  The  
> bootstrap server could coordinate UDP hole punching between the  
> existing peer and the new peer.
>

ok, now you have an "outbound" connection between the peer and the  
"bootstrap server", which I presume is some sort of centralized non- 
peer.  And you might have a new bootstrap protocol. No reason this  
couldn't work, and it's exact analog of the model I was explaining,  
just based on a centralized server doing the heavy lifting instead of  
having a peer do it.



>> So for something "outside" to talk to a peer that is behind a NAT, we
>> need to find another peer (or some other helper function) that is
>> outside the NAT and already has a connection to the peer-behind the
>> NAT. If this helper thing is a peer, then I'd say that there's an
>> "outbound peer" relationship between the peer-beind-the-nat and the
>> peer-outside-the-nat-that-is-helping.
>>
>
> If the UDP hole punching succeeds, then a 'link' between the two  
> peers (both behind NATs) is established.  Would they both be  
> 'outbound peers'?
>

Possibly.

> I was just wondering if you thought the 'outbound peer' behaviour  
> was optional, or just part of the regular responsibilities of a  
> peer, like storage and routing.

Good question. Possibly something that needs to be tracked as an open  
question for the WG. Personally, I think it is mandatory behavior for  
a "peer" but I suspect some of my co-authors might disagree. If it  
isn't mandatory, we'll need a locator mechanism to help us find a  
peer (or other helper function, like your bootsrap server) that  
provides the needed service.

This gets back to a fundamental "design pattern" question that I  
don't think most people are getting.

Either every peer is "full function" and supports every aspect of  
being a peer, OR we need to define a locator mechanism for finding  
peers that support specific aspects (or features).

--
Dean



From dean.willis at softarmor.com  Sun Nov  5 11:52:54 2006
From: dean.willis at softarmor.com (Dean Willis)
Date: Sun, 5 Nov 2006 10:52:54 -0600
Subject: [p2p-sip] comments on draft-willis-p2psip-concepts-03
In-Reply-To: <454BC4E7.5070408@cisco.com>
References: <454BC4E7.5070408@cisco.com>
Message-ID: <9EEAE2ED-D618-4DD5-A5A1-6E039F3DBB4A@softarmor.com>


On Nov 3, 2006, at 4:38 PM, Jonathan Rosenberg wrote:

> Overall, this document looks pretty decent to me. My comments below  
> come
> mostly from an 'outsider' to p2psip, in that I have not been actively
> following the list.
>
>
>
>> This document defines concepts and terminology for use of the Session
>>  Initiation Protocol in a peer-to-peer environment where the
>>
>>
>>
>> Bryan, et al.            Expires April 25, 2007                 [Page
>> 1]  Internet-Draft       P2PSIP Concepts and Terminology
>> October 2006
>>
>>
>> traditional proxy-registrar function is replaced by a distributed
>> mechanism that might be implemented using a distributed hash table or
>>  other distributed data mechanism with similar external properties.
>
> THe abstract makes it pretty clear that p2psip is meant to just  
> handle a
> distributed registrar, rather than a mechanism for finding other SIP
> resources (most notably, media relays/TURN servers). Generally I found
> the document unclear about whether discovery of those kinds of  
> resources
> is in scope.
>

That's a true statement. It is unclear about searching for  
differentiated resources. We need to make a decision -- either all  
peers support all resource functions, or we need to have a search  
mechanism.

>> P2PSIP:  A suite of communications protocols related to the Session
>> Initiation Protocol (SIP) [3] that extends SIP by using peer-to- peer
>> techniques for resolving the targets of SIP requests.  The
>
> OK - so not discovering relays then?
>

Open question.

>> A notable property of P2PSIP Peers is that they can be fully
>> functional peers while residing behind NATs.  Contrast this with some
>> other
>> P2P systems which make a distinction between "peer" and "super- peer"
>> (or "node" and "super-node") where the latter must have a public IP
>> address.
>
> This is a bold claim. Seems like you are dismissing the notion of
> supernodes as part of the design. Is that purposeful? Getting nat
> traversal to work without any kind of bootstramp element is very hard.
>

No, we're just saying that not ALL supernodes require public IP  
addresses. This position was established after months of debate.

>> P2PSIP Resource-ID:  A non-human friendly value that uniquely
>> determines which P2PSIP Peer is responsible for storing information
>> about this resource (user).  In a DHT approach, this is a numeric
>> value in the hash space resulting from hashing the P2PSIP Resource
>> Name.  Since Resource-ID is in the same space as the P2PSIP Peer-ID,
>> it allows for a mapping between the values, used to map a P2PSIP
>> Resource to the P2PSIP Peer that stores it.
>
>
> Its association with a node in the overlay is a consequence of the DHT
> algorithm, not an inherent property of the identifier itself. The
> identifier identifies the resource, not the node on which it is  
> stored.
>

Yeah, this def once said something like that, I think, and was  
subsumed by a DHT-centric definition.

>
>>  P2PSIP Bootstrap Server:  A network node used by other nodes which
>>       are attempting to contact a P2PSIP Bootstrap Peer.  It may  
>> return
>>       the address of a P2PSIP Bootstrap Peer, or act as a proxy to  
>> relay
>>       messages to and from the Bootstrap Peer, or act as a Bootstrap
>>       Peer itself.  The P2PSIP Bootstrap Peer itself should be a  
>> fairly
>>       stable and well-known host.
>
> I think you mean Bootstrap Server in this last sentence.

I agree.
>
>>  P2PSIP Peer Insertion:  The act of inserting a P2PSIP Peer into the
>>       current routing structure (presumably a distributed database or
>>       hash table) of a P2PSIP Overlay.  For example, the routing
>>       structure map the peer's IP address or DNS name to the peer's
>>       P2PSIP Peer-ID.
>
> This last sentence doesn't parse.
>

The noun "map" should be replaced with the verb "maps".

>>   2.  Media relay: a P2PSIP peer or client capable of relaying media
>>        sessions.  For example, an RTP relay as described in [14]
>
> This is the bit that made me think you could use the mesh to find TURN
> servers. If not, this needs to be clarified.
>

Indeed.

>> Gateway: a P2PSIP peer or client that converts from P2PSIP to
>>        some other protocol, such as PSTN (Public Switched Telephone
>>        Network).
>
> How would this work? Would you use the p2p mesh to find gateway  
> resources?
>

Indirectly. You'd need some telephony-routing mechanism to decide  
which gateway to use, and then use P2PSIP to find that gateway.

Of course, if we end up changing the model to provide differentiated  
resource searching, I suppose one could do telephony routing that way  
as well.

>> In a solution where the P2PSIP Client
>>    Protocol is SIP, such as is proposed in [7], there may no  
>> difference
>>    between UA client "C" and standard SIP UA "A".
>
> This contradicts your terminology, where you say that a P2PSIP UA is
> distinguished in that it knows about p2p operations. It seems to me  
> that
> even if the client protocol is sip, you'd still need to do something
> different with regard to bootstrapping...

Yeah, we had extended discussion on the above, and it's poorly  
worded. One might build a P2PSIP architecture that doesn't have a  
client protocol or clients as defined in this document. Just peers  
and plain old SIP UAs.



>
>> Both the "Proxy Peers" and "Redirect Peers" can serve as adapters
>>    between ordinary SIP devices and the the P2PSIP Overlay.  Each
>>    accepts standard SIP requests and resolves the next-hop by  
>> using the
>>    P2PSIP overlay Peer Protocol to interact with the routing  
>> knowledge
>>    of the P2PSIP Overlay, then processes the SIP requests as  
>> appropriate
>>    (proxying or redirecting towards the next-hop).  Note that proxy
>>    operation is bidirectional - the proxy may be forwarding a request
>>    from an ordinary SIP device to the P2PSIP overlay, or from the  
>> P2PSIP
>>    overlay to an ordinary SIP device.
>
> What about requests between p2psip peers? What if two peers can't
> directly reach each other due to nat. Would a proxy peer serve the  
> role
> of connecting them together?

Not normally. Peers route P2PSIP Peer Protocol request on behalf of  
other peers. Proxy Peers route SIP requests. They're different.

>
>> The Gateway Peer provides a similar sort of adaptation to and from
>>    the public switched telephone network (PSTN).  However, there is a
>>    subtle distinction.  The gateway function itself can be viewed  
>> as a
>>    "user" or within the P2PSIP overlay, and is addressed using a  
>> P2PSIP
>>    Overlay User Identifier.
>
> The latter is never defined by your terminology. Do you mean P2PSIP
> Resource-ID? I have a hard time understanding how this is going to  
> work.
> PSTN gateways don't have a (useful) unique identifier. Rather, when I
> want to mkae a call to a number, any gateway might potentially work.
> What resource name would I use when trying to find the gateway? The
> problem is that gateway selection is not easily modeled as a hash  
> table
> lookup operation, so I have a hard time understanding how it fits in.
>

One model would be to use TRIP to find a URI for a gateway servicing  
that phone number, then use P2PSIP to contact the gateway.

> I also think it would help a lot to spend some time on how SIP
> interworks with p2psip. Can a regular sip client call a p2p-sip  
> client?
>

Yes.

> Would this require that a overlay ID map 1-1 with a domain name (I  
> think
> so)?

While not required, such a mapping is one of our assumptions.


>
>>  o  Enrolling a Peer in the Overlay
>>
>>    o  Inserting a P2PSIP Peer into the Overlay
>>
>>    o  Enrolling a Resource or User in the Overlay
>>
>>    o  Inserting a Resource or User into the Overlay
>>
>>    o  Retrieve information about a Resource or User from the Overlay
>
> A notable omission here is "Reach a resource or user from the  
> overlay".
> Retrieving information (namely its IP) would seem to imply that it is
> always directly reachable and we know that is not going to be true.  
> What
> operation is used in that case? Or if it is a retrieval, would you
> retrieve a p2psip proxy to which that user has a connection/flow?
>

The information that is retrieved could be a route-vector instead of  
just an IP address.


>> Finally, it should be re-iterated that the Client Protocol is a
>>    strict subset of the Peer Protocol (at least logically, the two  
>> may
>>    differ syntactically).  This has two consequences:
>
> Actually its been reiterated like three times already....
>

Yeah, we're trying the "say it again in different words, louder,  
'cause they just ain't getting it" trick.

>> 3.4.3.  Conventional SIP Device using a Proxy Peer
>
> This flow here makes it seem like a proxy peer is only used by normal
> sip UA and not peers. Per above I don't see how this is working on the
> presence of nat.

One uses "outbound" between the proxy-peer and the SIP-UA.

>
>> 3.5.2.  More on The Difference Between a Peer, Client, and User Agent
>>
>>    P2PSIP Peers directly interact with and contain the routing and
>>    storage fabric of the overlay.  P2PSIP Clients simply use the  
>> routing
>>    and storage facilities provided by the peers to get and put
>>    information.  The peers speak the P2PSIP Peer Protocol, which can
>>    express all the necessary routing and storage operations  
>> required for
>>    the overlay.  Clients speak the P2PSIP Client protocol, which is
>>    presumably a subset of the peer protocol, and is limited to  
>> storage
>>    insertion (put) and storage retrieval (get).  Some designs do not
>>    require a separate client protocol.
>
> very repetitive. You've said this several times now.
>

yeah, we added repetition every time somebody seemed to think it was  
needed for comprehension.

>>  Presumably, the P2PSIP UA connects to a
>>    peer or client and uses the user's credentials to authenticate a
>>    route-set (Contact: plus Path:) to itself, and the peer or client
>>    stores the route-set into the overlay, using a key derived from  
>> the
>>    user's identity.
>
> Authenticate a route set or insert a route set? This is confusing.
> Insertion of a route-set would seem to imply that you ARE using  
> peers as
> proxies when going from one peer to another. This contradicts text and
> figures above.
>
>> In the typical case, the peer or client coming up is also a P2PSIP
>>    User Agent with one or more associated P2PSIP Resource (User)
>>    Identifiers.  The next step then is to insert a P2PSIP Resource
>>    Record (a Contact:) into the P2PSIP Overlay.
>
> Now, here its a Contact and not a Path/Contact.

Should be a "route vector" using my preferred vocabulary.


>
>> 4.3.  How Do We Find Gateways?
>>
>>    This needs to be not only netpath efficient, but also embodies
>>    elements of the TRIP and SPEERMINT problems.
>
> Right - this is my point. I don't see how this works.

Not our problem, I hope.

>
>> We presume that the question of how to send media between two peers
>>    or clients is already solved by the STUN/TURN/ICE suite of  
>> protocols,
>>    assuming that the necessary STUN servers and media relays can be
>>    located.
>
> :)
>
>> 4.13.  Admissions Control
>>
>>    What do we need to say about admissions control with respect to  
>> the
>>    enrollment of peers and users?  Do we need to discuss per-call
>>    admissions control in a P2P environment?
>
> terminology wise, I'd suggest not to use 'admission control' since it
> has well-defined meaning and a LOT of baggage already.
>

Actually, the "admissions control in a P2P environment" part is  
intended to be consistent with that well-defined meaning. And we're  
hoping the naswer (which you seem to agree with) is "no".

>>  This model presumes that all addressable elements, aka "users", are
>>    unique.  Are their other classes of resources that need some  
>> sort of
>>    class-addressable identifier that does not refer to a unique user?
>
> This is exactly the problem with gateways and turn servers.
>

And that should be "there", not "their". Probably my bad.

--
Dean



From philip_matthews at magma.ca  Sun Nov  5 18:44:28 2006
From: philip_matthews at magma.ca (Philip Matthews)
Date: Sun, 5 Nov 2006 18:44:28 -0500
Subject: [p2p-sip] comments on draft-willis-p2psip-concepts-03
In-Reply-To: <9EEAE2ED-D618-4DD5-A5A1-6E039F3DBB4A@softarmor.com>
References: <454BC4E7.5070408@cisco.com>
	<9EEAE2ED-D618-4DD5-A5A1-6E039F3DBB4A@softarmor.com>
Message-ID: <9AC6105A-F15B-4A8E-84EB-F82BCBB159C3@magma.ca>

Jonathan:

Just adding a bit to Dean's reply to your comments.

Most of my comments are in-line, but first a general comment.
Yes, the authors are aware that the current  version (-03) has
a number of internal inconsistencies and is very repetitive
in some areas.  A lot of this is because three of the four authors
have taken a hand at editing the document, where the last two iterations
were in rapid succession under a tight deadline. When I submitted
-03, I KNEW there were still problems, but I had run out of time.

We hope that the next version of the document will clean
some of this up.


The rest of my comments are inline.


On 5-Nov-06, at 11:52 , Dean Willis wrote:

>
> On Nov 3, 2006, at 4:38 PM, Jonathan Rosenberg wrote:
>
>>
>>
>> THe abstract makes it pretty clear that p2psip is meant to just
>> handle a
>> distributed registrar, rather than a mechanism for finding other SIP
>> resources (most notably, media relays/TURN servers). Generally I  
>> found
>> the document unclear about whether discovery of those kinds of
>> resources
>> is in scope.
>>
>
> That's a true statement. It is unclear about searching for
> differentiated resources. We need to make a decision -- either all
> peers support all resource functions, or we need to have a search
> mechanism.

I second Dean's reply. The draft IS unclear in this area, and we need
to make a decision on differentiated resources.

My own opinion is that only some peers and clients will be willing and
able to act as proxies or STUN servers or media relays or gateways or  
whatever,
so we need a way to find such peers.

>
>>> A notable property of P2PSIP Peers is that they can be fully
>>> functional peers while residing behind NATs.  Contrast this with  
>>> some
>>> other
>>> P2P systems which make a distinction between "peer" and "super-  
>>> peer"
>>> (or "node" and "super-node") where the latter must have a public IP
>>> address.
>>
>> This is a bold claim. Seems like you are dismissing the notion of
>> supernodes as part of the design. Is that purposeful? Getting nat
>> traversal to work without any kind of bootstramp element is very  
>> hard.
>>
>
> No, we're just saying that not ALL supernodes require public IP
> addresses. This position was established after months of debate.

Some people say that:
	P2PSIP peer 	= super-peer in some other P2P networks
	P2PSIP client	= ordinary peer in some other P2P networks
(Personally, I am leery of this, but this might help you.)
If you take this identification, then we are saying that superpeers
can have private IP addresses.

>
>>
>>>  P2PSIP Bootstrap Server:  A network node used by other nodes which
>>>       are attempting to contact a P2PSIP Bootstrap Peer.  It may
>>> return
>>>       the address of a P2PSIP Bootstrap Peer, or act as a proxy to
>>> relay
>>>       messages to and from the Bootstrap Peer, or act as a Bootstrap
>>>       Peer itself.  The P2PSIP Bootstrap Peer itself should be a
>>> fairly
>>>       stable and well-known host.
>>
>> I think you mean Bootstrap Server in this last sentence.
>
> I agree.

Yes. That is a typo on my part.


>
>
>>
>>>  o  Enrolling a Peer in the Overlay
>>>
>>>    o  Inserting a P2PSIP Peer into the Overlay
>>>
>>>    o  Enrolling a Resource or User in the Overlay
>>>
>>>    o  Inserting a Resource or User into the Overlay
>>>
>>>    o  Retrieve information about a Resource or User from the Overlay
>>
>> A notable omission here is "Reach a resource or user from the
>> overlay".
>> Retrieving information (namely its IP) would seem to imply that it is
>> always directly reachable and we know that is not going to be true.
>> What
>> operation is used in that case? Or if it is a retrieval, would you
>> retrieve a p2psip proxy to which that user has a connection/flow?
>>
>
> The information that is retrieved could be a route-vector instead of
> just an IP address.

The current version of the document mostly ignores the fact that
some peers may not be directly reachable because they are behind
NATs. There is a comment to this effect in  the examples section,
but the problem is pervasive throughout the document.
It is something we need to fix in the next version.



>>
>>> We presume that the question of how to send media between two peers
>>>    or clients is already solved by the STUN/TURN/ICE suite of
>>> protocols,
>>>    assuming that the necessary STUN servers and media relays can be
>>>    located.
>>
>> :)

Indeed. Thanks for your hard work on STUN/TURN/ICE.

- Philip


From salman at cs.columbia.edu  Mon Nov  6 13:30:55 2006
From: salman at cs.columbia.edu (Salman Abdul Baset)
Date: Mon, 6 Nov 2006 13:30:55 -0500 (EST)
Subject: [p2p-sip] draft-01: A protocol for implementing various DHT
	algorithms
Message-ID: <Pine.GSO.4.58.0611061320340.15545@flame.cs.columbia.edu>

We are in the process of developing a protocol for implementing various
DHT (p2p) algorithms. A snapshot of the current effort can be found at:

http://www1.cs.columbia.edu/~salman/drafts/draft-baset-sipping-p2pcommon-01.html
http://www1.cs.columbia.edu/~salman/drafts/draft-baset-sipping-p2pcommon-01.txt

Section 6 defines the messages and their fields and section 7 defines the
PDUs for these messages.

The draft-01 is still not complete and there might be [non-]obvious bugs.
However, I thought it will be useful to share it with the p2p-sip
community gathered at the IETF.

Regards,
Salman

From enrico.marocco at telecomitalia.it  Mon Nov  6 13:33:54 2006
From: enrico.marocco at telecomitalia.it (Marocco Enrico)
Date: Mon, 6 Nov 2006 19:33:54 +0100
Subject: [p2p-sip] comments on draft-willis-p2psip-concepts-03
References: <454BC4E7.5070408@cisco.com>
Message-ID: <9E577E61D0108D4F8CD482029B613F5C90E0C1@PTPEVS108BA020.idc.cww.telecomitalia.it>

Jonathan Rosenberg wrote:
> I also think it would help a lot to spend some time on how SIP 
> interworks with p2psip. Can a regular sip client call a p2p-sip client? 
> Would this require that a overlay ID map 1-1 with a domain name (I think 
> so)?

David and I have started thinking about it and and have proposed a
possible approach (draft-marocco-p2psip-interwork-00) which also deals
with discovery of relay resources.  However such a work is probably
worth doing after we have a WG formed, so no much attention has been
actually spent on it.

--
Ciao,
Enrico

Sorry..  the notice below is not my fault :-(
--------------------------------------------------------------------

CONFIDENTIALITY NOTICE

This message and its attachments are addressed solely to the persons above and may contain confidential information. If you have received the message in error, be informed that any use of the content hereof is prohibited. Please return it immediately to the sender and delete the message. Should you have any questions, please contact us by replying to webmaster at telecomitalia.it.

        Thank you

                                        www.telecomitalia.it

--------------------------------------------------------------------
                        


From rbarnes at bbn.com  Fri Nov 10 01:50:04 2006
From: rbarnes at bbn.com (Richard Barnes)
Date: Thu, 09 Nov 2006 22:50:04 -0800
Subject: [p2p-sip] Testing, please ignore
Message-ID: <4554211C.6040801@bbn.com>

(no body)


From rbarnes at bbn.com  Fri Nov 10 01:50:44 2006
From: rbarnes at bbn.com (Richard L. Barnes)
Date: Thu, 09 Nov 2006 22:50:44 -0800
Subject: [p2p-sip] Testing, please ignore
Message-ID: <45542144.6020209@bbn.com>



From jean-francois.wauthy at alcatel.fr  Fri Nov 10 10:30:09 2006
From: jean-francois.wauthy at alcatel.fr (=?ISO-8859-1?Q?Jean-Fran=E7ois_Wauthy?=)
Date: Fri, 10 Nov 2006 16:30:09 +0100
Subject: [p2p-sip] draft-bryan-sipping-p2p-03 and Chord's notify procedure
Message-ID: <45549B01.3030208@alcatel.fr>

Hi,

In subsection 10.8 of draft-bryan-sipping-p2p-03 describing the periodic 
stabilization, it is required to send a REGISTER request structured as 
if the stabilizing peer had just entered the system when the peer 
updates its successor so that that successor can eventually update its 
predecessor. IMO, this can't work because the successor will eventually, 
even almost every time in fact, route the REGISTER request to another 
peer in the overlay (may be the stabilizing peer itself).

This can be seen illustrated with the example of subsection 13.4, when 
peer 3 runs its periodic stabilization it sends a REGISTER request to 
peer 10, structured as if peer 3 had just entered the system. It is then 
said that:
"When Peer 10 receives the message, it then sends a 200 OK response to 
Peer 3."

How did peer 10 figure out that it handles registration of peer 3 ? 
According to the successor, predecessor and finger table information I 
think it should route the REGISTER request to peer 3 itself (as it 
doesn't have a predecessor it can't determine the interval it covers and 
thus must rely on its finger table to route the message).

Have i missed something ?

BTW i found a little repetition in subsection 10.2, the sentence : "The 
peer MUST initialize its finger table such that all entries point to 
itself." is useless as it is repeated in the next sentence.

--
Jean-Fran?ois Wauthy

From marcia at sipeerior.com  Fri Nov 10 15:33:47 2006
From: marcia at sipeerior.com (Marcia Z Bryan)
Date: Fri, 10 Nov 2006 15:33:47 -0500
Subject: [p2p-sip] draft-bryan-sipping-p2p-03 and Chord's notify
	procedure
In-Reply-To: <45549B01.3030208@alcatel.fr>
References: <45549B01.3030208@alcatel.fr>
Message-ID: <9e298a4a0611101233v2fee0efdx9bbde2055c66b459@mail.gmail.com>

On 11/10/06, Jean-Fran?ois Wauthy <jean-francois.wauthy at alcatel.fr> wrote:
>
> Hi,
>
> In subsection 10.8 of draft-bryan-sipping-p2p-03 describing the periodic
> stabilization, it is required to send a REGISTER request structured as
> if the stabilizing peer had just entered the system when the peer
> updates its successor so that that successor can eventually update its
> predecessor. IMO, this can't work because the successor will eventually,
> even almost every time in fact, route the REGISTER request to another
> peer in the overlay (may be the stabilizing peer itself).
>
> This can be seen illustrated with the example of subsection 13.4, when
> peer 3 runs its periodic stabilization it sends a REGISTER request to
> peer 10, structured as if peer 3 had just entered the system.


Peer 3, the stabilizing peer, sends a Peer Registration (see section 9.2.2.1)
message to Peer 10.


It is then
> said that:
> "When Peer 10 receives the message, it then sends a 200 OK response to
> Peer 3."
>
> How did peer 10 figure out that it handles registration of peer 3 ?


Peer 10 handles the message from Peer 3 as if Peer 10 is the admitting peer
(sections 9.2.2.2 and 10.4).  From 10.4 in the draft:

"...the admitting peer MUST reply with a 200 response if the joining peer
has a Peer-ID between the admitting peer's predecessor's Peer-ID and the
admitting Peer-ID, or the admitting peer's predcessor is NULL."

Peer 10 is checking to see if Peer 3 should be its predecessor --- basically
checking to see if Peer 3 is in  the interval [pred, Peer 10).  Because Peer
10 doesn't have a predecessor (pred = NULL), it sets its predecessor to Peer
3.

I believe it is still an open issue as to what the stabilizing peer, Peer 3
in this example, does if the response is not a 200.

does this help?

--Marcia

According to the successor, predecessor and finger table information I
> think it should route the REGISTER request to peer 3 itself (as it
> doesn't have a predecessor it can't determine the interval it covers and
> thus must rely on its finger table to route the message).
>
> Have i missed something ?
>
> BTW i found a little repetition in subsection 10.2, the sentence : "The
> peer MUST initialize its finger table such that all entries point to
> itself." is useless as it is repeated in the next sentence.
>
> --
> Jean-Fran?ois Wauthy
> _______________________________________________
> p2p-sip mailing list
> p2p-sip at cs.columbia.edu
> https://lists.cs.columbia.edu/cucslists/listinfo/p2p-sip
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: https://lists.cs.columbia.edu/pipermail/p2p-sip/attachments/20061110/16d3bd4e/attachment.html

From jean-francois.wauthy at alcatel.fr  Mon Nov 13 03:25:44 2006
From: jean-francois.wauthy at alcatel.fr (=?ISO-8859-1?Q?Jean-Fran=E7ois_Wauthy?=)
Date: Mon, 13 Nov 2006 09:25:44 +0100
Subject: [p2p-sip] draft-bryan-sipping-p2p-03 and Chord's notify
	procedure
In-Reply-To: <9e298a4a0611101233v2fee0efdx9bbde2055c66b459@mail.gmail.com>
References: <45549B01.3030208@alcatel.fr>
	<9e298a4a0611101233v2fee0efdx9bbde2055c66b459@mail.gmail.com>
Message-ID: <45582C08.8040401@alcatel.fr>

Marcia Z Bryan a ?crit :
> On 11/10/06, *Jean-Fran?ois Wauthy* <jean-francois.wauthy at alcatel.fr 
> <mailto:jean-francois.wauthy at alcatel.fr>> wrote:
>
> Peer 10 handles the message from Peer 3 as if Peer 10 is the admitting 
> peer (sections 9.2.2.2 <http://9.2.2.2> and 10.4).  From 10.4 in the 
> draft:
>
> "...the admitting peer MUST reply with a 200 response if the joining 
> peer has a Peer-ID between the admitting peer's predecessor's Peer-ID 
> and the admitting Peer-ID, or the admitting peer's predcessor is NULL."
>
Hmm, somehow i skipped that part. I understand now how peer 10 knows it 
is the admitting node for the request and must process it.
> Peer 10 is checking to see if Peer 3 should be its predecessor --- 
> basically checking to see if Peer 3 is in  the interval [pred, Peer 
> 10).  Because Peer 10 doesn't have a predecessor (pred = NULL), it 
> sets its predecessor to Peer 3. 
Shouldn't the interval be (pred, Peer 10] ? AFAICT the definition of 
that interval isn't clearly defined in the draft and may also cause some 
confusion.
> I believe it is still an open issue as to what the stabilizing peer, 
> Peer 3 in this example, does if the response is not a 200.
OK
> does this help?
Yes, thanks for the explanations.

-- 
Jean-Fran?ois Wauthy


From billmcc64 at gmail.com  Wed Nov 15 15:03:27 2006
From: billmcc64 at gmail.com (Bill Mccormick)
Date: Wed, 15 Nov 2006 15:03:27 -0500
Subject: [p2p-sip] Status of P2P SIP - did we form a WG at the IETF last
	week?
Message-ID: <fb4306550611151203g51343762w1d93cc08ba33a471@mail.gmail.com>

Hi everyone,

I couldn't attend... how did it go last week?

Bill

-- 
Bill McCormick

From eunsooshim at gmail.com  Wed Nov 15 21:57:31 2006
From: eunsooshim at gmail.com (Eunsoo Shim)
Date: Wed, 15 Nov 2006 21:57:31 -0500
Subject: [p2p-sip] Status of P2P SIP - did we form a WG at the IETF last
	week?
In-Reply-To: <fb4306550611151203g51343762w1d93cc08ba33a471@mail.gmail.com>
References: <fb4306550611151203g51343762w1d93cc08ba33a471@mail.gmail.com>
Message-ID: <7b683f3f0611151857u1276e371q614d72be9c5071c0@mail.gmail.com>

I believe the BOF went well.
Most people are confident that the WG creation will be approved even though
there are some procedures left before we hear the announcement.
According to our AD, Cullen, it will take one or two months to go through
the procedures.
We should also expect some minor editing of the charter to reflect the
comments at the BOF.
Thanks.

Eunsoo

On 11/15/06, Bill Mccormick <billmcc64 at gmail.com> wrote:
>
> Hi everyone,
>
> I couldn't attend... how did it go last week?
>
> Bill
>
> --
> Bill McCormick
> _______________________________________________
> p2p-sip mailing list
> p2p-sip at cs.columbia.edu
> https://lists.cs.columbia.edu/cucslists/listinfo/p2p-sip
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: https://lists.cs.columbia.edu/pipermail/p2p-sip/attachments/20061115/27d1d8a2/attachment.html

From philip_matthews at magma.ca  Thu Nov 16 14:16:03 2006
From: philip_matthews at magma.ca (Philip Matthews)
Date: Thu, 16 Nov 2006 14:16:03 -0500
Subject: [p2p-sip] Status of P2P SIP - did we form a WG at the IETF last
	week?
In-Reply-To: <7b683f3f0611151857u1276e371q614d72be9c5071c0@mail.gmail.com>
References: <fb4306550611151203g51343762w1d93cc08ba33a471@mail.gmail.com>
	<7b683f3f0611151857u1276e371q614d72be9c5071c0@mail.gmail.com>
Message-ID: <2CAD4BA2-033E-4DBA-8CEA-E73921550DF2@magma.ca>

I second  Eunsoo's comment that the BOF went very smoothly.

In fact, I heard that a number of people commented that
"This is the most boring BOF I have ever attended".
Believe it or not, boring is very good. It means that everyone agrees.
Exciting BOFs make good theater, but do not
necessarily lead to good working groups.


- Philip

On 15-Nov-06, at 21:57 , Eunsoo Shim wrote:

> I believe the BOF went well.
> Most people are confident that the WG creation will be approved  
> even though there are some procedures left before we hear the  
> announcement.
> According to our AD, Cullen, it will take one or two months to go  
> through the procedures.
> We should also expect some minor editing of the charter to reflect  
> the comments at the BOF.
> Thanks.
>
> Eunsoo
>
> On 11/15/06, Bill Mccormick < billmcc64 at gmail.com> wrote:
> Hi everyone,
>
> I couldn't attend... how did it go last week?
>
> Bill
>
> --
> Bill McCormick
> _______________________________________________
> p2p-sip mailing list
> p2p-sip at cs.columbia.edu
> https://lists.cs.columbia.edu/cucslists/listinfo/p2p-sip
>
> _______________________________________________
> p2p-sip mailing list
> p2p-sip at cs.columbia.edu
> https://lists.cs.columbia.edu/cucslists/listinfo/p2p-sip

-------------- next part --------------
An HTML attachment was scrubbed...
URL: https://lists.cs.columbia.edu/pipermail/p2p-sip/attachments/20061116/6789c7d2/attachment-0001.html

From slavitch at gmail.com  Thu Nov 16 14:48:10 2006
From: slavitch at gmail.com (Michael Slavitch)
Date: Thu, 16 Nov 2006 14:48:10 -0500
Subject: [p2p-sip] Status of P2P SIP - did we form a WG at the IETF last
	week?
In-Reply-To: <2CAD4BA2-033E-4DBA-8CEA-E73921550DF2@magma.ca>
References: <fb4306550611151203g51343762w1d93cc08ba33a471@mail.gmail.com>
	<7b683f3f0611151857u1276e371q614d72be9c5071c0@mail.gmail.com>
	<2CAD4BA2-033E-4DBA-8CEA-E73921550DF2@magma.ca>
Message-ID: <1fc32c8f0611161148s16c31c7ei407b0f305ff68709@mail.gmail.com>

"This is the most boring BOF I have ever attended".

Good news.  A BOF should be about as exciting as a tax audit, for
similar reasons.

From enrico.marocco at telecomitalia.it  Fri Nov 17 04:32:15 2006
From: enrico.marocco at telecomitalia.it (Enrico Marocco)
Date: Fri, 17 Nov 2006 10:32:15 +0100
Subject: [p2p-sip] Status of P2P SIP - did we form a WG at the IETF last
 week?
In-Reply-To: <7b683f3f0611151857u1276e371q614d72be9c5071c0@mail.gmail.com>
References: <fb4306550611151203g51343762w1d93cc08ba33a471@mail.gmail.com>
	<7b683f3f0611151857u1276e371q614d72be9c5071c0@mail.gmail.com>
Message-ID: <455D819F.7030904@telecomitalia.it>

Eunsoo Shim wrote:
> I believe the BOF went well.

Defenitely the best of the many BOFs we've had :-)

-- 
Ciao,
Enrico

From drage at lucent.com  Fri Nov 17 09:51:38 2006
From: drage at lucent.com (Drage, Keith (Keith))
Date: Fri, 17 Nov 2006 15:51:38 +0100
Subject: [p2p-sip] Status of P2P SIP - did we form a WG at the IETF
	lastweek?
Message-ID: <5D1A7985295922448D5550C94DE291807F3385@DEEXC1U01.de.lucent.com>

Yes, but I do wonder why the IESG (I assume it was them) had to run the
process to the fullest extent. To me it was clear six months ago that
the proposed WG met all the rules, and the final polishing of the
charter could have been handled on the IETF discussion list (where I
believe it still has to go anyway). Having the extra BOF is not required
by the rules.

We seemed to end up having 300 people in a room to confirm a done deal.

Regards

Keith

> -----Original Message-----
> From: p2p-sip-bounces at cs.columbia.edu 
> [mailto:p2p-sip-bounces at cs.columbia.edu] On Behalf Of Michael Slavitch
> Sent: 16 November 2006 19:48
> To: Philip Matthews; P2P-SIP
> Subject: Re: [p2p-sip] Status of P2P SIP - did we form a WG 
> at the IETF lastweek?
> 
> "This is the most boring BOF I have ever attended".
> 
> Good news.  A BOF should be about as exciting as a tax audit, 
> for similar reasons.
> _______________________________________________
> p2p-sip mailing list
> p2p-sip at cs.columbia.edu
> https://lists.cs.columbia.edu/cucslists/listinfo/p2p-sip
> 


From slavitch at gmail.com  Fri Nov 17 10:29:12 2006
From: slavitch at gmail.com (Michael Slavitch)
Date: Fri, 17 Nov 2006 10:29:12 -0500
Subject: [p2p-sip] Status of P2P SIP - did we form a WG at the IETF
	lastweek?
In-Reply-To: <5D1A7985295922448D5550C94DE291807F3385@DEEXC1U01.de.lucent.com>
References: <5D1A7985295922448D5550C94DE291807F3385@DEEXC1U01.de.lucent.com>
Message-ID: <1fc32c8f0611170729n5ed4ebeap79b2cf8de3578e3f@mail.gmail.com>

You need the 300 people in a room to get a done deal.

On 11/17/06, Drage, Keith (Keith) <drage at lucent.com> wrote:

> We seemed to end up having 300 people in a room to confirm a done deal.
>
> Regards
>
> Keith
>

From fluffy at cisco.com  Sat Nov 18 21:58:43 2006
From: fluffy at cisco.com (Cullen Jennings)
Date: Sat, 18 Nov 2006 21:58:43 -0500
Subject: [p2p-sip] Status of P2P SIP - did we form a WG at the IETF
	lastweek?
In-Reply-To: <1fc32c8f0611170729n5ed4ebeap79b2cf8de3578e3f@mail.gmail.com>
References: <5D1A7985295922448D5550C94DE291807F3385@DEEXC1U01.de.lucent.com>
	<1fc32c8f0611170729n5ed4ebeap79b2cf8de3578e3f@mail.gmail.com>
Message-ID: <391DDCE7-6177-453E-ADC3-662B66E17D07@cisco.com>


The BOF went very well - the odds are very high a WG will be formed.  
What happens next might seem like needless bureaucracy but it really  
does add some value (ok, ok, I realize my viewpoint might be  
questionable here :-) The IETF believes in things like being able to  
work on a mailing list and letting the whole community comment - that  
means that forming a WG actually has to go to IETF Last Call on the  
IETF mailing list. This is a good thing IMHO. The IESG/IAB also needs  
some time to try and figure out how to make sure this work lines up  
with other WGs, IRTF groups, and try and set up the work so it has  
maximums chance of making good progress.

Don't worry - the BOF was a success and you are over the hard part of  
getting to a WG.

Cullen <with my AD hat on>


On Nov 17, 2006, at 10:29 AM, Michael Slavitch wrote:

> You need the 300 people in a room to get a done deal.
>
> On 11/17/06, Drage, Keith (Keith) <drage at lucent.com> wrote:
>
>> We seemed to end up having 300 people in a room to confirm a done  
>> deal.
>>
>> Regards
>>
>> Keith
>>
> _______________________________________________
> p2p-sip mailing list
> p2p-sip at cs.columbia.edu
> https://lists.cs.columbia.edu/cucslists/listinfo/p2p-sip

From adamfisk at gmail.com  Sun Nov 19 22:51:32 2006
From: adamfisk at gmail.com (Adam Fisk)
Date: Sun, 19 Nov 2006 22:51:32 -0500
Subject: [p2p-sip] what's wrong with DNS?
Message-ID: <f40963db0611191951r5638ebc8q94c38e28275cf2d5@mail.gmail.com>

I had a conversation with Paul Francis at the recent networking conference
at NYU.  I brought up p2p SIP and the implementation of it I've been working
on, and he said "What's wrong with DNS?"  Specifically, with dynamic DNS.  I
didn't understand his point at first because my work incorporates HTTP web
servers, and I thought he missed that you still need SIP/STUN/TURN/ICE to
negotiate the NATs.  He was talking about for the registrars, though.  If
your registrar goes down for, say, sip:afisk at mysubdomain.p2psip.org, and I
get a new one, why not just use dynamic DNS to remap
mysubdomain.p2psip.orgto the new registrar?

It was something I hadn't thought about before.  Has it been discussed
here?  Granted, there's a lag to the DNS updating, but what if you give each
node a couple of sudmomains and register in a couple of places?  Also, is
churn really high enough to make the lag a major concern?

It's certainly a far simpler solution to the problem, and it would reduce
call latency tremendously in the typical case over a DHT lookup.  Apologies
if this has been covered in detail.

Thanks.

-Adam
-------------- next part --------------
An HTML attachment was scrubbed...
URL: https://lists.cs.columbia.edu/pipermail/p2p-sip/attachments/20061119/1b8a294f/attachment.html

From karstbj at yahoo.com  Mon Nov 20 03:30:05 2006
From: karstbj at yahoo.com (Karst Bjorgson)
Date: Mon, 20 Nov 2006 00:30:05 -0800 (PST)
Subject: [p2p-sip] what's wrong with DNS?
In-Reply-To: <f40963db0611191951r5638ebc8q94c38e28275cf2d5@mail.gmail.com>
Message-ID: <482549.16699.qm@web53003.mail.yahoo.com>

> I had a conversation with Paul Francis at the recent
> networking conference
> at NYU.  I brought up p2p SIP and the implementation
> of it I've been working
> on, and he said "What's wrong with DNS?" 
> Specifically, with dynamic DNS.  I
> didn't understand his point at first because my work
> incorporates HTTP web
> servers, and I thought he missed that you still need
> SIP/STUN/TURN/ICE to
> negotiate the NATs.  He was talking about for the
> registrars, though.  If
> your registrar goes down for, say,
> sip:afisk at mysubdomain.p2psip.org, and I
> get a new one, why not just use dynamic DNS to remap
> mysubdomain.p2psip.orgto the new registrar?

Also, it is entirely possible to build a P2P system
that simply let you register names in a subtree of the
DNS. For example, the PNRP feature in Vista allows
each Vista host to publish a name of the form
"JohnDoe.pnrp.net", or a secure variant thereof,
according to Noah Horton's blog
(http://blogs.msdn.com/noahh/archive/2006/08/23/715894.aspx).
Once you can publish a DNS name, running SIP appears
rather simple, right?

-- Karst


 
____________________________________________________________________________________
Sponsored Link

$420k for $1,399/mo. 
Think You Pay Too Much For Your Mortgage? 
Find Out! www.LowerMyBills.com/lre

From ggb at tid.es  Mon Nov 20 05:30:17 2006
From: ggb at tid.es (=?iso-8859-1?Q?Gustavo_Garc=EDa_Bernardo?=)
Date: Mon, 20 Nov 2006 11:30:17 +0100
Subject: [p2p-sip] what's wrong with DNS?
In-Reply-To: <482549.16699.qm@web53003.mail.yahoo.com>
Message-ID: <039901c70c8e$e000b2e0$01595f0a@hi.inet>

> > I had a conversation with Paul Francis at the recent
> > networking conference
> > at NYU.  I brought up p2p SIP and the implementation
> > of it I've been working
> > on, and he said "What's wrong with DNS?"
> > Specifically, with dynamic DNS.  I
> > didn't understand his point at first because my work
> > incorporates HTTP web
> > servers, and I thought he missed that you still need
> > SIP/STUN/TURN/ICE to
> > negotiate the NATs.  He was talking about for the
> > registrars, though.  If
> > your registrar goes down for, say,
> > sip:afisk at mysubdomain.p2psip.org, and I
> > get a new one, why not just use dynamic DNS to remap
> > mysubdomain.p2psip.orgto the new registrar?

That solution continues being centralized (you are only changing the
registrar when the main goes down), you aren't getting the advantages of P2P
like zero infrastructure, load distribution among nodes, working in
isolated/ad-hoc environments... 

> Also, it is entirely possible to build a P2P system
> that simply let you register names in a subtree of the
> DNS. For example, the PNRP feature in Vista allows
> each Vista host to publish a name of the form
> "JohnDoe.pnrp.net", or a secure variant thereof,
> according to Noah Horton's blog
> (http://blogs.msdn.com/noahh/archive/2006/08/23/715894.aspx).
> Once you can publish a DNS name, running SIP appears
> rather simple, right?

As far as I know, PNRP uses a DHT like algorithm.  PNRP doesn't register
anything in DNS despite of the fact that it employs IDs like DNS ones
(subdomain.domain).

> -- Karst
> 


G.
 
> __________________________________________________________________________
> __________
> Sponsored Link
> 
> $420k for $1,399/mo.
> Think You Pay Too Much For Your Mortgage?
> Find Out! www.LowerMyBills.com/lre
> _______________________________________________
> p2p-sip mailing list
> p2p-sip at cs.columbia.edu
> https://lists.cs.columbia.edu/cucslists/listinfo/p2p-sip


From karstbj at yahoo.com  Mon Nov 20 10:14:33 2006
From: karstbj at yahoo.com (Karst Bjorgson)
Date: Mon, 20 Nov 2006 07:14:33 -0800 (PST)
Subject: [p2p-sip] what's wrong with DNS?
In-Reply-To: <039901c70c8e$e000b2e0$01595f0a@hi.inet>
Message-ID: <20061120151433.25866.qmail@web53008.mail.yahoo.com>

> As far as I know, PNRP uses a DHT like algorithm. 
> PNRP doesn't register
> anything in DNS despite of the fact that it employs
> IDs like DNS ones
> (subdomain.domain). 

Sure, but why is that a problem? Suppose that your
implementation of DHT allows for the participants to
publih names like
<some-peer-name>.<DHT-specific-domain>, and to resolve
them using the standard "getaddrinfo" API. You get a
system in which registration is effectively
peer-to-peer, but where the code is very much
compatible with a standard DNS based SIP routing. 

You also get a very neat solution to the "multiple
DHT" issue: just make sure that each
"DHT-specific-domain" incorporates a couple of
gateways to the DNS, so that requests by non-DHT aware
hosts can be resolved. If you want operation on
isolated subnets, you may also incorporate a gateway
with mDNS/LLMNR.

-- Karst 


 
____________________________________________________________________________________
Sponsored Link

Degrees online in as fast as 1 Yr
MBA, Bachelor's, Master's, Assoc
http://yahoo.degrees.info

From adamfisk at gmail.com  Mon Nov 20 10:41:59 2006
From: adamfisk at gmail.com (Adam Fisk)
Date: Mon, 20 Nov 2006 10:41:59 -0500
Subject: [p2p-sip] what's wrong with DNS?
In-Reply-To: <039901c70c8e$e000b2e0$01595f0a@hi.inet>
References: <482549.16699.qm@web53003.mail.yahoo.com>
	<039901c70c8e$e000b2e0$01595f0a@hi.inet>
Message-ID: <f40963db0611200741y3a878091p5c83b4f8d9b54b58@mail.gmail.com>

No -- I'm talking about continuing to have non-firewalled nodes act as
registrars and proxies.  So, you still have a voluntary, fully distributed
system.  In fact, it would be just as decentralized as the current vision
for p2p SIP, except it would use dynamic DNS instead of a DHT.

-Adam


On 11/20/06, Gustavo Garc?a Bernardo <ggb at tid.es> wrote:
>
> > > I had a conversation with Paul Francis at the recent
> > > networking conference
> > > at NYU.  I brought up p2p SIP and the implementation
> > > of it I've been working
> > > on, and he said "What's wrong with DNS?"
> > > Specifically, with dynamic DNS.  I
> > > didn't understand his point at first because my work
> > > incorporates HTTP web
> > > servers, and I thought he missed that you still need
> > > SIP/STUN/TURN/ICE to
> > > negotiate the NATs.  He was talking about for the
> > > registrars, though.  If
> > > your registrar goes down for, say,
> > > sip:afisk at mysubdomain.p2psip.org, and I
> > > get a new one, why not just use dynamic DNS to remap
> > > mysubdomain.p2psip.orgto the new registrar?
>
> That solution continues being centralized (you are only changing the
> registrar when the main goes down), you aren't getting the advantages of
> P2P
> like zero infrastructure, load distribution among nodes, working in
> isolated/ad-hoc environments...
>
> > Also, it is entirely possible to build a P2P system
> > that simply let you register names in a subtree of the
> > DNS. For example, the PNRP feature in Vista allows
> > each Vista host to publish a name of the form
> > "JohnDoe.pnrp.net", or a secure variant thereof,
> > according to Noah Horton's blog
> > (http://blogs.msdn.com/noahh/archive/2006/08/23/715894.aspx).
> > Once you can publish a DNS name, running SIP appears
> > rather simple, right?
>
> As far as I know, PNRP uses a DHT like algorithm.  PNRP doesn't register
> anything in DNS despite of the fact that it employs IDs like DNS ones
> (subdomain.domain).
>
> > -- Karst
> >
>
>
> G.
>
> >
> __________________________________________________________________________
> > __________
> > Sponsored Link
> >
> > $420k for $1,399/mo.
> > Think You Pay Too Much For Your Mortgage?
> > Find Out! www.LowerMyBills.com/lre
> > _______________________________________________
> > p2p-sip mailing list
> > p2p-sip at cs.columbia.edu
> > https://lists.cs.columbia.edu/cucslists/listinfo/p2p-sip
>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: https://lists.cs.columbia.edu/pipermail/p2p-sip/attachments/20061120/d10de158/attachment.html

From philip_matthews at magma.ca  Mon Nov 20 15:08:59 2006
From: philip_matthews at magma.ca (Philip Matthews)
Date: Mon, 20 Nov 2006 15:08:59 -0500
Subject: [p2p-sip] what's wrong with DNS?
In-Reply-To: <f40963db0611191951r5638ebc8q94c38e28275cf2d5@mail.gmail.com>
References: <f40963db0611191951r5638ebc8q94c38e28275cf2d5@mail.gmail.com>
Message-ID: <EA3E4C3E-DA53-4DBB-855A-7D857D8EB416@magma.ca>

Adam:

Are you suggesting that the P2PSIP Peer  Protocol (as described in
draft-willis-p2psip-concepts-03) could be a combination of SIP and  
dynamic DNS?
Or are you suggesting that we use the existing DNS and SIP  
infrastructure
in some P2P manner?
Perhaps you could go into a bit more detail on how all this would work?


- Philip

On 19-Nov-06, at 22:51 , Adam Fisk wrote:

> I had a conversation with Paul Francis at the recent networking  
> conference at NYU.  I brought up p2p SIP and the implementation of  
> it I've been working on, and he said "What's wrong with DNS?"   
> Specifically, with dynamic DNS.  I didn't understand his point at  
> first because my work incorporates HTTP web servers, and I thought  
> he missed that you still need SIP/STUN/TURN/ICE to negotiate the  
> NATs.  He was talking about for the registrars, though.  If your  
> registrar goes down for, say, sip:afisk at mysubdomain.p2psip.org, and  
> I get a new one, why not just use dynamic DNS to remap  
> mysubdomain.p2psip.org to the new registrar?
>
> It was something I hadn't thought about before.  Has it been  
> discussed here?  Granted, there's a lag to the DNS updating, but  
> what if you give each node a couple of sudmomains and register in a  
> couple of places?  Also, is churn really high enough to make the  
> lag a major concern?
>
> It's certainly a far simpler solution to the problem, and it would  
> reduce call latency tremendously in the typical case over a DHT  
> lookup.  Apologies if this has been covered in detail.
>
> Thanks.
>
> -Adam
> _______________________________________________
> p2p-sip mailing list
> p2p-sip at cs.columbia.edu
> https://lists.cs.columbia.edu/cucslists/listinfo/p2p-sip

-------------- next part --------------
An HTML attachment was scrubbed...
URL: https://lists.cs.columbia.edu/pipermail/p2p-sip/attachments/20061120/30c2446b/attachment.html

From hsinnrei at adobe.com  Mon Nov 20 18:52:51 2006
From: hsinnrei at adobe.com (Henry Sinnreich)
Date: Mon, 20 Nov 2006 15:52:51 -0800
Subject: [p2p-sip] what's wrong with DNS?
In-Reply-To: <039901c70c8e$e000b2e0$01595f0a@hi.inet>
Message-ID: <24CCCC428EFEA2469BF046DB3C7A8D2214133E@namail5.corp.adobe.com>

>>What's wrong with DNS?

> you aren't getting the advantages of P2P
Correct.

P2P overlays are supposed to work even when disconnected from the 
big Internet and besides, have their own name space.

Peers can act as proxies to connect the overlay to DNS.

DNS needs some improvements too as argued in 

http://codeen.cs.princeton.edu/codns/ 

Thanks Henry

-----Original Message-----
From: p2p-sip-bounces at cs.columbia.edu [mailto:p2p-sip-bounces at cs.columbia.edu] On Behalf Of Gustavo Garc?a Bernardo
Sent: Monday, November 20, 2006 4:30 AM
To: 'Karst Bjorgson'; 'Adam Fisk'; p2p-sip at cs.columbia.edu
Subject: Re: [p2p-sip] what's wrong with DNS?

> > I had a conversation with Paul Francis at the recent
> > networking conference
> > at NYU.  I brought up p2p SIP and the implementation
> > of it I've been working
> > on, and he said "What's wrong with DNS?"
> > Specifically, with dynamic DNS.  I
> > didn't understand his point at first because my work
> > incorporates HTTP web
> > servers, and I thought he missed that you still need
> > SIP/STUN/TURN/ICE to
> > negotiate the NATs.  He was talking about for the
> > registrars, though.  If
> > your registrar goes down for, say,
> > sip:afisk at mysubdomain.p2psip.org, and I
> > get a new one, why not just use dynamic DNS to remap
> > mysubdomain.p2psip.orgto the new registrar?

That solution continues being centralized (you are only changing the
registrar when the main goes down), you aren't getting the advantages of P2P
like zero infrastructure, load distribution among nodes, working in
isolated/ad-hoc environments... 

> Also, it is entirely possible to build a P2P system
> that simply let you register names in a subtree of the
> DNS. For example, the PNRP feature in Vista allows
> each Vista host to publish a name of the form
> "JohnDoe.pnrp.net", or a secure variant thereof,
> according to Noah Horton's blog
> (http://blogs.msdn.com/noahh/archive/2006/08/23/715894.aspx).
> Once you can publish a DNS name, running SIP appears
> rather simple, right?

As far as I know, PNRP uses a DHT like algorithm.  PNRP doesn't register
anything in DNS despite of the fact that it employs IDs like DNS ones
(subdomain.domain).

> -- Karst
> 


G.
 
> __________________________________________________________________________
> __________
> Sponsored Link
> 
> $420k for $1,399/mo.
> Think You Pay Too Much For Your Mortgage?
> Find Out! www.LowerMyBills.com/lre
> _______________________________________________
> p2p-sip mailing list
> p2p-sip at cs.columbia.edu
> https://lists.cs.columbia.edu/cucslists/listinfo/p2p-sip

_______________________________________________
p2p-sip mailing list
p2p-sip at cs.columbia.edu
https://lists.cs.columbia.edu/cucslists/listinfo/p2p-sip


From hsinnrei at adobe.com  Mon Nov 20 19:19:23 2006
From: hsinnrei at adobe.com (Henry Sinnreich)
Date: Mon, 20 Nov 2006 16:19:23 -0800
Subject: [p2p-sip] what's wrong with DNS?
In-Reply-To: <f40963db0611200741y3a878091p5c83b4f8d9b54b58@mail.gmail.com>
Message-ID: <24CCCC428EFEA2469BF046DB3C7A8D22141340@namail5.corp.adobe.com>

>except it would use dynamic DNS instead of a DHT

 

This is a terrible idea.

 

P2P overlays are supposed to work even when disconnected from the big Internet and besides, have their own name space.

 

Peers can act as proxies to connect the overlay to DNS and ENUM. 

 

(BTW: In the trapezoid model, adding the outgoing SIP proxy lookup to an ENUM lookup gives you total of 6 (six) DNS lookups).

 

DNS needs some improvements too as argued in 

 

http://codeen.cs.princeton.edu/codns/ <http://codeen.cs.princeton.edu/codns/> 

 

Thanks, Henry

 

________________________________

From: p2p-sip-bounces at cs.columbia.edu [mailto:p2p-sip-bounces at cs.columbia.edu] On Behalf Of Adam Fisk
Sent: Monday, November 20, 2006 9:42 AM
To: Gustavo Garc?a Bernardo
Cc: p2p-sip at cs.columbia.edu
Subject: Re: [p2p-sip] what's wrong with DNS?

 

No -- I'm talking about continuing to have non-firewalled nodes act as registrars and proxies.  So, you still have a voluntary, fully distributed system.  In fact, it would be just as decentralized as the current vision for p2p SIP, except it would use dynamic DNS instead of a DHT. 

-Adam

On 11/20/06, Gustavo Garc?a Bernardo <ggb at tid.es> wrote:

> > I had a conversation with Paul Francis at the recent
> > networking conference
> > at NYU.  I brought up p2p SIP and the implementation
> > of it I've been working
> > on, and he said "What's wrong with DNS?" 
> > Specifically, with dynamic DNS.  I
> > didn't understand his point at first because my work
> > incorporates HTTP web
> > servers, and I thought he missed that you still need
> > SIP/STUN/TURN/ICE to 
> > negotiate the NATs.  He was talking about for the
> > registrars, though.  If
> > your registrar goes down for, say,
> > sip:afisk at mysubdomain.p2psip.org , and I
> > get a new one, why not just use dynamic DNS to remap
> > mysubdomain.p2psip.orgto the new registrar?

That solution continues being centralized (you are only changing the
registrar when the main goes down), you aren't getting the advantages of P2P 
like zero infrastructure, load distribution among nodes, working in
isolated/ad-hoc environments...

> Also, it is entirely possible to build a P2P system
> that simply let you register names in a subtree of the 
> DNS. For example, the PNRP feature in Vista allows
> each Vista host to publish a name of the form
> "JohnDoe.pnrp.net", or a secure variant thereof,
> according to Noah Horton's blog
> (http://blogs.msdn.com/noahh/archive/2006/08/23/715894.aspx).
> Once you can publish a DNS name, running SIP appears 
> rather simple, right?

As far as I know, PNRP uses a DHT like algorithm.  PNRP doesn't register
anything in DNS despite of the fact that it employs IDs like DNS ones
(subdomain.domain).

> -- Karst 
>


G.

> __________________________________________________________________________
> __________
> Sponsored Link
>
> $420k for $1,399/mo.
> Think You Pay Too Much For Your Mortgage? 
> Find Out! www.LowerMyBills.com/lre
> _______________________________________________
> p2p-sip mailing list
> p2p-sip at cs.columbia.edu
> https://lists.cs.columbia.edu/cucslists/listinfo/p2p-sip

 

-------------- next part --------------
An HTML attachment was scrubbed...
URL: https://lists.cs.columbia.edu/pipermail/p2p-sip/attachments/20061120/4dbd40ac/attachment-0001.html

From saikat at cs.cornell.edu  Mon Nov 20 20:07:30 2006
From: saikat at cs.cornell.edu (Saikat Guha)
Date: Mon, 20 Nov 2006 20:07:30 -0500
Subject: [p2p-sip] what's wrong with DNS?
In-Reply-To: <24CCCC428EFEA2469BF046DB3C7A8D22141340@namail5.corp.adobe.com>
References: <24CCCC428EFEA2469BF046DB3C7A8D22141340@namail5.corp.adobe.com>
Message-ID: <1164071250.14261.4.camel@sioux.systems.cs.cornell.edu>

On Mon, 2006-11-20 at 16:19 -0800, Henry Sinnreich wrote:
> >except it would use dynamic DNS instead of a DHT
>
> P2P overlays are supposed to work even when disconnected from the big
> Internet and besides, have their own name space.

Even P2P overlays need some way to bootstrap in disconnected networks.
While DNS _typically_ bootstraps with hardcoded IP addresses of the 13
root servers, it can bootstrap in a disconnected scenario with a link
level broadcast. See Apple's rendezvous/zero-conf protocol and mDNS.


-- 
Saikat
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 189 bytes
Desc: This is a digitally signed message part
Url : https://lists.cs.columbia.edu/pipermail/p2p-sip/attachments/20061120/dfec6988/attachment.bin

From slavitch at gmail.com  Mon Nov 20 22:14:24 2006
From: slavitch at gmail.com (Michael Slavitch)
Date: Mon, 20 Nov 2006 22:14:24 -0500
Subject: [p2p-sip] what's wrong with DNS?
In-Reply-To: <1164071250.14261.4.camel@sioux.systems.cs.cornell.edu>
References: <24CCCC428EFEA2469BF046DB3C7A8D22141340@namail5.corp.adobe.com>
	<1164071250.14261.4.camel@sioux.systems.cs.cornell.edu>
Message-ID: <1fc32c8f0611201914pcba511bka99ecbdaa65136b@mail.gmail.com>

DNS was designed in the days when long-term store was scarce and so
was bandwidth. Technology was difficult and obtuse, with the number of
human implementors measured in the hundreds, and the number of users
in the thousands. I remember that because I was there. There was no
other way to pass this data. Something had to be first and that was
DNS, which reflected the times and the organizations that invented it
in the first place.

Those days are long gone now in the days of the $20 grocery-store 2G
flash RAM, millions of implementors, and billions of users.

Today's Internet is a consumer-driven multi-modal world which has no
patience for that.
To cripple bootstrap by leashing it to creaky DNS is a disservice to
the future.

Bootstrap should have some criteria requirements for some obvious
things such as satisfying the security AD but the actual
implementation should allow for individual implementations that
reflect a rapidly changing modern world.

Hardcoding is a quaint way of solving the problem, and it is far too
easy for protocol engineers to design in a vacuum. When a single
handheld device can receive inputs by SMS, email, walled and unwalled
IP,  and HTTP the number of sources for a bootstrap is openended and
impossible to mandate without creating limitations on functionality
and usefulness in a hybridized multi-modal world..

A mime message, encrypted by a private key known to that device, could
contain a myriad number of bootstraps, be they DNS, hardcoded IP (IPV4
or IPV6? Which?), local names,  peer names, SMS addresses,sip uris,
whathaveyou.  A list based on operating criteria, that can evolve over
time.

This message, more 'carrier pigeon'  than boostrap, can be delivered
via a variety of means, including via C/S SIP or stored from a
previous DHT. A primitive form of unencrypted messaging is called
'paper'.

In this environment a 'friend' network, of names associated with keys,
could be seeded entirely by SMS messages and never ever need DNS,, and
it works when zeroconf fails, which is often.  It could later use
broadcasts or multicasts to find other 'friends', using one of a
variety of techniques, zeroconf being one.

Create the requirements for the bootstrap that guarantee security,
functionality and interopability.  Create the list of appropriate
beacons, allowing room for more in the future.

Then stop, please.



On 11/20/06, Saikat Guha <saikat at cs.cornell.edu> wrote:
> On Mon, 2006-11-20 at 16:19 -0800, Henry Sinnreich wrote:
> > >except it would use dynamic DNS instead of a DHT
> >
> > P2P overlays are supposed to work even when disconnected from the big
> > Internet and besides, have their own name space.
>
> Even P2P overlays need some way to bootstrap in disconnected networks.
> While DNS _typically_ bootstraps with hardcoded IP addresses of the 13
> root servers, it can bootstrap in a disconnected scenario with a link
> level broadcast. See Apple's rendezvous/zero-conf protocol and mDNS.
>
>
> --
> Saikat
>
>
> _______________________________________________
> p2p-sip mailing list
> p2p-sip at cs.columbia.edu
> https://lists.cs.columbia.edu/cucslists/listinfo/p2p-sip
>
>
>
>

From Mehdi.Mani at int-evry.fr  Tue Nov 21 00:08:33 2006
From: Mehdi.Mani at int-evry.fr (Mehdi MANI)
Date: Tue, 21 Nov 2006 06:08:33 +0100
Subject: [p2p-sip] what's wrong with DNS?
In-Reply-To: <f40963db0611200741y3a878091p5c83b4f8d9b54b58@mail.gmail.com>
References: <482549.16699.qm@web53003.mail.yahoo.com>
	<039901c70c8e$e000b2e0$01595f0a@hi.inet>
	<f40963db0611200741y3a878091p5c83b4f8d9b54b58@mail.gmail.com>
Message-ID: <1164085713.456289d1027a9@imp.int-evry.fr>

Hi Adam,

According to your suggestion two approaches are possible:
Firstly, the super nodes (or at least the boot-strapt nodes) support DNS
functions instead of DHT.
Secondly, the super nodes only function as registrar and proxy and in addition
some dynamic DNS servers be implemented. In this case, when a node whant to be
localized should contact a DNS, then DNS lets it know on which super node (as a
registrar) it should register its ID. And the same process should be done for a
query.

If you consider the first approach, what would be advantage? DHT works much more
better in that case because it is able to organize efficiently the registery
information.
And the second approach, will be something like the architecture of Napster that
was not scalable enough.

-Mehdi

Selon Adam Fisk <adamfisk at gmail.com>:

> No -- I'm talking about continuing to have non-firewalled nodes act as
> registrars and proxies.  So, you still have a voluntary, fully distributed
> system.  In fact, it would be just as decentralized as the current vision
> for p2p SIP, except it would use dynamic DNS instead of a DHT.
>
> -Adam
>
>
> On 11/20/06, Gustavo Garc?a Bernardo <ggb at tid.es> wrote:
> >
> > > > I had a conversation with Paul Francis at the recent
> > > > networking conference
> > > > at NYU.  I brought up p2p SIP and the implementation
> > > > of it I've been working
> > > > on, and he said "What's wrong with DNS?"
> > > > Specifically, with dynamic DNS.  I
> > > > didn't understand his point at first because my work
> > > > incorporates HTTP web
> > > > servers, and I thought he missed that you still need
> > > > SIP/STUN/TURN/ICE to
> > > > negotiate the NATs.  He was talking about for the
> > > > registrars, though.  If
> > > > your registrar goes down for, say,
> > > > sip:afisk at mysubdomain.p2psip.org, and I
> > > > get a new one, why not just use dynamic DNS to remap
> > > > mysubdomain.p2psip.orgto the new registrar?
> >
> > That solution continues being centralized (you are only changing the
> > registrar when the main goes down), you aren't getting the advantages of
> > P2P
> > like zero infrastructure, load distribution among nodes, working in
> > isolated/ad-hoc environments...
> >
> > > Also, it is entirely possible to build a P2P system
> > > that simply let you register names in a subtree of the
> > > DNS. For example, the PNRP feature in Vista allows
> > > each Vista host to publish a name of the form
> > > "JohnDoe.pnrp.net", or a secure variant thereof,
> > > according to Noah Horton's blog
> > > (http://blogs.msdn.com/noahh/archive/2006/08/23/715894.aspx).
> > > Once you can publish a DNS name, running SIP appears
> > > rather simple, right?
> >
> > As far as I know, PNRP uses a DHT like algorithm.  PNRP doesn't register
> > anything in DNS despite of the fact that it employs IDs like DNS ones
> > (subdomain.domain).
> >
> > > -- Karst
> > >
> >
> >
> > G.
> >
> > >
> > __________________________________________________________________________
> > > __________
> > > Sponsored Link
> > >
> > > $420k for $1,399/mo.
> > > Think You Pay Too Much For Your Mortgage?
> > > Find Out! www.LowerMyBills.com/lre
> > > _______________________________________________
> > > p2p-sip mailing list
> > > p2p-sip at cs.columbia.edu
> > > https://lists.cs.columbia.edu/cucslists/listinfo/p2p-sip
> >
> >
>




----------------------------------------------------------------
This message was sent using IMP, the Internet Messaging Program.

From dbarrett at quinthar.com  Tue Nov 21 00:32:39 2006
From: dbarrett at quinthar.com (David Barrett)
Date: Mon, 20 Nov 2006 21:32:39 -0800
Subject: [p2p-sip] what's wrong with DNS?
In-Reply-To: <1164085713.456289d1027a9@imp.int-evry.fr>
References: <482549.16699.qm@web53003.mail.yahoo.com>
	<1164085713.456289d1027a9@imp.int-evry.fr>
Message-ID: <1164087163.391D096A@ba12.dngr.org>

On Mon, 20 Nov 2006 9:19 pm, Mehdi MANI wrote:
>
> And the second approach, will be something like the architecture of 
> Napster that
> was not scalable enough.

If it weren't for the lawsuits, nobody would have ever bothered with 
decentralized indexing.  Scalability was not the problem.

I like Adam's idea.  DNS works and scales in the real world; it's a 
proven solution.  For totally disconnected, ad hoc networks, perhaps we 
should start a P2P-DNS WG and solve that problem by itself.  That's 
bigger than SIP and might benefit from being separated out.

-david

From br at brianrosen.net  Tue Nov 21 09:21:17 2006
From: br at brianrosen.net (Brian Rosen)
Date: Tue, 21 Nov 2006 09:21:17 -0500
Subject: [p2p-sip] what's wrong with DNS?
In-Reply-To: <1164087163.391D096A@ba12.dngr.org>
Message-ID: <01e701c70d78$509a13e0$640fa8c0@cis.neustar.com>

Agree

Let's stick to things we know will work.  

Brian

> -----Original Message-----
> From: p2p-sip-bounces at cs.columbia.edu [mailto:p2p-sip-
> bounces at cs.columbia.edu] On Behalf Of David Barrett
> Sent: Tuesday, November 21, 2006 12:33 AM
> To: Mehdi MANI; Adam Fisk
> Cc: p2p-sip at cs.columbia.edu
> Subject: Re: [p2p-sip] what's wrong with DNS?
> 
> On Mon, 20 Nov 2006 9:19 pm, Mehdi MANI wrote:
> >
> > And the second approach, will be something like the architecture of
> > Napster that
> > was not scalable enough.
> 
> If it weren't for the lawsuits, nobody would have ever bothered with
> decentralized indexing.  Scalability was not the problem.
> 
> I like Adam's idea.  DNS works and scales in the real world; it's a
> proven solution.  For totally disconnected, ad hoc networks, perhaps we
> should start a P2P-DNS WG and solve that problem by itself.  That's
> bigger than SIP and might benefit from being separated out.
> 
> -david
> _______________________________________________
> p2p-sip mailing list
> p2p-sip at cs.columbia.edu
> https://lists.cs.columbia.edu/cucslists/listinfo/p2p-sip


From tme at multicasttech.com  Tue Nov 21 09:45:01 2006
From: tme at multicasttech.com (Marshall Eubanks)
Date: Tue, 21 Nov 2006 09:45:01 -0500
Subject: [p2p-sip] what's wrong with DNS?
In-Reply-To: <01e701c70d78$509a13e0$640fa8c0@cis.neustar.com>
References: <01e701c70d78$509a13e0$640fa8c0@cis.neustar.com>
Message-ID: <2F405B1C-A843-4CA9-848C-17CB4BE9E9F3@multicasttech.com>

Hello;

On Nov 21, 2006, at 9:21 AM, Brian Rosen wrote:

> Agree
>
> Let's stick to things we know will work.
>
> Brian
>

I am not sure what is actually being proposed here, so I am not sure  
what you are saying will work.

To use Dave Singer's mDNS ideas for LAN resource discovery ? Cool.  
There are RFC's - apply.

To extend these ideas, maybe using P2P application layer multicast,  
to the wide area - sounds
like an interesting idea, and could be very cool. I could see this  
being used all over the place.

To expand the core DNS into P2P discovery - i.e., to have P2P RR's  
added into the
root servers ? Is not a good idea, and (I dare say) will not happen  
in the IETF.

I think Michael Dillon put it well in DKIM :

"if someone wants to leverage the DNS protocol and
DNS server software in order to operate a global
distributed database, then there are no barriers to
doing this on ports other than port 53. The IETF barely
needs to be involved at all other than defining the
new RR.

But when someone suggests that port 53 servers should
all support some new RR or anything else that is new,
now we are talking about a major upgrade to the Internet's
mission critical infrastructure impacting millions of
people worldwide. Here the IETF has a major role to play
and the IETF should tread very carefully."

So, I would like to see a clearer description of just what path is  
actually being proposed here.

Regards
Marshall


>> -----Original Message-----
>> From: p2p-sip-bounces at cs.columbia.edu [mailto:p2p-sip-
>> bounces at cs.columbia.edu] On Behalf Of David Barrett
>> Sent: Tuesday, November 21, 2006 12:33 AM
>> To: Mehdi MANI; Adam Fisk
>> Cc: p2p-sip at cs.columbia.edu
>> Subject: Re: [p2p-sip] what's wrong with DNS?
>>
>> On Mon, 20 Nov 2006 9:19 pm, Mehdi MANI wrote:
>>>
>>> And the second approach, will be something like the architecture of
>>> Napster that
>>> was not scalable enough.
>>
>> If it weren't for the lawsuits, nobody would have ever bothered with
>> decentralized indexing.  Scalability was not the problem.
>>
>> I like Adam's idea.  DNS works and scales in the real world; it's a
>> proven solution.  For totally disconnected, ad hoc networks,  
>> perhaps we
>> should start a P2P-DNS WG and solve that problem by itself.  That's
>> bigger than SIP and might benefit from being separated out.
>>
>> -david
>> _______________________________________________
>> p2p-sip mailing list
>> p2p-sip at cs.columbia.edu
>> https://lists.cs.columbia.edu/cucslists/listinfo/p2p-sip
>
> _______________________________________________
> p2p-sip mailing list
> p2p-sip at cs.columbia.edu
> https://lists.cs.columbia.edu/cucslists/listinfo/p2p-sip


From br at brianrosen.net  Tue Nov 21 10:01:53 2006
From: br at brianrosen.net (Brian Rosen)
Date: Tue, 21 Nov 2006 10:01:53 -0500
Subject: [p2p-sip] what's wrong with DNS?
In-Reply-To: <2F405B1C-A843-4CA9-848C-17CB4BE9E9F3@multicasttech.com>
Message-ID: <022b01c70d7e$01803bd0$640fa8c0@cis.neustar.com>

I meant use of existing mechanisms such as DNS rather than inventing new
ones.

I think the dynamic DNS idea can work.  I think simpler NAPTR based
mechanisms can work (especially if, as I think we're heading for, we have a
"domain" for the community the P2P operates over, where we support multiple
such domains (the concepts document calls these "overlay IDs").

I'm sanguine about multicast mechanisms because multicast in general doesn't
work across the Internet :(

I don't think we need new RRs.

Brian

> -----Original Message-----
> From: Marshall Eubanks [mailto:tme at multicasttech.com]
> Sent: Tuesday, November 21, 2006 9:45 AM
> To: Brian Rosen
> Cc: 'David Barrett'; 'Mehdi MANI'; 'Adam Fisk'; p2p-sip at cs.columbia.edu
> Subject: Re: [p2p-sip] what's wrong with DNS?
> 
> Hello;
> 
> On Nov 21, 2006, at 9:21 AM, Brian Rosen wrote:
> 
> > Agree
> >
> > Let's stick to things we know will work.
> >
> > Brian
> >
> 
> I am not sure what is actually being proposed here, so I am not sure
> what you are saying will work.
> 
> To use Dave Singer's mDNS ideas for LAN resource discovery ? Cool.
> There are RFC's - apply.
> 
> To extend these ideas, maybe using P2P application layer multicast,
> to the wide area - sounds
> like an interesting idea, and could be very cool. I could see this
> being used all over the place.
> 
> To expand the core DNS into P2P discovery - i.e., to have P2P RR's
> added into the
> root servers ? Is not a good idea, and (I dare say) will not happen
> in the IETF.
> 
> I think Michael Dillon put it well in DKIM :
> 
> "if someone wants to leverage the DNS protocol and
> DNS server software in order to operate a global
> distributed database, then there are no barriers to
> doing this on ports other than port 53. The IETF barely
> needs to be involved at all other than defining the
> new RR.
> 
> But when someone suggests that port 53 servers should
> all support some new RR or anything else that is new,
> now we are talking about a major upgrade to the Internet's
> mission critical infrastructure impacting millions of
> people worldwide. Here the IETF has a major role to play
> and the IETF should tread very carefully."
> 
> So, I would like to see a clearer description of just what path is
> actually being proposed here.
> 
> Regards
> Marshall
> 
> 
> >> -----Original Message-----
> >> From: p2p-sip-bounces at cs.columbia.edu [mailto:p2p-sip-
> >> bounces at cs.columbia.edu] On Behalf Of David Barrett
> >> Sent: Tuesday, November 21, 2006 12:33 AM
> >> To: Mehdi MANI; Adam Fisk
> >> Cc: p2p-sip at cs.columbia.edu
> >> Subject: Re: [p2p-sip] what's wrong with DNS?
> >>
> >> On Mon, 20 Nov 2006 9:19 pm, Mehdi MANI wrote:
> >>>
> >>> And the second approach, will be something like the architecture of
> >>> Napster that
> >>> was not scalable enough.
> >>
> >> If it weren't for the lawsuits, nobody would have ever bothered with
> >> decentralized indexing.  Scalability was not the problem.
> >>
> >> I like Adam's idea.  DNS works and scales in the real world; it's a
> >> proven solution.  For totally disconnected, ad hoc networks,
> >> perhaps we
> >> should start a P2P-DNS WG and solve that problem by itself.  That's
> >> bigger than SIP and might benefit from being separated out.
> >>
> >> -david
> >> _______________________________________________
> >> p2p-sip mailing list
> >> p2p-sip at cs.columbia.edu
> >> https://lists.cs.columbia.edu/cucslists/listinfo/p2p-sip
> >
> > _______________________________________________
> > p2p-sip mailing list
> > p2p-sip at cs.columbia.edu
> > https://lists.cs.columbia.edu/cucslists/listinfo/p2p-sip


From bryan at ethernot.org  Tue Nov 21 10:39:05 2006
From: bryan at ethernot.org (David A. Bryan)
Date: Tue, 21 Nov 2006 10:39:05 -0500
Subject: [p2p-sip] what's wrong with DNS?
In-Reply-To: <022b01c70d7e$01803bd0$640fa8c0@cis.neustar.com>
References: <2F405B1C-A843-4CA9-848C-17CB4BE9E9F3@multicasttech.com>
	<022b01c70d7e$01803bd0$640fa8c0@cis.neustar.com>
Message-ID: <8b2769930611210739u2a7a115bne91ccd6805e23313@mail.gmail.com>

We've actually discussed this several times at several ad-hocs, and
the general conclusion of the group was that dynamic DNS did not meet
the needs on several fronts:

1) mobility. We want to preserve the (rapid) update ability of a SIP
registrar, for example going from home to the office to the mobile
phone in a period of a few minutes. There is a great deal of concern
if dynamic DNS could meet these requirements.

2) it still requires centralized infrastructure, in the sense that
some known entity is the server -- even if that isn't a fixed server
but one of the "peers", it will be doing much of the work.

3) if it is using the existing DNS approach, it requires Internet
connectivity -- not very useful for disconnected scenarios.

The charter that went around for several months was a reflection of
this -- it was explicitly ruled out of scope as item 4. It is very
interesting as an approach for resolution, but doesn't meet the needs
of the folks looking at this WG because of the points above.

David

On 11/21/06, Brian Rosen <br at brianrosen.net> wrote:
> I meant use of existing mechanisms such as DNS rather than inventing new
> ones.
>
> I think the dynamic DNS idea can work.  I think simpler NAPTR based
> mechanisms can work (especially if, as I think we're heading for, we have a
> "domain" for the community the P2P operates over, where we support multiple
> such domains (the concepts document calls these "overlay IDs").
>
> I'm sanguine about multicast mechanisms because multicast in general doesn't
> work across the Internet :(
>
> I don't think we need new RRs.
>
> Brian
>
> > -----Original Message-----
> > From: Marshall Eubanks [mailto:tme at multicasttech.com]
> > Sent: Tuesday, November 21, 2006 9:45 AM
> > To: Brian Rosen
> > Cc: 'David Barrett'; 'Mehdi MANI'; 'Adam Fisk'; p2p-sip at cs.columbia.edu
> > Subject: Re: [p2p-sip] what's wrong with DNS?
> >
> > Hello;
> >
> > On Nov 21, 2006, at 9:21 AM, Brian Rosen wrote:
> >
> > > Agree
> > >
> > > Let's stick to things we know will work.
> > >
> > > Brian
> > >
> >
> > I am not sure what is actually being proposed here, so I am not sure
> > what you are saying will work.
> >
> > To use Dave Singer's mDNS ideas for LAN resource discovery ? Cool.
> > There are RFC's - apply.
> >
> > To extend these ideas, maybe using P2P application layer multicast,
> > to the wide area - sounds
> > like an interesting idea, and could be very cool. I could see this
> > being used all over the place.
> >
> > To expand the core DNS into P2P discovery - i.e., to have P2P RR's
> > added into the
> > root servers ? Is not a good idea, and (I dare say) will not happen
> > in the IETF.
> >
> > I think Michael Dillon put it well in DKIM :
> >
> > "if someone wants to leverage the DNS protocol and
> > DNS server software in order to operate a global
> > distributed database, then there are no barriers to
> > doing this on ports other than port 53. The IETF barely
> > needs to be involved at all other than defining the
> > new RR.
> >
> > But when someone suggests that port 53 servers should
> > all support some new RR or anything else that is new,
> > now we are talking about a major upgrade to the Internet's
> > mission critical infrastructure impacting millions of
> > people worldwide. Here the IETF has a major role to play
> > and the IETF should tread very carefully."
> >
> > So, I would like to see a clearer description of just what path is
> > actually being proposed here.
> >
> > Regards
> > Marshall
> >
> >
> > >> -----Original Message-----
> > >> From: p2p-sip-bounces at cs.columbia.edu [mailto:p2p-sip-
> > >> bounces at cs.columbia.edu] On Behalf Of David Barrett
> > >> Sent: Tuesday, November 21, 2006 12:33 AM
> > >> To: Mehdi MANI; Adam Fisk
> > >> Cc: p2p-sip at cs.columbia.edu
> > >> Subject: Re: [p2p-sip] what's wrong with DNS?
> > >>
> > >> On Mon, 20 Nov 2006 9:19 pm, Mehdi MANI wrote:
> > >>>
> > >>> And the second approach, will be something like the architecture of
> > >>> Napster that
> > >>> was not scalable enough.
> > >>
> > >> If it weren't for the lawsuits, nobody would have ever bothered with
> > >> decentralized indexing.  Scalability was not the problem.
> > >>
> > >> I like Adam's idea.  DNS works and scales in the real world; it's a
> > >> proven solution.  For totally disconnected, ad hoc networks,
> > >> perhaps we
> > >> should start a P2P-DNS WG and solve that problem by itself.  That's
> > >> bigger than SIP and might benefit from being separated out.
> > >>
> > >> -david
> > >> _______________________________________________
> > >> p2p-sip mailing list
> > >> p2p-sip at cs.columbia.edu
> > >> https://lists.cs.columbia.edu/cucslists/listinfo/p2p-sip
> > >
> > > _______________________________________________
> > > p2p-sip mailing list
> > > p2p-sip at cs.columbia.edu
> > > https://lists.cs.columbia.edu/cucslists/listinfo/p2p-sip
>
> _______________________________________________
> p2p-sip mailing list
> p2p-sip at cs.columbia.edu
> https://lists.cs.columbia.edu/cucslists/listinfo/p2p-sip
>

From hsinnrei at adobe.com  Tue Nov 21 11:29:44 2006
From: hsinnrei at adobe.com (Henry Sinnreich)
Date: Tue, 21 Nov 2006 08:29:44 -0800
Subject: [p2p-sip] what's wrong with DNS?
In-Reply-To: <1fc32c8f0611201914pcba511bka99ecbdaa65136b@mail.gmail.com>
Message-ID: <24CCCC428EFEA2469BF046DB3C7A8D22141344@namail5.corp.adobe.com>

I agree with Michael. 

Research folks are reporting on why and how to replace the DNS. 
Here is some good weekend reading to start with :-)

http://www.acm.org/sigs/sigcomm/sigcomm2004/papers/p292-ramasubramanian1
111.pdf 

After homework - reading this paper, we can formulate something better
for p2p bootstrap.

Thanks, Henry

-----Original Message-----
From: Michael Slavitch [mailto:slavitch at gmail.com] 
Sent: Monday, November 20, 2006 9:14 PM
To: Saikat Guha
Cc: Henry Sinnreich; p2p-sip at cs.columbia.edu
Subject: Re: [p2p-sip] what's wrong with DNS?

DNS was designed in the days when long-term store was scarce and so
was bandwidth. Technology was difficult and obtuse, with the number of
human implementors measured in the hundreds, and the number of users
in the thousands. I remember that because I was there. There was no
other way to pass this data. Something had to be first and that was
DNS, which reflected the times and the organizations that invented it
in the first place.

Those days are long gone now in the days of the $20 grocery-store 2G
flash RAM, millions of implementors, and billions of users.

Today's Internet is a consumer-driven multi-modal world which has no
patience for that.
To cripple bootstrap by leashing it to creaky DNS is a disservice to
the future.

Bootstrap should have some criteria requirements for some obvious
things such as satisfying the security AD but the actual
implementation should allow for individual implementations that
reflect a rapidly changing modern world.

Hardcoding is a quaint way of solving the problem, and it is far too
easy for protocol engineers to design in a vacuum. When a single
handheld device can receive inputs by SMS, email, walled and unwalled
IP,  and HTTP the number of sources for a bootstrap is openended and
impossible to mandate without creating limitations on functionality
and usefulness in a hybridized multi-modal world..

A mime message, encrypted by a private key known to that device, could
contain a myriad number of bootstraps, be they DNS, hardcoded IP (IPV4
or IPV6? Which?), local names,  peer names, SMS addresses,sip uris,
whathaveyou.  A list based on operating criteria, that can evolve over
time.

This message, more 'carrier pigeon'  than boostrap, can be delivered
via a variety of means, including via C/S SIP or stored from a
previous DHT. A primitive form of unencrypted messaging is called
'paper'.

In this environment a 'friend' network, of names associated with keys,
could be seeded entirely by SMS messages and never ever need DNS,, and
it works when zeroconf fails, which is often.  It could later use
broadcasts or multicasts to find other 'friends', using one of a
variety of techniques, zeroconf being one.

Create the requirements for the bootstrap that guarantee security,
functionality and interopability.  Create the list of appropriate
beacons, allowing room for more in the future.

Then stop, please.



On 11/20/06, Saikat Guha <saikat at cs.cornell.edu> wrote:
> On Mon, 2006-11-20 at 16:19 -0800, Henry Sinnreich wrote:
> > >except it would use dynamic DNS instead of a DHT
> >
> > P2P overlays are supposed to work even when disconnected from the
big
> > Internet and besides, have their own name space.
>
> Even P2P overlays need some way to bootstrap in disconnected networks.
> While DNS _typically_ bootstraps with hardcoded IP addresses of the 13
> root servers, it can bootstrap in a disconnected scenario with a link
> level broadcast. See Apple's rendezvous/zero-conf protocol and mDNS.
>
>
> --
> Saikat
>
>
> _______________________________________________
> p2p-sip mailing list
> p2p-sip at cs.columbia.edu
> https://lists.cs.columbia.edu/cucslists/listinfo/p2p-sip
>
>
>
>


From adamfisk at gmail.com  Tue Nov 21 12:24:33 2006
From: adamfisk at gmail.com (Adam Fisk)
Date: Tue, 21 Nov 2006 12:24:33 -0500
Subject: [p2p-sip] what's wrong with DNS?
In-Reply-To: <8b2769930611210739u2a7a115bne91ccd6805e23313@mail.gmail.com>
References: <2F405B1C-A843-4CA9-848C-17CB4BE9E9F3@multicasttech.com>
	<022b01c70d7e$01803bd0$640fa8c0@cis.neustar.com>
	<8b2769930611210739u2a7a115bne91ccd6805e23313@mail.gmail.com>
Message-ID: <f40963db0611210924l378030e3sc7f4aa9982ad7db2@mail.gmail.com>

I was travelling yesterday, so my apologies for not responding.  Comments
are inline.


1) mobility. We want to preserve the (rapid) update ability of a SIP
> registrar, for example going from home to the office to the mobile
> phone in a period of a few minutes. There is a great deal of concern
> if dynamic DNS could meet these requirements.


This is my biggest area of concern too.  Just the latency of updates.  In my
own use of dynamic DNS, I've known updates to happen extremely quickly.
I've heard of them taking more like 7 minutes to fully propagate though.  To
me this is less a question of mobility and more a question of churn.  How
often are the registrars going up and down?  I think you can influence this
in many ways as well, such as running daemon processes automatically on
system startup.

2) it still requires centralized infrastructure, in the sense that
> some known entity is the server -- even if that isn't a fixed server
> but one of the "peers", it will be doing much of the work.


Correct.  The simplest way to do it would be to just use existing dynamic
DNS providers out there.  I suspect the load on these servers is extremely
small because of the "distributed" nature of DNS, but I've never run them
myself so can't say for sure.  You'd still be decentralizing the heavy
lifting of the registrars and SIP proxies, and that certainly seems like the
bulk of the work.

3) if it is using the existing DNS approach, it requires Internet
> connectivity -- not very useful for disconnected scenarios.



I certainly agree with you there, although I must say I'm confused on
exactly how and where this system would ever realistically be deployed in
disconnected scenarios.  I can see it for disaster relief, and I've worked
extensively with disaster relief groups at the DoD and elsewhere around how
to use SIP in those scenarios, but I think that's a miniscule fraction of
the realistic use cases.  It's a cool one to think about, but I'm not
convinced on the practicality of it.


The charter that went around for several months was a reflection of
> this -- it was explicitly ruled out of scope as item 4. It is very
> interesting as an approach for resolution, but doesn't meet the needs
> of the folks looking at this WG because of the points above.


I just read that.  I have to apologize because I only participate in the
group intermittently, and I'm certainly not trying to derail the WG
efforts.  I'm not at all sure why it's out of scope though.  If it's a
simpler system that works better for 98% of the realistic use cases, it
seems in scope to me.  Does it work better is the real question, and your
point #1 is the biggest issue in my mind.

-Adam
-------------- next part --------------
An HTML attachment was scrubbed...
URL: https://lists.cs.columbia.edu/pipermail/p2p-sip/attachments/20061121/1c6eb228/attachment.html

From br at brianrosen.net  Tue Nov 21 12:48:49 2006
From: br at brianrosen.net (Brian Rosen)
Date: Tue, 21 Nov 2006 12:48:49 -0500
Subject: [p2p-sip] what's wrong with DNS?
In-Reply-To: <8b2769930611210739u2a7a115bne91ccd6805e23313@mail.gmail.com>
Message-ID: <02cc01c70d95$4fa685f0$640fa8c0@cis.neustar.com>

I don't want to beat a dead horse, but:
1. Most of the dynamic DNS implementations I'm familiar with have update
times measured in a couple of minutes.  It's not "real time", but it's
pretty short.  This is reflected in the TTL of the answer.

2. Any form of DNS requires a set of caching resolvers and authoritative
servers.  If you call that "centralized", then you are correct.  I always
think of DNS as being very decentralized except for the roots, and most
implementations I know work if connection to the root is lost, but the
answer is in the cache. 

3. If you combine a short TTL with a disconnected server, then caching
doesn't work; resolvers need access to the authoritative servers (plural)
that provide the data.  It is servers plural, not singular.  I think this is
the reason dynamic DNS is not the best choice, although we may have been
hasty in coming to that conclusion.

4. I think identifiers in these systems are in the form of user at domain,
where domain is the "Overlay Id".  Humans may never see that, but it's
there.  Resolving "domain" to at least a starter set of peers has to happen
somehow.  We have only a few tricks in our bag for that:
	a. Configuration
	b. Multicast
	c. DNS
Are there other mechanisms?  There are multiple configuration mechanisms,
but they all depend on static information, not dynamic, learned data.

Agree that we explicitly declared dynamic DNS and multicast out of scope.
That leaves basic DNS and Configuration.

When networks become disconnected, local caching resolvers work.  If you are
within a domain, and you want a DNS answer from that domain, it works.  

Many existing systems work by configuration.  I'd rather not do that.

Brian

> -----Original Message-----
> From: davidbryan at gmail.com [mailto:davidbryan at gmail.com] On Behalf Of
> David A. Bryan
> Sent: Tuesday, November 21, 2006 10:39 AM
> To: Brian Rosen
> Cc: Marshall Eubanks; p2p-sip at cs.columbia.edu; Mehdi MANI
> Subject: Re: [p2p-sip] what's wrong with DNS?
> 
> We've actually discussed this several times at several ad-hocs, and
> the general conclusion of the group was that dynamic DNS did not meet
> the needs on several fronts:
> 
> 1) mobility. We want to preserve the (rapid) update ability of a SIP
> registrar, for example going from home to the office to the mobile
> phone in a period of a few minutes. There is a great deal of concern
> if dynamic DNS could meet these requirements.
> 
> 2) it still requires centralized infrastructure, in the sense that
> some known entity is the server -- even if that isn't a fixed server
> but one of the "peers", it will be doing much of the work.
> 
> 3) if it is using the existing DNS approach, it requires Internet
> connectivity -- not very useful for disconnected scenarios.
> 
> The charter that went around for several months was a reflection of
> this -- it was explicitly ruled out of scope as item 4. It is very
> interesting as an approach for resolution, but doesn't meet the needs
> of the folks looking at this WG because of the points above.
> 
> David
> 
> On 11/21/06, Brian Rosen <br at brianrosen.net> wrote:
> > I meant use of existing mechanisms such as DNS rather than inventing new
> > ones.
> >
> > I think the dynamic DNS idea can work.  I think simpler NAPTR based
> > mechanisms can work (especially if, as I think we're heading for, we
> have a
> > "domain" for the community the P2P operates over, where we support
> multiple
> > such domains (the concepts document calls these "overlay IDs").
> >
> > I'm sanguine about multicast mechanisms because multicast in general
> doesn't
> > work across the Internet :(
> >
> > I don't think we need new RRs.
> >
> > Brian
> >
> > > -----Original Message-----
> > > From: Marshall Eubanks [mailto:tme at multicasttech.com]
> > > Sent: Tuesday, November 21, 2006 9:45 AM
> > > To: Brian Rosen
> > > Cc: 'David Barrett'; 'Mehdi MANI'; 'Adam Fisk'; p2p-
> sip at cs.columbia.edu
> > > Subject: Re: [p2p-sip] what's wrong with DNS?
> > >
> > > Hello;
> > >
> > > On Nov 21, 2006, at 9:21 AM, Brian Rosen wrote:
> > >
> > > > Agree
> > > >
> > > > Let's stick to things we know will work.
> > > >
> > > > Brian
> > > >
> > >
> > > I am not sure what is actually being proposed here, so I am not sure
> > > what you are saying will work.
> > >
> > > To use Dave Singer's mDNS ideas for LAN resource discovery ? Cool.
> > > There are RFC's - apply.
> > >
> > > To extend these ideas, maybe using P2P application layer multicast,
> > > to the wide area - sounds
> > > like an interesting idea, and could be very cool. I could see this
> > > being used all over the place.
> > >
> > > To expand the core DNS into P2P discovery - i.e., to have P2P RR's
> > > added into the
> > > root servers ? Is not a good idea, and (I dare say) will not happen
> > > in the IETF.
> > >
> > > I think Michael Dillon put it well in DKIM :
> > >
> > > "if someone wants to leverage the DNS protocol and
> > > DNS server software in order to operate a global
> > > distributed database, then there are no barriers to
> > > doing this on ports other than port 53. The IETF barely
> > > needs to be involved at all other than defining the
> > > new RR.
> > >
> > > But when someone suggests that port 53 servers should
> > > all support some new RR or anything else that is new,
> > > now we are talking about a major upgrade to the Internet's
> > > mission critical infrastructure impacting millions of
> > > people worldwide. Here the IETF has a major role to play
> > > and the IETF should tread very carefully."
> > >
> > > So, I would like to see a clearer description of just what path is
> > > actually being proposed here.
> > >
> > > Regards
> > > Marshall
> > >
> > >
> > > >> -----Original Message-----
> > > >> From: p2p-sip-bounces at cs.columbia.edu [mailto:p2p-sip-
> > > >> bounces at cs.columbia.edu] On Behalf Of David Barrett
> > > >> Sent: Tuesday, November 21, 2006 12:33 AM
> > > >> To: Mehdi MANI; Adam Fisk
> > > >> Cc: p2p-sip at cs.columbia.edu
> > > >> Subject: Re: [p2p-sip] what's wrong with DNS?
> > > >>
> > > >> On Mon, 20 Nov 2006 9:19 pm, Mehdi MANI wrote:
> > > >>>
> > > >>> And the second approach, will be something like the architecture
> of
> > > >>> Napster that
> > > >>> was not scalable enough.
> > > >>
> > > >> If it weren't for the lawsuits, nobody would have ever bothered
> with
> > > >> decentralized indexing.  Scalability was not the problem.
> > > >>
> > > >> I like Adam's idea.  DNS works and scales in the real world; it's a
> > > >> proven solution.  For totally disconnected, ad hoc networks,
> > > >> perhaps we
> > > >> should start a P2P-DNS WG and solve that problem by itself.  That's
> > > >> bigger than SIP and might benefit from being separated out.
> > > >>
> > > >> -david
> > > >> _______________________________________________
> > > >> p2p-sip mailing list
> > > >> p2p-sip at cs.columbia.edu
> > > >> https://lists.cs.columbia.edu/cucslists/listinfo/p2p-sip
> > > >
> > > > _______________________________________________
> > > > p2p-sip mailing list
> > > > p2p-sip at cs.columbia.edu
> > > > https://lists.cs.columbia.edu/cucslists/listinfo/p2p-sip
> >
> > _______________________________________________
> > p2p-sip mailing list
> > p2p-sip at cs.columbia.edu
> > https://lists.cs.columbia.edu/cucslists/listinfo/p2p-sip
> >


From adamfisk at gmail.com  Tue Nov 21 12:46:38 2006
From: adamfisk at gmail.com (Adam Fisk)
Date: Tue, 21 Nov 2006 12:46:38 -0500
Subject: [p2p-sip] what's wrong with DNS?
In-Reply-To: <022b01c70d7e$01803bd0$640fa8c0@cis.neustar.com>
References: <2F405B1C-A843-4CA9-848C-17CB4BE9E9F3@multicasttech.com>
	<022b01c70d7e$01803bd0$640fa8c0@cis.neustar.com>
Message-ID: <f40963db0611210946v6d48b8f7xf948f4b8bc617926@mail.gmail.com>

On 11/21/06, Brian Rosen <br at brianrosen.net> wrote:
>
> I meant use of existing mechanisms such as DNS rather than inventing new
> ones.


That's certainly my thinking in bringing this up.  Just do what's easy and
knock off the biggest win -- a low-cost SIP infrastructure.

As I see it, this approach has the following advantages:

1) Zero changes to proxy servers and registrars.
    - Non-firewalled peers become proxies and registrars just like before.
They should be able to literally use the same code as the open source
proxies and registrars out there, though, because nothing changes in the
lookup.

2) Eliminating the most complicated part of the current architecture -- the
DHT.  It's also the thorniest to standardize.

3) Lower latency for completing calls.  No more waiting around for the DHT
to respond.  Call completion happens just like it does with any other SIP
call (if the DNS is up-to-date -- more below).


Then there are the disadvantages:

1) Delay of DNS propagation when registrars go down or clients move.  I
think this is the toughest to get a good read on because I don't know how it
would play out in the field.

2) You'd need to run a dynamic DNS *client* on every node (at least that
seems like the best approach to me).

3) You'd be relying on some servers out there to do the dynamic DNS.
Possibly existing dynamic DNS providers?

4) Doesn't work in completely disconnected scenarios.

Keep in mind I'm not necessarily advocating this shift, but it certainly
resonated with me once I realized the full implications of Paul's question
to me the other day.

-Adam
-------------- next part --------------
An HTML attachment was scrubbed...
URL: https://lists.cs.columbia.edu/pipermail/p2p-sip/attachments/20061121/1e4764fa/attachment.html

From br at brianrosen.net  Tue Nov 21 13:00:07 2006
From: br at brianrosen.net (Brian Rosen)
Date: Tue, 21 Nov 2006 13:00:07 -0500
Subject: [p2p-sip] what's wrong with DNS?
In-Reply-To: <f40963db0611210946v6d48b8f7xf948f4b8bc617926@mail.gmail.com>
Message-ID: <02e701c70d96$e86de480$640fa8c0@cis.neustar.com>

You are going out way too far for me.

 

I want a simple mechanism to learn peers to get started (bootstrap), not a
replacement for the DHT.

 

The fundamental problem I think you have is that if you have a classic
registrar, if that registrar fails, you are out of luck.  The DHT stays
running.  I don't think you are going to find a lot of support for this line
of thinking.  

 

Brian

 

  _____  

From: Adam Fisk [mailto:adamfisk at gmail.com] 
Sent: Tuesday, November 21, 2006 12:47 PM
To: Brian Rosen
Cc: Marshall Eubanks; David Barrett; Mehdi MANI; p2p-sip at cs.columbia.edu
Subject: Re: [p2p-sip] what's wrong with DNS?

 

On 11/21/06, Brian Rosen <br at brianrosen.net> wrote:

I meant use of existing mechanisms such as DNS rather than inventing new
ones.


That's certainly my thinking in bringing this up.  Just do what's easy and
knock off the biggest win -- a low-cost SIP infrastructure. 

As I see it, this approach has the following advantages:

1) Zero changes to proxy servers and registrars.
    - Non-firewalled peers become proxies and registrars just like before.
They should be able to literally use the same code as the open source
proxies and registrars out there, though, because nothing changes in the
lookup. 

2) Eliminating the most complicated part of the current architecture -- the
DHT.  It's also the thorniest to standardize.

3) Lower latency for completing calls.  No more waiting around for the DHT
to respond.  Call completion happens just like it does with any other SIP
call (if the DNS is up-to-date -- more below). 



Then there are the disadvantages:

1) Delay of DNS propagation when registrars go down or clients move.  I
think this is the toughest to get a good read on because I don't know how it
would play out in the field. 

2) You'd need to run a dynamic DNS *client* on every node (at least that
seems like the best approach to me).

3) You'd be relying on some servers out there to do the dynamic DNS.
Possibly existing dynamic DNS providers?  

4) Doesn't work in completely disconnected scenarios.

Keep in mind I'm not necessarily advocating this shift, but it certainly
resonated with me once I realized the full implications of Paul's question
to me the other day. 

-Adam

-------------- next part --------------
An HTML attachment was scrubbed...
URL: https://lists.cs.columbia.edu/pipermail/p2p-sip/attachments/20061121/2e677995/attachment.html

From adamfisk at gmail.com  Tue Nov 21 13:30:14 2006
From: adamfisk at gmail.com (Adam Fisk)
Date: Tue, 21 Nov 2006 13:30:14 -0500
Subject: [p2p-sip] what's wrong with DNS?
In-Reply-To: <02e701c70d96$e86de480$640fa8c0@cis.neustar.com>
References: <f40963db0611210946v6d48b8f7xf948f4b8bc617926@mail.gmail.com>
	<02e701c70d96$e86de480$640fa8c0@cis.neustar.com>
Message-ID: <f40963db0611211030n79457435p459636a0af71eb5d@mail.gmail.com>

On 11/21/06, Brian Rosen <br at brianrosen.net> wrote:
>
>  You are going out way too far for me.
>

I'm again just tossing around ideas here, but from a more traditional
perspective outside of this working group, I'd argue the dynamic DNS
approach is the much less "far out" of the two.  That's what attracted me to
it in the first place.  You're using off-the-shelf working technology to
solve a known problem.  The p2p part of SIP becomes just having some way to
find all the non-firewalled nodes out there running the service, a problem
you have in any scenario.

The fundamental problem I think you have is that if you have a classic
> registrar, if that registrar fails, you are out of luck.  The DHT stays
> running.  I don't think you are going to find a lot of support for this line
> of thinking.
>

I'd say more precisely that there are non-firewalled nodes out there that
stay running.  If you lose your connection to a registrar, you just connect
to another registrar.  If you lose your connection to a DHT node, it's the
same thing -- you just connect to another one.

The fundamental robustness comes from the fact that there will always be
non-firewalled peers who can connect you to the network, not that they're
structured into a DHT or not.

-Adam
-------------- next part --------------
An HTML attachment was scrubbed...
URL: https://lists.cs.columbia.edu/pipermail/p2p-sip/attachments/20061121/4016256d/attachment.html

From br at brianrosen.net  Tue Nov 21 13:53:44 2006
From: br at brianrosen.net (Brian Rosen)
Date: Tue, 21 Nov 2006 13:53:44 -0500
Subject: [p2p-sip] what's wrong with DNS?
In-Reply-To: <f40963db0611211030n79457435p459636a0af71eb5d@mail.gmail.com>
Message-ID: <030601c70d9e$60cbc7b0$640fa8c0@cis.neustar.com>

I don't know what you are smoking.

 

The point of a Distributed Hash Table is that no one node has the whole
thing, and the network can maintain the database if individual nodes go
down.  Any form of classic SIP registration doesn't do anything like that;
you are registered at ONE registrar, and if that registrar becomes
unavailable, you can't get calls.   Commercial implementations often use
redundant databases and proxies to make their registration service reliable,
but you can' use classic registration to do the kind of peer to peer things
a DHT will do.

 

The problem isn't necessarily you losing connection to a registrar, it's
your peer being unable to connect to your registrar and thus being unable to
contact you.

 

The fundamental robustness is that the database is distributed and
replicated; any node can get you to any part of the database even when peers
come and go.

 

Brian

 

  _____  

From: Adam Fisk [mailto:adamfisk at gmail.com] 
Sent: Tuesday, November 21, 2006 1:30 PM
To: Brian Rosen
Cc: Marshall Eubanks; David Barrett; Mehdi MANI; p2p-sip at cs.columbia.edu
Subject: Re: [p2p-sip] what's wrong with DNS?

 

 

On 11/21/06, Brian Rosen <br at brianrosen.net> wrote:

You are going out way too far for me.


I'm again just tossing around ideas here, but from a more traditional
perspective outside of this working group, I'd argue the dynamic DNS
approach is the much less "far out" of the two.  That's what attracted me to
it in the first place.  You're using off-the-shelf working technology to
solve a known problem.  The p2p part of SIP becomes just having some way to
find all the non-firewalled nodes out there running the service, a problem
you have in any scenario.  

 

The fundamental problem I think you have is that if you have a classic
registrar, if that registrar fails, you are out of luck.  The DHT stays
running.  I don't think you are going to find a lot of support for this line
of thinking. 


I'd say more precisely that there are non-firewalled nodes out there that
stay running.  If you lose your connection to a registrar, you just connect
to another registrar.  If you lose your connection to a DHT node, it's the
same thing -- you just connect to another one.  

The fundamental robustness comes from the fact that there will always be
non-firewalled peers who can connect you to the network, not that they're
structured into a DHT or not.
 
-Adam

 

 

-------------- next part --------------
An HTML attachment was scrubbed...
URL: https://lists.cs.columbia.edu/pipermail/p2p-sip/attachments/20061121/f1c36622/attachment-0001.html

From spencer at mcsr-labs.org  Tue Nov 21 14:02:48 2006
From: spencer at mcsr-labs.org (Spencer Dawkins)
Date: Tue, 21 Nov 2006 13:02:48 -0600
Subject: [p2p-sip] what's wrong with DNS?
References: <02e701c70d96$e86de480$640fa8c0@cis.neustar.com>
Message-ID: <076c01c70d9f$a3773c10$6a01a8c0@china.huawei.com>

Hi, Brian,

I'm not sure who else may have misunderstood - when I saw your note that said

> Agree
> 
> Let's stick to things we know will work.  
> 
> Brian

I thought you were talking about a replacement for the DHT, and others may have as well.

Sorry for (my) misunderstanding - I was confused by the thread you were following up in.

Thanks,

Spencer
  ----- Original Message ----- 
  From: Brian Rosen 
  To: 'Adam Fisk' 
  Cc: p2p-sip at cs.columbia.edu ; 'Mehdi MANI' 
  Sent: Tuesday, November 21, 2006 12:00 PM
  Subject: Re: [p2p-sip] what's wrong with DNS?


  You are going out way too far for me.

   

  I want a simple mechanism to learn peers to get started (bootstrap), not a replacement for the DHT.

   

  The fundamental problem I think you have is that if you have a classic registrar, if that registrar fails, you are out of luck.  The DHT stays running.  I don't think you are going to find a lot of support for this line of thinking.  

   

  Brian

   


------------------------------------------------------------------------------

  From: Adam Fisk [mailto:adamfisk at gmail.com] 
  Sent: Tuesday, November 21, 2006 12:47 PM
  To: Brian Rosen
  Cc: Marshall Eubanks; David Barrett; Mehdi MANI; p2p-sip at cs.columbia.edu
  Subject: Re: [p2p-sip] what's wrong with DNS?

   

  On 11/21/06, Brian Rosen <br at brianrosen.net> wrote:

    I meant use of existing mechanisms such as DNS rather than inventing new
    ones.


  That's certainly my thinking in bringing this up.  Just do what's easy and knock off the biggest win -- a low-cost SIP infrastructure. 

  As I see it, this approach has the following advantages:

  1) Zero changes to proxy servers and registrars.
      - Non-firewalled peers become proxies and registrars just like before.  They should be able to literally use the same code as the open source proxies and registrars out there, though, because nothing changes in the lookup. 

  2) Eliminating the most complicated part of the current architecture -- the DHT.  It's also the thorniest to standardize.

  3) Lower latency for completing calls.  No more waiting around for the DHT to respond.  Call completion happens just like it does with any other SIP call (if the DNS is up-to-date -- more below). 



  Then there are the disadvantages:

  1) Delay of DNS propagation when registrars go down or clients move.  I think this is the toughest to get a good read on because I don't know how it would play out in the field. 

  2) You'd need to run a dynamic DNS *client* on every node (at least that seems like the best approach to me).

  3) You'd be relying on some servers out there to do the dynamic DNS.  Possibly existing dynamic DNS providers?  

  4) Doesn't work in completely disconnected scenarios.

  Keep in mind I'm not necessarily advocating this shift, but it certainly resonated with me once I realized the full implications of Paul's question to me the other day. 

  -Adam



------------------------------------------------------------------------------


  _______________________________________________
  p2p-sip mailing list
  p2p-sip at cs.columbia.edu
  https://lists.cs.columbia.edu/cucslists/listinfo/p2p-sip
-------------- next part --------------
An HTML attachment was scrubbed...
URL: https://lists.cs.columbia.edu/pipermail/p2p-sip/attachments/20061121/cd5029d6/attachment.html

From br at brianrosen.net  Tue Nov 21 14:22:20 2006
From: br at brianrosen.net (Brian Rosen)
Date: Tue, 21 Nov 2006 14:22:20 -0500
Subject: [p2p-sip] what's wrong with DNS?
In-Reply-To: <076c01c70d9f$a3773c10$6a01a8c0@china.huawei.com>
Message-ID: <032d01c70da2$61b8ea00$640fa8c0@cis.neustar.com>

Sorry that I confused people.  I was thinking about bootstrap and not DHT
replacement.

 

Brian

 

  _____  

From: p2p-sip-bounces at cs.columbia.edu
[mailto:p2p-sip-bounces at cs.columbia.edu] On Behalf Of Spencer Dawkins
Sent: Tuesday, November 21, 2006 2:03 PM
To: p2p-sip at cs.columbia.edu
Subject: Re: [p2p-sip] what's wrong with DNS?

 

Hi, Brian,

 

I'm not sure who else may have misunderstood - when I saw your note that
said

 

> Agree
> 
> Let's stick to things we know will work.  
> 
> Brian

 

I thought you were talking about a replacement for the DHT, and others may
have as well.

 

Sorry for (my) misunderstanding - I was confused by the thread you were
following up in.

 

Thanks,

 

Spencer

----- Original Message ----- 

From: Brian Rosen <mailto:br at brianrosen.net>  

To: 'Adam Fisk' <mailto:adamfisk at gmail.com>  

Cc: p2p-sip at cs.columbia.edu ; 'Mehdi <mailto:Mehdi.Mani at int-evry.fr>  MANI' 

Sent: Tuesday, November 21, 2006 12:00 PM

Subject: Re: [p2p-sip] what's wrong with DNS?

 

You are going out way too far for me.

 

I want a simple mechanism to learn peers to get started (bootstrap), not a
replacement for the DHT.

 

The fundamental problem I think you have is that if you have a classic
registrar, if that registrar fails, you are out of luck.  The DHT stays
running.  I don't think you are going to find a lot of support for this line
of thinking.  

 

Brian

 


  _____  


From: Adam Fisk [mailto:adamfisk at gmail.com] 
Sent: Tuesday, November 21, 2006 12:47 PM
To: Brian Rosen
Cc: Marshall Eubanks; David Barrett; Mehdi MANI; p2p-sip at cs.columbia.edu
Subject: Re: [p2p-sip] what's wrong with DNS?

 

On 11/21/06, Brian Rosen <br at brianrosen.net> wrote:

I meant use of existing mechanisms such as DNS rather than inventing new
ones.


That's certainly my thinking in bringing this up.  Just do what's easy and
knock off the biggest win -- a low-cost SIP infrastructure. 

As I see it, this approach has the following advantages:

1) Zero changes to proxy servers and registrars.
    - Non-firewalled peers become proxies and registrars just like before.
They should be able to literally use the same code as the open source
proxies and registrars out there, though, because nothing changes in the
lookup. 

2) Eliminating the most complicated part of the current architecture -- the
DHT.  It's also the thorniest to standardize.

3) Lower latency for completing calls.  No more waiting around for the DHT
to respond.  Call completion happens just like it does with any other SIP
call (if the DNS is up-to-date -- more below). 



Then there are the disadvantages:

1) Delay of DNS propagation when registrars go down or clients move.  I
think this is the toughest to get a good read on because I don't know how it
would play out in the field. 

2) You'd need to run a dynamic DNS *client* on every node (at least that
seems like the best approach to me).

3) You'd be relying on some servers out there to do the dynamic DNS.
Possibly existing dynamic DNS providers?  

4) Doesn't work in completely disconnected scenarios.

Keep in mind I'm not necessarily advocating this shift, but it certainly
resonated with me once I realized the full implications of Paul's question
to me the other day. 

-Adam


  _____  


_______________________________________________
p2p-sip mailing list
p2p-sip at cs.columbia.edu
https://lists.cs.columbia.edu/cucslists/listinfo/p2p-sip

-------------- next part --------------
An HTML attachment was scrubbed...
URL: https://lists.cs.columbia.edu/pipermail/p2p-sip/attachments/20061121/88873614/attachment-0001.html

From adamfisk at gmail.com  Tue Nov 21 14:25:35 2006
From: adamfisk at gmail.com (Adam Fisk)
Date: Tue, 21 Nov 2006 14:25:35 -0500
Subject: [p2p-sip] what's wrong with DNS?
In-Reply-To: <030601c70d9e$60cbc7b0$640fa8c0@cis.neustar.com>
References: <f40963db0611211030n79457435p459636a0af71eb5d@mail.gmail.com>
	<030601c70d9e$60cbc7b0$640fa8c0@cis.neustar.com>
Message-ID: <f40963db0611211125t3bea2aebo5f4bf59a48956d8a@mail.gmail.com>

>  I don't know what you are smoking.
>

The good weed.

The point of a Distributed Hash Table is that no one node has the whole
> thing, and the network can maintain the database if individual nodes go
> down.  Any form of classic SIP registration doesn't do anything like that;
> you are registered at ONE registrar, and if that registrar becomes
> unavailable, you can't get calls.   Commercial implementations often use
> redundant databases and proxies to make their registration service reliable,
> but you can' use classic registration to do the kind of peer to peer things
> a DHT will do.
>

OK, let's step back for a moment.  I'm making a pretty simple suggestion
here to offer what to me is the primary benefit of p2p SIP, namely low cost
telephony for the masses, with a much simpler implementation.  In this
system, every node would be a dynamic DNS client.  Every node would be
registered with some non-firewalled peer(s) out there.  If that connection
is severed, the client connects to a new non-firewalled registrar and
updates the DNS.  It's as simple as that.

The problem isn't necessarily you losing connection to a registrar, it's
> your peer being unable to connect to your registrar and thus being unable to
> contact you.
>

That's where DNS comes in.  Your peer just calls you as your peer normally
would.

The fundamental robustness is that the database is distributed and
> replicated; any node can get you to any part of the database even when peers
> come and go.
>

The system I just described does the same thing, and it does it using
existing technology that's far less complicated.

-Adam
-------------- next part --------------
An HTML attachment was scrubbed...
URL: https://lists.cs.columbia.edu/pipermail/p2p-sip/attachments/20061121/bf3f4df5/attachment.html

From ggb at tid.es  Tue Nov 21 14:50:08 2006
From: ggb at tid.es (=?iso-8859-1?Q?Gustavo_Garc=EDa_Bernardo?=)
Date: Tue, 21 Nov 2006 20:50:08 +0100
Subject: [p2p-sip] what's wrong with DNS?
In-Reply-To: <f40963db0611210924l378030e3sc7f4aa9982ad7db2@mail.gmail.com>
Message-ID: <009e01c70da6$3fe14a90$01595f0a@hi.inet>


I certainly agree with you there, although I must say I'm confused on
exactly how and where this system would ever realistically be deployed in
disconnected scenarios.  

A meeting room without internet connectivity.

A corporate LAN with the ?dynamic dns update? protocol filtered in their
firewalls.

Vehicle2Vehicle communications or other kinds of MANETs.

 

[http://www.p2psip.org/drafts/draft-bryan-sipping-p2p-usecases-00.txt]

 

-------------- next part --------------
An HTML attachment was scrubbed...
URL: https://lists.cs.columbia.edu/pipermail/p2p-sip/attachments/20061121/e67c015b/attachment.html

From victor.pascual.avila at gmail.com  Tue Nov 21 14:55:41 2006
From: victor.pascual.avila at gmail.com (=?ISO-8859-1?Q?Victor_Pascual_=C1vila?=)
Date: Tue, 21 Nov 2006 20:55:41 +0100
Subject: [p2p-sip] what's wrong with DNS?
In-Reply-To: <009e01c70da6$3fe14a90$01595f0a@hi.inet>
References: <f40963db0611210924l378030e3sc7f4aa9982ad7db2@mail.gmail.com>
	<009e01c70da6$3fe14a90$01595f0a@hi.inet>
Message-ID: <618e24240611211155q18de8fd3xfce234a1675242f8@mail.gmail.com>

2006/11/21, Gustavo Garc?a Bernardo <ggb at tid.es>:
>
>
>
>
>
>
>
>
>  I certainly agree with you there, although I must say I'm confused on
> exactly how and where this system would ever realistically be deployed in
> disconnected scenarios.
>
> A meeting room without internet connectivity.
>
> A corporate LAN with the "dynamic dns update" protocol filtered in their
> firewalls.
>
> Vehicle2Vehicle communications or other kinds of MANETs.
>
When no infrastructure is available... is it correct?
>
>
> [http://www.p2psip.org/drafts/draft-bryan-sipping-p2p-usecases-00.txt]
>
>
> _______________________________________________
> p2p-sip mailing list
> p2p-sip at cs.columbia.edu
> https://lists.cs.columbia.edu/cucslists/listinfo/p2p-sip
>
>
>


-- 
Victor Pascual ?vila

mail:victor.pascual.avila at gmail.com
sip:victor.pascual at iptel.org
linux user: 224443


From slavitch at gmail.com  Tue Nov 21 15:01:42 2006
From: slavitch at gmail.com (Michael Slavitch)
Date: Tue, 21 Nov 2006 15:01:42 -0500
Subject: [p2p-sip] Fwd:  what's wrong with DNS?
In-Reply-To: <1fc32c8f0611211200j3f4bc570kf9382bb3664e5bf6@mail.gmail.com>
References: <f40963db0611210924l378030e3sc7f4aa9982ad7db2@mail.gmail.com>
	<009e01c70da6$3fe14a90$01595f0a@hi.inet>
	<618e24240611211155q18de8fd3xfce234a1675242f8@mail.gmail.com>
	<1fc32c8f0611211200j3f4bc570kf9382bb3664e5bf6@mail.gmail.com>
Message-ID: <1fc32c8f0611211201s404c391cu6b2ca2a6c662ef6c@mail.gmail.com>

Trade shows.
Emergency Measures Response.
Search and Rescue.
First Responder.
I am working with a company developing systems for the latter three.
Depending On DNS is a non-starter for them as under their operatingrequirements DNS may no longer exist.
M
On 11/21/06, Victor Pascual ?vila <victor.pascual.avila at gmail.com> wrote:> 2006/11/21, Gustavo Garc?a Bernardo <ggb at tid.es>:> >> >> >> > A meeting room without internet connectivity.> >> > A corporate LAN with the "dynamic dns update" protocol filtered in their> > firewalls.> >> > Vehicle2Vehicle communications or other kinds of MANETs.> >> When no infrastructure is available... is it correct?> >

From br at brianrosen.net  Tue Nov 21 15:08:24 2006
From: br at brianrosen.net (Brian Rosen)
Date: Tue, 21 Nov 2006 15:08:24 -0500
Subject: [p2p-sip] what's wrong with DNS?
In-Reply-To: <f40963db0611211125t3bea2aebo5f4bf59a48956d8a@mail.gmail.com>
Message-ID: <035a01c70da8$d225d4f0$640fa8c0@cis.neustar.com>

I'll play along for just a little bit more, then I have to go home.

 

I'm not sure I understand your model.  You start with a domain name
(alice.example.com).  You dynamically change the IP address of, let's say,
the sip "server" at alice.example.com to the non-firewalled peer address.
Now a caller who calls alice.example.com looks that domain name up in the
DNS, gets your non-firewalled peer's address and sends the call to it.  It
sends the call to you.  Have I got that right?  The DNS entry can't point to
you, generally speaking if you are behind a firewall, you can't get an
arbitrary SIP message sent to you. 

 

What if the non firewalled peer loses connectivity to the caller, but not to
you?  You don't know there is a problem, but a caller can't get to you.  DNS
doesn't help you.  The caller resolves alice.example.com to the
non-firewalled peer, but it can't get the call to the peer.  You don't know,
and he can't tell you.

 

Also, dynamic DNS layers something on top of the static DNS.  The domain
(alice.example.com) has one (okay, a set of) authoritative nameserver(s).
The dynamic part changes the value (in this case, the IP address for the sip
SRV maybe for the domain).  If the namesever(s) is/are not available, you
are up the creek.

 

Brian

 

 

 

 

  _____  

From: Adam Fisk [mailto:adamfisk at gmail.com] 
Sent: Tuesday, November 21, 2006 2:26 PM
To: Brian Rosen
Cc: Marshall Eubanks; David Barrett; Mehdi MANI; p2p-sip at cs.columbia.edu
Subject: Re: [p2p-sip] what's wrong with DNS?

 

 

I don't know what you are smoking.


The good weed. 

 

The point of a Distributed Hash Table is that no one node has the whole
thing, and the network can maintain the database if individual nodes go
down.  Any form of classic SIP registration doesn't do anything like that;
you are registered at ONE registrar, and if that registrar becomes
unavailable, you can't get calls.   Commercial implementations often use
redundant databases and proxies to make their registration service reliable,
but you can' use classic registration to do the kind of peer to peer things
a DHT will do.


OK, let's step back for a moment.  I'm making a pretty simple suggestion
here to offer what to me is the primary benefit of p2p SIP, namely low cost
telephony for the masses, with a much simpler implementation.  In this
system, every node would be a dynamic DNS client.  Every node would be
registered with some non-firewalled peer(s) out there.  If that connection
is severed, the client connects to a new non-firewalled registrar and
updates the DNS.  It's as simple as that. 

 

The problem isn't necessarily you losing connection to a registrar, it's
your peer being unable to connect to your registrar and thus being unable to
contact you.


That's where DNS comes in.  Your peer just calls you as your peer normally
would.

 

The fundamental robustness is that the database is distributed and
replicated; any node can get you to any part of the database even when peers
come and go.


The system I just described does the same thing, and it does it using
existing technology that's far less complicated.  

-Adam 

 

 

-------------- next part --------------
An HTML attachment was scrubbed...
URL: https://lists.cs.columbia.edu/pipermail/p2p-sip/attachments/20061121/02b81405/attachment-0001.html

From slavitch at gmail.com  Tue Nov 21 15:08:56 2006
From: slavitch at gmail.com (Michael Slavitch)
Date: Tue, 21 Nov 2006 15:08:56 -0500
Subject: [p2p-sip] what's wrong with DNS?
In-Reply-To: <1fc32c8f0611211201s404c391cu6b2ca2a6c662ef6c@mail.gmail.com>
References: <f40963db0611210924l378030e3sc7f4aa9982ad7db2@mail.gmail.com>
	<009e01c70da6$3fe14a90$01595f0a@hi.inet>
	<618e24240611211155q18de8fd3xfce234a1675242f8@mail.gmail.com>
	<1fc32c8f0611211200j3f4bc570kf9382bb3664e5bf6@mail.gmail.com>
	<1fc32c8f0611211201s404c391cu6b2ca2a6c662ef6c@mail.gmail.com>
Message-ID: <1fc32c8f0611211208l1f21076bxff9430830f174eb5@mail.gmail.com>

Oh I forgot to add:
Marine nets: (HAM+802.11+WiMax+broadband satellite)Industrial sites.Forestry.3rd world/emerging market.
It goes on.
On 11/21/06, Michael Slavitch <slavitch at gmail.com> wrote:> Trade shows.>> Emergency Measures Response.>> Search and Rescue.>> First Responder.>> I am working with a company developing systems for the latter three.>> Depending On DNS is a non-starter for them as under their operating> requirements DNS may no longer exist.>> M>> On 11/21/06, Victor Pascual ?vila <victor.pascual.avila at gmail.com> wrote:> > 2006/11/21, Gustavo Garc?a Bernardo <ggb at tid.es>:> > >> > >> > >> > > A meeting room without internet connectivity.> > >> > > A corporate LAN with the "dynamic dns update" protocol filtered in their> > > firewalls.> > >> > > Vehicle2Vehicle communications or other kinds of MANETs.> > >> > When no infrastructure is available... is it correct?> > >>

From adamfisk at gmail.com  Tue Nov 21 15:09:01 2006
From: adamfisk at gmail.com (Adam Fisk)
Date: Tue, 21 Nov 2006 15:09:01 -0500
Subject: [p2p-sip] Fwd: what's wrong with DNS?
In-Reply-To: <1fc32c8f0611211201s404c391cu6b2ca2a6c662ef6c@mail.gmail.com>
References: <f40963db0611210924l378030e3sc7f4aa9982ad7db2@mail.gmail.com>
	<009e01c70da6$3fe14a90$01595f0a@hi.inet>
	<618e24240611211155q18de8fd3xfce234a1675242f8@mail.gmail.com>
	<1fc32c8f0611211200j3f4bc570kf9382bb3664e5bf6@mail.gmail.com>
	<1fc32c8f0611211201s404c391cu6b2ca2a6c662ef6c@mail.gmail.com>
Message-ID: <f40963db0611211209o4c3aa936v46a0c6f795b5187f@mail.gmail.com>

These are very important scenarios to be sure.  Maybe there's a different
set of requirements for a completely disconnected environment?

As an implementor of this technology in a connected environment, the dynamic
DNS approach just seems much simpler to me.  I also worry about how much the
added complexity of the DHT routing would deter widespread adoption.

-Adam


On 11/21/06, Michael Slavitch <slavitch at gmail.com> wrote:
>
> Trade shows.
> Emergency Measures Response.
> Search and Rescue.
> First Responder.
> I am working with a company developing systems for the latter three.
> Depending On DNS is a non-starter for them as under their
> operatingrequirements DNS may no longer exist.
> M
> On 11/21/06, Victor Pascual ?vila <victor.pascual.avila at gmail.com> wrote:>
> 2006/11/21, Gustavo Garc?a Bernardo <ggb at tid.es>:> >> >> >> > A meeting
> room without internet connectivity.> >> > A corporate LAN with the "dynamic
> dns update" protocol filtered in their> > firewalls.> >> > Vehicle2Vehicle
> communications or other kinds of MANETs.> >> When no infrastructure is
> available... is it correct?> >
> _______________________________________________
> p2p-sip mailing list
> p2p-sip at cs.columbia.edu
> https://lists.cs.columbia.edu/cucslists/listinfo/p2p-sip
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: https://lists.cs.columbia.edu/pipermail/p2p-sip/attachments/20061121/ac9f9523/attachment.html

From br at brianrosen.net  Tue Nov 21 15:15:36 2006
From: br at brianrosen.net (Brian Rosen)
Date: Tue, 21 Nov 2006 15:15:36 -0500
Subject: [p2p-sip] Fwd:  what's wrong with DNS?
In-Reply-To: <1fc32c8f0611211201s404c391cu6b2ca2a6c662ef6c@mail.gmail.com>
Message-ID: <036d01c70da9$d02bbd30$640fa8c0@cis.neustar.com>

Michael

While I think the idea of replacing the DHT with dynamic DNS is a
non-starter, I don't understand your problem with DNS in general.

I'm working on systems that deal with the later three.  DNS seems to be
something you CAN rely on within a network you have connectivity to.  You
may lose connectivity to the global Internet, but within, say, an incident
network, assuming that there is some DNS infrastructure seems pretty
reasonable.

I have some sympathy with trying to make things work when, say the only
entities in a network are a pair of first responders (a MANET network with
exactly two members).  It should be possible for those two to talk to one
another.  It IS possible to make such systems use DNS.  One or both of them
have to enable DNS servers.  Working without such mechanisms seems to
require multicast, also on our list of no-no's.

How do you think it works?

Brian

> -----Original Message-----
> From: p2p-sip-bounces at cs.columbia.edu [mailto:p2p-sip-
> bounces at cs.columbia.edu] On Behalf Of Michael Slavitch
> Sent: Tuesday, November 21, 2006 3:02 PM
> To: Henry Sinnreich; P2P-SIP
> Subject: [p2p-sip] Fwd: what's wrong with DNS?
> 
> Trade shows.
> Emergency Measures Response.
> Search and Rescue.
> First Responder.
> I am working with a company developing systems for the latter three.
> Depending On DNS is a non-starter for them as under their
> operatingrequirements DNS may no longer exist.
> M
> On 11/21/06, Victor Pascual ?vila <victor.pascual.avila at gmail.com> wrote:>
> 2006/11/21, Gustavo Garc?a Bernardo <ggb at tid.es>:> >> >> >> > A meeting
> room without internet connectivity.> >> > A corporate LAN with the
> "dynamic dns update" protocol filtered in their> > firewalls.> >> >
> Vehicle2Vehicle communications or other kinds of MANETs.> >> When no
> infrastructure is available... is it correct?> >
> _______________________________________________
> p2p-sip mailing list
> p2p-sip at cs.columbia.edu
> https://lists.cs.columbia.edu/cucslists/listinfo/p2p-sip



From dbarrett at quinthar.com  Tue Nov 21 15:29:47 2006
From: dbarrett at quinthar.com (David Barrett)
Date: Tue, 21 Nov 2006 12:29:47 -0800
Subject: [p2p-sip] what's wrong with DNS?
In-Reply-To: <f40963db0611211125t3bea2aebo5f4bf59a48956d8a@mail.gmail.com>
Message-ID: <027a01c70dab$cd2c26e0$6b02a8c0@Quinthar>

Help me to understand the proposal more clearly.  Are you suggesting:

 

1)       Every user chooses a domain name.  Mine might be
'dbarrett.p2psip.org'



2)       When I start my client up, I choose any non-firewalled peer (how do
I find it?) and REGISTER myself with it as something like
'dbarrett at dbarrett.p2psip.org'



3)       After registering with the peer/proxy, I contact some
highly-redundant set of dynamic DNS servers and say "this is the latest IP
address of my proxy, point to it whenever anyone does a DNS lookup on
dbarrett.p2psip.org"



4)       Your off the shelf, totally unaltered SIP client asks DNS for the
IP of 'dbarrett.p2psip.org', gets the latest IP of my proxy, and looks up
the user 'dbarrett at dbarrett.p2psip.org', and then contacts me as normal.



5)       If the peer I chose as my proxy goes down, I detect it immediately
and choose a new non-firewwalled peer (again, how do I find it?), REGISTER
with that as in step (2), and update my dynamic DNS binding as in step (3).



6)       Repeat.

 

Is this correct? 

 

-david

 

PS: And presumably, if my computer is itself non-firewalled, I can act as my
own proxy; then dbarrett.p2psip.org would point directly to me, and I
wouldn't depend on anyone.

 

 

  _____  

From: Adam Fisk [mailto:adamfisk at gmail.com] 
Sent: Tuesday, November 21, 2006 11:26 AM
To: Brian Rosen
Cc: Marshall Eubanks; David Barrett; Mehdi MANI; p2p-sip at cs.columbia.edu
Subject: Re: [p2p-sip] what's wrong with DNS?

 

 

I don't know what you are smoking.


The good weed. 

 

The point of a Distributed Hash Table is that no one node has the whole
thing, and the network can maintain the database if individual nodes go
down.  Any form of classic SIP registration doesn't do anything like that;
you are registered at ONE registrar, and if that registrar becomes
unavailable, you can't get calls.   Commercial implementations often use
redundant databases and proxies to make their registration service reliable,
but you can' use classic registration to do the kind of peer to peer things
a DHT will do.


OK, let's step back for a moment.  I'm making a pretty simple suggestion
here to offer what to me is the primary benefit of p2p SIP, namely low cost
telephony for the masses, with a much simpler implementation.  In this
system, every node would be a dynamic DNS client.  Every node would be
registered with some non-firewalled peer(s) out there.  If that connection
is severed, the client connects to a new non-firewalled registrar and
updates the DNS.  It's as simple as that. 

 

The problem isn't necessarily you losing connection to a registrar, it's
your peer being unable to connect to your registrar and thus being unable to
contact you.


That's where DNS comes in.  Your peer just calls you as your peer normally
would.

 

The fundamental robustness is that the database is distributed and
replicated; any node can get you to any part of the database even when peers
come and go.


The system I just described does the same thing, and it does it using
existing technology that's far less complicated.  

-Adam 

 

 

-------------- next part --------------
An HTML attachment was scrubbed...
URL: https://lists.cs.columbia.edu/pipermail/p2p-sip/attachments/20061121/af38c105/attachment-0001.html

From slavitch at gmail.com  Tue Nov 21 15:33:36 2006
From: slavitch at gmail.com (Michael Slavitch)
Date: Tue, 21 Nov 2006 15:33:36 -0500
Subject: [p2p-sip] Fwd: what's wrong with DNS?
In-Reply-To: <f40963db0611211209o4c3aa936v46a0c6f795b5187f@mail.gmail.com>
References: <f40963db0611210924l378030e3sc7f4aa9982ad7db2@mail.gmail.com>
	<009e01c70da6$3fe14a90$01595f0a@hi.inet>
	<618e24240611211155q18de8fd3xfce234a1675242f8@mail.gmail.com>
	<1fc32c8f0611211200j3f4bc570kf9382bb3664e5bf6@mail.gmail.com>
	<1fc32c8f0611211201s404c391cu6b2ca2a6c662ef6c@mail.gmail.com>
	<f40963db0611211209o4c3aa936v46a0c6f795b5187f@mail.gmail.com>
Message-ID: <1fc32c8f0611211233q180cdf2dh34c6ae580fa58f93@mail.gmail.com>

As far as I understand existing SIP, when combined with a well-donestable DNS database, is already largely peer to peer, and always hasbeen for a end-to-end Internet.  The trouble is that DNS is not stableor universal in application, the Internet is not end to end, and it isgetting more fragmented not less.
Your implementation given your requirements could be fully implementedas peer to peer using existing protocols albeit with the addition ofNAT traversal using ICE and using DDNS for uri location. This worksfine in an environment with a few thousand or even a few hundredthousand of users and devices.  It starts to break down when you go100x that or more. It works only in a single DNS regime.  It doesn'tfederate well with others.
There are several impediments that make such implementations somewhatbrittle in the real world, some of them are technical, (ie NAT) andsome are anti-competitive, ie walled gardens. DNS itself is a form ofwalled garden in of itself.  Can I get the internal DNS table of youremployer? If I gave out mine, I'd be fired, and quick,
Other issues, like the forced used of root-level commercialcertificate authorities, create other problems in a world that is veryad-hoc and not nearly as centrally managed as administrators wouldlike to think.
The world is a self-similar, chaotic, place.  To the top-down worldthat is bad. To the self-organizing world that is a mathematicalrequirement. DNS does not meet the requirements for self-organizingsystems.
SIP/DNS is fine for a top-down Internet. The top-down people are building IMS.
A user-centric environment, where users and communities create theirown namespaces, is exactly counter to the heavily provisionedcentrally authorized centrally administered way of doing things,canonically described by the root servers of DNS.
Military and first-responders are the first to understand  the needfor self-organization, top-down in today's world is the 21st Centuryedition of the Maginot Line.
While the  P2PSIP proposal can use DNS it should not require the useof it.  The best solutions for the generic problems regarding DNSbrittleness and NAT are the ones that allow for flexibility androbustness for the long term.  This allows for a user andapplication-centric edge-defined Internet that can scale into thebillions and hundreds of billions of end-devices.
Henry Sinnreich's presentations on this very issue describe the circleof complexity far better than I can.

Regards
MOn 11/21/06, Adam Fisk <adamfisk at gmail.com> wrote:> These are very important scenarios to be sure.  Maybe there's a different> set of requirements for a completely disconnected environment?>> As an implementor of this technology in a connected environment, the dynamic> DNS approach just seems much simpler to me.  I also worry about how much the> added complexity of the DHT routing would deter widespread adoption.>> -Adam>>> On 11/21/06, Michael Slavitch <slavitch at gmail.com> wrote:> > Trade shows.> > Emergency Measures Response.> > Search and Rescue.> > First Responder.> > I am working with a company developing systems for the latter three.> > Depending On DNS is a non-starter for them as under their> operatingrequirements DNS may no longer exist.> > M> > On 11/21/06, Victor Pascual ?vila <victor.pascual.avila at gmail.com> wrote:>> 2006/11/21, Gustavo Garc?a Bernardo <ggb at tid.es >:> >> >> >> > A meeting> room without internet connectivity.> >> > A corporate LAN with the "dynamic> dns update" protocol filtered in their> > firewalls.> >> > Vehicle2Vehicle> communications or other kinds of MANETs.> >> When no infrastructure is> available... is it correct?> >> > _______________________________________________> > p2p-sip mailing list> > p2p-sip at cs.columbia.edu> > https://lists.cs.columbia.edu/cucslists/listinfo/p2p-sip> >>>

From hsinnrei at adobe.com  Tue Nov 21 15:38:59 2006
From: hsinnrei at adobe.com (Henry Sinnreich)
Date: Tue, 21 Nov 2006 12:38:59 -0800
Subject: [p2p-sip] Fwd:  what's wrong with DNS?
In-Reply-To: <036d01c70da9$d02bbd30$640fa8c0@cis.neustar.com>
Message-ID: <24CCCC428EFEA2469BF046DB3C7A8D2214134B@namail5.corp.adobe.com>

In a disconnected p2p scenario, peers can be provisioned with the IP address of bootstrap peers in the neighborhood. 

There is no need to rely on the DNS except when proxying SIP messages between different p2p systems or when talking p2psip to cs-sip.

For what it's worth, I remember to have seen quite a number of p2p designs and none of them relies on the DNS. See for example http://opendht.org/ and the P2P SIP system reported at Columbia University.

Thanks, Henry

-----Original Message-----
From: Brian Rosen [mailto:br at brianrosen.net] 
Sent: Tuesday, November 21, 2006 2:16 PM
To: 'Michael Slavitch'; 'P2P-SIP'
Cc: Henry Sinnreich
Subject: RE: [p2p-sip] Fwd: what's wrong with DNS?

Michael

While I think the idea of replacing the DHT with dynamic DNS is a
non-starter, I don't understand your problem with DNS in general.

I'm working on systems that deal with the later three.  DNS seems to be
something you CAN rely on within a network you have connectivity to.  You
may lose connectivity to the global Internet, but within, say, an incident
network, assuming that there is some DNS infrastructure seems pretty
reasonable.

I have some sympathy with trying to make things work when, say the only
entities in a network are a pair of first responders (a MANET network with
exactly two members).  It should be possible for those two to talk to one
another.  It IS possible to make such systems use DNS.  One or both of them
have to enable DNS servers.  Working without such mechanisms seems to
require multicast, also on our list of no-no's.

How do you think it works?

Brian

> -----Original Message-----
> From: p2p-sip-bounces at cs.columbia.edu [mailto:p2p-sip-
> bounces at cs.columbia.edu] On Behalf Of Michael Slavitch
> Sent: Tuesday, November 21, 2006 3:02 PM
> To: Henry Sinnreich; P2P-SIP
> Subject: [p2p-sip] Fwd: what's wrong with DNS?
> 
> Trade shows.
> Emergency Measures Response.
> Search and Rescue.
> First Responder.
> I am working with a company developing systems for the latter three.
> Depending On DNS is a non-starter for them as under their
> operatingrequirements DNS may no longer exist.
> M
> On 11/21/06, Victor Pascual ?vila <victor.pascual.avila at gmail.com> wrote:>
> 2006/11/21, Gustavo Garc?a Bernardo <ggb at tid.es>:> >> >> >> > A meeting
> room without internet connectivity.> >> > A corporate LAN with the
> "dynamic dns update" protocol filtered in their> > firewalls.> >> >
> Vehicle2Vehicle communications or other kinds of MANETs.> >> When no
> infrastructure is available... is it correct?> >
> _______________________________________________
> p2p-sip mailing list
> p2p-sip at cs.columbia.edu
> https://lists.cs.columbia.edu/cucslists/listinfo/p2p-sip



From dbarrett at quinthar.com  Tue Nov 21 15:40:19 2006
From: dbarrett at quinthar.com (David Barrett)
Date: Tue, 21 Nov 2006 12:40:19 -0800
Subject: [p2p-sip] what's wrong with DNS?
In-Reply-To: <035a01c70da8$d225d4f0$640fa8c0@cis.neustar.com>
Message-ID: <029101c70dad$42a1c5f0$6b02a8c0@Quinthar>

  _____  

From: Brian Rosen [mailto:br at brianrosen.net] 
Subject: RE: [p2p-sip] what's wrong with DNS?

 

What if the non firewalled peer loses connectivity to the caller, but not to
you?  You don't know there is a problem, but a caller can't get to you.  DNS
doesn't help you.  The caller resolves alice.example.com to the
non-firewalled peer, but it can't get the call to the peer.  You don't know,
and he can't tell you.

 

Also, dynamic DNS layers something on top of the static DNS.  The domain
(alice.example.com) has one (okay, a set of) authoritative nameserver(s).
The dynamic part changes the value (in this case, the IP address for the sip
SRV maybe for the domain).  If the namesever(s) is/are not available, you
are up the creek.

 

Well, if you lose the nameservers every other internet application that
depends on DNS (like, all of them) is totally screwed anyway - it's no
surprise that your VoIP would stop working if your web browser does too.  So
in an internet application, it seems reasonable to assume that the basic
foundation of the internet continues to function.

 

The first question is more interesting: a peer isn't a hardened server in a
solid datacenter, so what if it just flakes out for some peers more than for
others.  This is a really good point.  But I think DNS offers a solution
here, too: just register your client with multiple proxies (as many as you
like) and use dynamic DNS to associate all of those IPs with your domain
(alice.example.com).  When the SIP client gets a list of IPs back from that
DNS lookup it tries each in order.

 

But the real question is how this works "off the net" in an ad-hoc wireless
or other network that lacks nameservers, DNS, and so on.  I don't have a
good answer to that, though any P2P solution might be stuck under the mask
of DNS (and thus transparent to SIP), thereby paving the way for other P2P
applications who will all have the exact same problem.

 

-david

-------------- next part --------------
An HTML attachment was scrubbed...
URL: https://lists.cs.columbia.edu/pipermail/p2p-sip/attachments/20061121/2e02f16c/attachment.html

From adamfisk at gmail.com  Tue Nov 21 15:45:36 2006
From: adamfisk at gmail.com (Adam Fisk)
Date: Tue, 21 Nov 2006 15:45:36 -0500
Subject: [p2p-sip] what's wrong with DNS?
In-Reply-To: <035a01c70da8$d225d4f0$640fa8c0@cis.neustar.com>
References: <f40963db0611211125t3bea2aebo5f4bf59a48956d8a@mail.gmail.com>
	<035a01c70da8$d225d4f0$640fa8c0@cis.neustar.com>
Message-ID: <f40963db0611211245v10e8343eq636308b20ef8ef92@mail.gmail.com>

On 11/21/06, Brian Rosen <br at brianrosen.net> wrote:
>
>  I'll play along for just a little bit more, then I have to go home.
>
Golly gee, thanks for "playing" Brian.  Gotta love mailing lists.


> I'm not sure I understand your model.  You start with a domain name (
> alice.example.com).  You dynamically change the IP address of, let's say,
> the sip "server" at alice.example.com to the non-firewalled peer address.
> Now a caller who calls alice.example.com looks that domain name up in the
> DNS, gets your non-firewalled peer's address and sends the call to it.  It
> sends the call to you.  Have I got that right?
>

Yes.  The non-firewalled peer is the registrar for Alice, say
registrar1.example.com.  It's a 100% standard SIP registrar/proxy.  Alice
would be a SIP "client" connected to registrar1.example.com (the
non-firewalled peer), and Alice would have set the DNS for
alice.example.comto point to
registrar1.example.com.

The DNS entry can't point to you, generally speaking if you are behind a
> firewall, you can't get an arbitrary SIP message sent to you.
>

Right - it's sending it to registrar1.example.com, the non-firewalled peer
that can get an arbitrary SIP message sent to it.

What if the non firewalled peer loses connectivity to the caller, but not to
> you?  You don't know there is a problem, but a caller can't get to you.  DNS
> doesn't help you.  The caller resolves alice.example.com to the
> non-firewalled peer, but it can't get the call to the peer.  You don't know,
> and he can't tell you.
>

Are you talking about during the call?  At this point we'd just be talking
about a completely normal SIP session, so any failures on any end would be
handled as they normally would.  Let's call the caller bob.example.com.
Let's say Bob is registered at registrar2.example.com.  Bob initiates a
call, his completely standard SIP registrar looks up the authority for Alice
in DNS and sends it along.  You're asking what if the SIP session is broken
because the connection between the 2 proxies in the standard SIP trapezoid
is broken?  All of these scenarios just follow the standard state diagrams.

Also, dynamic DNS layers something on top of the static DNS.  The domain (
> alice.example.com) has one (okay, a set of) authoritative nameserver(s).
> The dynamic part changes the value (in this case, the IP address for the sip
> SRV maybe for the domain).  If the namesever(s) is/are not available, you
> are up the creek.
>

I think you'd probably just use standard name servers.  You could use any
public dynamic DNS provider out there.  Sure, they could go out, but that
doesn't happen often.

I think part of the confusion here might be that you need some way of
knowing about all the non-firewalled nodes out there so that when you lose a
connection, you just re-register with another node and remap the DNS.  This
is essentially the bootstrapping problem all over again though.  You just
need some way to know about those nodes.

-Adam
-------------- next part --------------
An HTML attachment was scrubbed...
URL: https://lists.cs.columbia.edu/pipermail/p2p-sip/attachments/20061121/698ea6cf/attachment-0001.html

From slavitch at gmail.com  Tue Nov 21 15:51:45 2006
From: slavitch at gmail.com (Michael Slavitch)
Date: Tue, 21 Nov 2006 15:51:45 -0500
Subject: [p2p-sip] Fwd:  what's wrong with DNS?
In-Reply-To: <1fc32c8f0611211250i7ae67db6pd45f59dc19eec81b@mail.gmail.com>
References: <035a01c70da8$d225d4f0$640fa8c0@cis.neustar.com>
	<029101c70dad$42a1c5f0$6b02a8c0@Quinthar>
	<1fc32c8f0611211246j7d999e24q48bb1c17e06ccf2a@mail.gmail.com>
	<1fc32c8f0611211250i7ae67db6pd45f59dc19eec81b@mail.gmail.com>
Message-ID: <1fc32c8f0611211251r3eeed24ao99d68aab11009a61@mail.gmail.com>

"I think you'd probably just use standard name servers.  You could use
any public dynamic DNS provider out there.  Sure, they could go out,
but that doesn't happen often. "

My own requirements have to assume that the root nameserver no longer exists.

What if a government turned it off?  Or a corporation? What then? What
if there is a flood?
Or an attack? Does the system go dark?

Katrina exposed such loopholes in systems design.  Let's not design
them in after the fact when there are known alternatives that offer
rewards even when things are working well.

From adamfisk at gmail.com  Tue Nov 21 15:54:05 2006
From: adamfisk at gmail.com (Adam Fisk)
Date: Tue, 21 Nov 2006 15:54:05 -0500
Subject: [p2p-sip] what's wrong with DNS?
In-Reply-To: <027a01c70dab$cd2c26e0$6b02a8c0@Quinthar>
References: <f40963db0611211125t3bea2aebo5f4bf59a48956d8a@mail.gmail.com>
	<027a01c70dab$cd2c26e0$6b02a8c0@Quinthar>
Message-ID: <f40963db0611211254v5f1f839bg1aa02b069ff655cb@mail.gmail.com>

Hi David-

>
> 1)       Every user chooses a domain name.  Mine might be
> 'dbarrett.p2psip.org'
>

Yes.

2)       When I start my client up, I choose any non-firewalled peer (how do
> I find it?) and REGISTER myself with it as something like 'dbarrett@
> dbarrett.p2psip.org'
>

Yes.  You'd find it in the same way you would with any of the bootstrap
methods.

3)       After registering with the peer/proxy, I contact some
> highly-redundant set of dynamic DNS servers and say "this is the latest IP
> address of my proxy, point to it whenever anyone does a DNS lookup on
> dbarrett.p2psip.org"
>

Yeah.  You could even just use existing free dynamic DNS providers out
there.

4)       Your off the shelf, totally unaltered SIP client asks DNS for the
> IP of 'dbarrett.p2psip.org', gets the latest IP of my proxy, and looks up
> the user 'dbarrett at dbarrett.p2psip.org', and then contacts me as normal.
>

Yes again.  At this stage it would be a standard SIP call with the standard
trapezoid.

5)       If the peer I chose as my proxy goes down, I detect it immediately
> and choose a new non-firewwalled peer (again, how do I find it?), REGISTER
> with that as in step (2), and update my dynamic DNS binding as in step (3).
>

Again exactly.

PS: And presumably, if my computer is itself non-firewalled, I can act as my
> own proxy; then dbarrett.p2psip.org would point directly to me, and I
> wouldn't depend on anyone.
>

Yup.  The "how do I find it" question is certainly one of the missing
pieces, but it's there in the DHT scenario anyway.

Thanks for elucidating those steps -- I should have done that first thing
this morning.

-Adam
-------------- next part --------------
An HTML attachment was scrubbed...
URL: https://lists.cs.columbia.edu/pipermail/p2p-sip/attachments/20061121/5c1b5c92/attachment.html

From slavitch at gmail.com  Tue Nov 21 16:01:08 2006
From: slavitch at gmail.com (Michael Slavitch)
Date: Tue, 21 Nov 2006 16:01:08 -0500
Subject: [p2p-sip] Using DNS (was Fwd: [p2p-hackers] designing with
	back-of-the-envelope estimates)
Message-ID: <1fc32c8f0611211301j473e01e1r93c10ad69b93870@mail.gmail.com>

I'm forwarding this to the P2SIP mailing list. It is very interestingreading especially when comparing DNS based systems to systems thatmaintain their own DHT, even large.
DNS in the quad-core age is not needed when 1G of db entries can bere-sorted in 30 seconds.
Regards
M

---------- Forwarded message ----------From: Ryan Barrett <p2p-hackers at ryanb.org>Date: Nov 21, 2006 3:34 PMSubject: [p2p-hackers] designing with back-of-the-envelope estimatesTo: p2p-hackers at lists.zooko.com

i caught a great talk the other day by jeff dean on using back-of-the-envelopecalculations to inform design decisions in large-scale systems. it's relevantto the latency vs. bandwidth thread, and also just good to know in general, soi figured i'd share a few of the more interesting slides.
no link for the talk, but jeff is http://labs.google.com/people/jeff/ .
(i'm sure we can dispute all of the numbers to varying degrees. the pointisn't to get them exactly right, just within an order of magnitude or so. :P)

Numbers Everyone Should Know
L1 cache reference                            0.5 nsBranch mispredict                             5 nsL2 cache reference                            7 nsMutex lock/unlock                           100 nsMain memory reference                       100 nsCompress 1K bytes with XXX               10,000 nsSend 2K bytes over 1 Gbps network        20,000 nsRead 1 MB sequentially from memory      250,000 nsRound trip within same datacenter       500,000 nsDisk seek                            10,000,000 nsRead 1 MB sequentially from network  10,000,000 nsRead 1 MB sequentially from disk     30,000,000 nsSend packet CA->Netherlands->CA     150,000,000 ns

Back of the Envelope Calculations
How long to generate image results page (30 thumbnails)?
Design 1: Read serially, thumbnail 256K images on the fly     30 seeks * 10 ms/seek + 30 * 256K / 30 MB/s = 560 ms
Design 2: Issue reads in parallel:     10 ms/seek + 256K read / 30 MB/s = 18 ms     (Ignores variance, so really more like 30-60 ms, probably)
Lots of variations:     ? caching (single images? whole sets of thumbnails?)     ? pre-computing thumbnails     ?...
Back of the envelope helps identify most promising...

How long to quicksort 1 GB of 4 byte numbers?
Comparisons: lots of unpredictable branches     log(2^28) passes over 2^28 numbers = ~2^33 comparisons     ~1/2 will mispredict, so 2^32 mispredicts * 5 ns/mispredict = 21 secs
Memory bandwidth: mostly sequential streaming     2^30 bytes * 28 passes = 28 GB. Memory BW is ~4 GB/s, so ~7 secs
So, it should take ~30 seconds to sort 1 GB on one CPU
Estimate only! Ignores initial read from disk, parallelism, etc...
-Ryan
--http://snarfed.org/
_______________________________________________p2p-hackers mailing listp2p-hackers at lists.zooko.comhttp://lists.zooko.com/mailman/listinfo/p2p-hackers

From dbarrett at quinthar.com  Tue Nov 21 16:01:01 2006
From: dbarrett at quinthar.com (David Barrett)
Date: Tue, 21 Nov 2006 13:01:01 -0800
Subject: [p2p-sip] Fwd:  what's wrong with DNS?
In-Reply-To: <24CCCC428EFEA2469BF046DB3C7A8D2214134B@namail5.corp.adobe.com>
Message-ID: <02b101c70db0$2c861b60$6b02a8c0@Quinthar>

> -----Original Message-----
> From: Henry Sinnreich
> Sent: Tuesday, November 21, 2006 12:39 PM
> To: Brian Rosen; Michael Slavitch; P2P-SIP
> Subject: Re: [p2p-sip] Fwd: what's wrong with DNS?
> 
> For what it's worth, I remember to have seen quite a number of p2p designs
> and none of them relies on the DNS. See for example http://opendht.org/
> and the P2P SIP system reported at Columbia University.

Well, it depends on how you define P2P.  By volume today the vast majority
of P2P traffic is BitTorrent, and it uses DNS for tracker lookups and
torrent distribution.  Or even historically, P2P was invented by Napster and
Scour, and they used central indexing servers located via DNS.

I agree that today's research in P2P has given up on DNS, but the actual
implementations that defined the term swore by it.

-david


From adamfisk at gmail.com  Tue Nov 21 16:18:48 2006
From: adamfisk at gmail.com (Adam Fisk)
Date: Tue, 21 Nov 2006 16:18:48 -0500
Subject: [p2p-sip] Fwd: what's wrong with DNS?
In-Reply-To: <1fc32c8f0611211251r3eeed24ao99d68aab11009a61@mail.gmail.com>
References: <035a01c70da8$d225d4f0$640fa8c0@cis.neustar.com>
	<029101c70dad$42a1c5f0$6b02a8c0@Quinthar>
	<1fc32c8f0611211246j7d999e24q48bb1c17e06ccf2a@mail.gmail.com>
	<1fc32c8f0611211250i7ae67db6pd45f59dc19eec81b@mail.gmail.com>
	<1fc32c8f0611211251r3eeed24ao99d68aab11009a61@mail.gmail.com>
Message-ID: <f40963db0611211318u5ccfc84aq21413f8b8e547075@mail.gmail.com>

Right.  Stepping back again for a second, I think we're just talking about
different sets of requirements.  The fascinating and important scenarios
like Katrina where you simply cannot rely on DNS need a solution, and I
think the various ad-hoc bootstrapping ideas may very well be the best
route.  I also ideologically agree with your other points about walled
gardens and user-centric, bottom up design.

For my own use case, though, basically providing a low-cost SIP
infrastructure where you use peers instead of running your own servers, the
dynamic DNS approach struck me as much simpler.  We could call this the
Skype use case.

-Adam


On 11/21/06, Michael Slavitch <slavitch at gmail.com> wrote:
>
> "I think you'd probably just use standard name servers.  You could use
> any public dynamic DNS provider out there.  Sure, they could go out,
> but that doesn't happen often. "
>
> My own requirements have to assume that the root nameserver no longer
> exists.
>
> What if a government turned it off?  Or a corporation? What then? What
> if there is a flood?
> Or an attack? Does the system go dark?
>
> Katrina exposed such loopholes in systems design.  Let's not design
> them in after the fact when there are known alternatives that offer
> rewards even when things are working well.
> _______________________________________________
> p2p-sip mailing list
> p2p-sip at cs.columbia.edu
> https://lists.cs.columbia.edu/cucslists/listinfo/p2p-sip
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: https://lists.cs.columbia.edu/pipermail/p2p-sip/attachments/20061121/a1ae8937/attachment.html

From br at brianrosen.net  Tue Nov 21 18:09:42 2006
From: br at brianrosen.net (Brian Rosen)
Date: Tue, 21 Nov 2006 18:09:42 -0500
Subject: [p2p-sip] Fwd: what's wrong with DNS?
In-Reply-To: <f40963db0611211318u5ccfc84aq21413f8b8e547075@mail.gmail.com>
Message-ID: <042301c70dc2$2400b0b0$640fa8c0@cis.neustar.com>

Please be careful.

 

If you are on a network that suddenly becomes disconnected from the
Internet, but your local network works, then your local DNS works, and any
question within your domain will be answered.  If you ask a question outside
your domain, it won't work, or you get a cached answer if it's still valid.

 

If your access network blows up, and you have no infrastructure, then
nothing really matters.

 

So when you say "scenarios like Katrina" the most common circumstances was
that the access network connection was lost.  Nothing you can do about that.
The second most common one was that you had an access network, but no
connections onward to the Internet.  In those cases, DNS works to anyone you
can talk to.

 

Brian

 

  _____  

From: p2p-sip-bounces at cs.columbia.edu
[mailto:p2p-sip-bounces at cs.columbia.edu] On Behalf Of Adam Fisk
Sent: Tuesday, November 21, 2006 4:19 PM
To: Michael Slavitch
Cc: P2P-SIP
Subject: Re: [p2p-sip] Fwd: what's wrong with DNS?

 

Right.  Stepping back again for a second, I think we're just talking about
different sets of requirements.  The fascinating and important scenarios
like Katrina where you simply cannot rely on DNS need a solution, and I
think the various ad-hoc bootstrapping ideas may very well be the best
route.  I also ideologically agree with your other points about walled
gardens and user-centric, bottom up design.  

For my own use case, though, basically providing a low-cost SIP
infrastructure where you use peers instead of running your own servers, the
dynamic DNS approach struck me as much simpler.  We could call this the
Skype use case.  

-Adam



On 11/21/06, Michael Slavitch <slavitch at gmail.com> wrote:

"I think you'd probably just use standard name servers.  You could use
any public dynamic DNS provider out there.  Sure, they could go out,
but that doesn't happen often. "

My own requirements have to assume that the root nameserver no longer
exists. 

What if a government turned it off?  Or a corporation? What then? What
if there is a flood?
Or an attack? Does the system go dark?

Katrina exposed such loopholes in systems design.  Let's not design
them in after the fact when there are known alternatives that offer
rewards even when things are working well.
_______________________________________________
p2p-sip mailing list
p2p-sip at cs.columbia.edu
https://lists.cs.columbia.edu/cucslists/listinfo/p2p-sip

 

-------------- next part --------------
An HTML attachment was scrubbed...
URL: https://lists.cs.columbia.edu/pipermail/p2p-sip/attachments/20061121/dd4fa644/attachment-0001.html

From slavitch at gmail.com  Tue Nov 21 18:50:54 2006
From: slavitch at gmail.com (Michael Slavitch)
Date: Tue, 21 Nov 2006 18:50:54 -0500
Subject: [p2p-sip] Fwd: what's wrong with DNS?
In-Reply-To: <042301c70dc2$2400b0b0$640fa8c0@cis.neustar.com>
References: <f40963db0611211318u5ccfc84aq21413f8b8e547075@mail.gmail.com>
	<042301c70dc2$2400b0b0$640fa8c0@cis.neustar.com>
Message-ID: <1fc32c8f0611211550leca6de4reea6b9fa23dc538c@mail.gmail.com>

There is this miracle called radio spectrum.

On 11/21/06, Brian Rosen <br at brianrosen.net> wrote:
>
>
>
>
> Please be careful.
>
>
>
> If you are on a network that suddenly becomes disconnected from the
> Internet, but your local network works, then your local DNS works, and any
> question within your domain will be answered.  If you ask a question outside
> your domain, it won't work, or you get a cached answer if it's still valid.
>
>
>
> If your access network blows up, and you have no infrastructure, then
> nothing really matters.
>
>
>
> So when you say "scenarios like Katrina" the most common circumstances was
> that the access network connection was lost.  Nothing you can do about that.
>  The second most common one was that you had an access network, but no
> connections onward to the Internet.  In those cases, DNS works to anyone you
> can talk to.
>
>
>
> Brian
>
>
>
>
>  ________________________________
>
>
> From: p2p-sip-bounces at cs.columbia.edu
> [mailto:p2p-sip-bounces at cs.columbia.edu] On Behalf Of Adam Fisk
>  Sent: Tuesday, November 21, 2006 4:19 PM
>  To: Michael Slavitch
>  Cc: P2P-SIP
>  Subject: Re: [p2p-sip] Fwd: what's wrong with DNS?
>
>
>
>
> Right.  Stepping back again for a second, I think we're just talking about
> different sets of requirements.  The fascinating and important scenarios
> like Katrina where you simply cannot rely on DNS need a solution, and I
> think the various ad-hoc bootstrapping ideas may very well be the best
> route.  I also ideologically agree with your other points about walled
> gardens and user-centric, bottom up design.
>
>  For my own use case, though, basically providing a low-cost SIP
> infrastructure where you use peers instead of running your own servers, the
> dynamic DNS approach struck me as much simpler.  We could call this the
> Skype use case.
>
>  -Adam
>
>
>
>
> On 11/21/06, Michael Slavitch <slavitch at gmail.com> wrote:
>
> "I think you'd probably just use standard name servers.  You could use
>  any public dynamic DNS provider out there.  Sure, they could go out,
>  but that doesn't happen often. "
>
>  My own requirements have to assume that the root nameserver no longer
> exists.
>
>  What if a government turned it off?  Or a corporation? What then? What
>  if there is a flood?
>  Or an attack? Does the system go dark?
>
>  Katrina exposed such loopholes in systems design.  Let's not design
>  them in after the fact when there are known alternatives that offer
>  rewards even when things are working well.
>  _______________________________________________
>  p2p-sip mailing list
>  p2p-sip at cs.columbia.edu
>  https://lists.cs.columbia.edu/cucslists/listinfo/p2p-sip
>
>

From dbarrett at quinthar.com  Tue Nov 21 19:24:41 2006
From: dbarrett at quinthar.com (David Barrett)
Date: Tue, 21 Nov 2006 16:24:41 -0800
Subject: [p2p-sip] Fwd:  what's wrong with DNS?
In-Reply-To: <036d01c70da9$d02bbd30$640fa8c0@cis.neustar.com>
Message-ID: <035401c70dcc$9ac6ae70$6b02a8c0@Quinthar>

> -----Original Message-----
> From: Brian Rosen
> Subject: Re: [p2p-sip] Fwd: what's wrong with DNS?
> 
> I'm working on systems that deal with the later three.  DNS seems to be
> something you CAN rely on within a network you have connectivity to.  You
> may lose connectivity to the global Internet, but within, say, an incident
> network, assuming that there is some DNS infrastructure seems pretty
> reasonable.
> 
> I have some sympathy with trying to make things work when, say the only
> entities in a network are a pair of first responders (a MANET network with
> exactly two members).  It should be possible for those two to talk to one
> another.  It IS possible to make such systems use DNS.  One or both of
> them
> have to enable DNS servers.  Working without such mechanisms seems to
> require multicast, also on our list of no-no's.

Actually, this is a good point.  There's a lot of talk about emergency ad
hoc networks constructed in disaster zones.  But even in these cases, won't
one of the very first tasks be setting up internet connectivity so they can
get messages out to the outside world?  And won't first on that list be
getting DNS set up?

I mean, VoIP is cool, but email and web are probably even more important.
There are things like satellite telephones, HAM radios, and even walkie
talkies -- hell, loudspeakers that can do wonders in a disaster.  But
there's simply no alternative to email and web access in an emergency, and
thus the priority (if I were FEMA) would be on getting those up first.  If
VoIP comes with it, neat, but I'd have plenty of other voice communication
systems set up well before waiting on net access.

So if we can generally agree that DNS is generally reliable for the 99% of
users in typical, non-disaster usage, can we also agree that DNS will be
available very quickly to most disaster networks as well -- whether the same
DNS as used by the internet, or a private DNS used only within the disaster
network itself (and then later hooked up to the rest of the world?.

If that's the case, then who *are* we targeting this to?  Is this really
being designed exclusively around the needs of the very first hours after
disaster strikes as an alternative to walkie talkies or yelling really loud?

-david


From spencer at mcsr-labs.org  Tue Nov 21 20:32:50 2006
From: spencer at mcsr-labs.org (Spencer Dawkins)
Date: Tue, 21 Nov 2006 19:32:50 -0600
Subject: [p2p-sip] Fwd:  what's wrong with DNS?
References: <035401c70dcc$9ac6ae70$6b02a8c0@Quinthar>
Message-ID: <090601c70dd6$1fec97f0$6a01a8c0@china.huawei.com>

And, on this note...

> If that's the case, then who *are* we targeting this to?

I can almost PROMISE that the authors of 
http://www.p2psip.org/drafts/draft-bryan-sipping-p2p-usecases-00.html would 
love to hear comments about their work to date, and I can GUARANTEE that the 
editor of http://www.p2psip.org/P2PUseCases.php is waiting with bated breath 
for any thoughts that occur to people who are thinking about "then who *are* 
we targeting this to?"

Thanks,

Spencer 



From dbarrett at quinthar.com  Tue Nov 21 21:33:11 2006
From: dbarrett at quinthar.com (David Barrett)
Date: Tue, 21 Nov 2006 18:33:11 -0800
Subject: [p2p-sip] Fwd:  what's wrong with DNS?
In-Reply-To: <090601c70dd6$1fec97f0$6a01a8c0@china.huawei.com>
Message-ID: <038401c70dde$92b477f0$6b02a8c0@Quinthar>

Excellent summary.  All but five of the use cases (in fact, all the "big"
use cases affecting 99% of users) support access to the global internet DNS.
Those that do not include: 

- Anonymous communication
- Ad-host and ephemeral groups
- Emergency first responder networks
- Extending the reach of mobile devices
- Deployments in the developing world

It seems that many or even all of these could support private DNS if we set
our minds to it and made it easy.  The problem of resolving names to IPs is
a big one -- it matters to much more than just VoIP.  Rather than creating a
SIP-only solution that lets people talk but do nothing else, why not just
tackle this at the DNS layer and thus make everything "just work"?

Anyway, the reason I liked Adam's solution so much is it seems like *such a
cleaner interface*.  Rather than trying to recreate a new fancy P2P naming
system *and* solve VoIP, why not just take each separately?  Dynamic DNS is
a really clever solution to the naming problem that seems workable on the
internet.  Furthermore, it doesn't seem too difficult (or, rather, any more
difficult) to invent a P2P-DNS system that works for VoIP as well as
everything else, and that handles the five scenarios listed above.

So here's my proposal:
- Use DynamicDNS for 99% of users
- Invent some P2P-DNS to handle the remaining 1%
- Make VoIP work like every other internet application (use DNS)
- Make all internet applications "just work" with P2P-DNS in the cases above

Sometimes circling the wagons differently can make a really hard problem
simple.  I like Alex's proposal because it might offer just that.

-david

> -----Original Message-----
> From: p2p-sip-bounces at cs.columbia.edu [mailto:p2p-sip-
> bounces at cs.columbia.edu] On Behalf Of Spencer Dawkins
> Sent: Tuesday, November 21, 2006 5:33 PM
> To: 'P2P-SIP'
> Subject: Re: [p2p-sip] Fwd: what's wrong with DNS?
> 
> And, on this note...
> 
> > If that's the case, then who *are* we targeting this to?
> 
> I can almost PROMISE that the authors of
> http://www.p2psip.org/drafts/draft-bryan-sipping-p2p-usecases-00.html
> would
> love to hear comments about their work to date, and I can GUARANTEE that
> the
> editor of http://www.p2psip.org/P2PUseCases.php is waiting with bated
> breath
> for any thoughts that occur to people who are thinking about "then who
> *are*
> we targeting this to?"
> 
> Thanks,
> 
> Spencer
> 
> 
> _______________________________________________
> p2p-sip mailing list
> p2p-sip at cs.columbia.edu
> https://lists.cs.columbia.edu/cucslists/listinfo/p2p-sip


From dbarrett at quinthar.com  Wed Nov 22 02:52:07 2006
From: dbarrett at quinthar.com (David Barrett)
Date: Tue, 21 Nov 2006 23:52:07 -0800
Subject: [p2p-sip] Fwd:  what's wrong with DNS?
In-Reply-To: <090601c70dd6$1fec97f0$6a01a8c0@china.huawei.com>
Message-ID: <001001c70e0b$26a27a80$0202a8c0@Quinthar>

Excellent summary.  All but five of the use cases (in fact, all the "big"
use cases affecting 99% of users) support access to the global internet DNS.
Those that do not include: 

- Anonymous communication
- Ad-host and ephemeral groups
- Emergency first responder networks
- Extending the reach of mobile devices
- Deployments in the developing world

It seems that many or even all of these could support private DNS if we set
our minds to it and made it easy.  The problem of resolving names to IPs is
a big one -- it matters to much more than just VoIP.  Rather than creating a
SIP-only solution that lets people talk but do nothing else, why not just
tackle this at the DNS layer and thus make everything "just work"?

Anyway, the reason I liked Adam's solution so much is it seems like *such a
cleaner interface*.  Rather than trying to recreate a new fancy P2P naming
system *and* solve VoIP, why not just take each separately?  Dynamic DNS is
a really clever solution to the naming problem that seems workable on the
internet.  Furthermore, it doesn't seem too difficult (or, rather, any more
difficult) to invent a P2P-DNS system that works for VoIP as well as
everything else, and that handles the five scenarios listed above.

So here's my proposal:
- Use DynamicDNS for 99% of users
- Invent some P2P-DNS to handle the remaining 1%
- Make VoIP work like every other internet application (use DNS)
- Make all internet applications "just work" with P2P-DNS in the cases above

Sometimes circling the wagons differently can make a really hard problem
simple.  I like Alex's proposal because it might offer just that.

-david

> -----Original Message-----
> From: p2p-sip-bounces at cs.columbia.edu [mailto:p2p-sip-
> bounces at cs.columbia.edu] On Behalf Of Spencer Dawkins
> Sent: Tuesday, November 21, 2006 5:33 PM
> To: 'P2P-SIP'
> Subject: Re: [p2p-sip] Fwd: what's wrong with DNS?
> 
> And, on this note...
> 
> > If that's the case, then who *are* we targeting this to?
> 
> I can almost PROMISE that the authors of
> http://www.p2psip.org/drafts/draft-bryan-sipping-p2p-usecases-00.html
> would
> love to hear comments about their work to date, and I can GUARANTEE that
> the
> editor of http://www.p2psip.org/P2PUseCases.php is waiting with bated
> breath
> for any thoughts that occur to people who are thinking about "then who
> *are*
> we targeting this to?"
> 
> Thanks,
> 
> Spencer
> 
> 
> _______________________________________________
> p2p-sip mailing list
> p2p-sip at cs.columbia.edu
> https://lists.cs.columbia.edu/cucslists/listinfo/p2p-sip


From dbildikar at ipunity.com  Wed Nov 22 03:25:44 2006
From: dbildikar at ipunity.com (Darshan Bildikar)
Date: Wed, 22 Nov 2006 13:55:44 +0530
Subject: [p2p-sip] Doubt in draft draft-bryan-sipping-p2p-03
In-Reply-To: <038401c70dde$92b477f0$6b02a8c0@Quinthar>
Message-ID: <EXCHANGEVM2mQPoHEpx00000be3@exchangevm.ipunity.com>

I've been going through the draft (draft-bryan-sipping-p2p) and there's one
thing I don't understand. I'm hoping someone can help me understand it
better. 

Section 5.2 states

"When locating a resource with a particular Resource-ID, the peer will send
the request to the finger table entry with the Peer-ID closest to the
desired Resource-ID"

It also states earlier that the resource ID is a hash of user name whereas
peer-id is a hash of IP and port. In this regard how will a peer determine
closeness of resource id to peer id when the hash values are actually based
on different parameters?

I suspect the explanation lies in this line ("The Resource-IDs map to the
same space as the Peer-IDs") that I don't really understand. Clarification
would help. What is it about the relation between the peer-id and the
resource-id that helps a peer identify where to send the request?

Additionally, it would help if this section were to clearly define what a
resource is and what a peer is. And the overlap between them. 

BR,
Darshan


From jean-francois.wauthy at alcatel.fr  Wed Nov 22 07:27:49 2006
From: jean-francois.wauthy at alcatel.fr (=?ISO-8859-1?Q?Jean-Fran=E7ois_Wauthy?=)
Date: Wed, 22 Nov 2006 13:27:49 +0100
Subject: [p2p-sip] Doubt in draft draft-bryan-sipping-p2p-03
In-Reply-To: <EXCHANGEVM2mQPoHEpx00000be3@exchangevm.ipunity.com>
References: <EXCHANGEVM2mQPoHEpx00000be3@exchangevm.ipunity.com>
Message-ID: <45644245.8030004@alcatel.fr>

Darshan Bildikar a ?crit :
> It also states earlier that the resource ID is a hash of user name whereas
> peer-id is a hash of IP and port. In this regard how will a peer determine
> closeness of resource id to peer id when the hash values are actually based
> on different parameters?
>
> I suspect the explanation lies in this line ("The Resource-IDs map to the
> same space as the Peer-IDs") that I don't really understand. Clarification
> would help. What is it about the relation between the peer-id and the
> resource-id that helps a peer identify where to send the request?
>
>   
The hash function (SHA1 here) is the same for the resource-IDs and 
peer-IDs, the results being a 160bit number in both cases you can 
compare them and therefore find the closest peer to send the request to.

> Additionally, it would help if this section were to clearly define what a
> resource is and what a peer is. And the overlap between them. 
>
>   
A resource is an information stored in the DHT and a peer can be seen as 
a host participating to the DHT overlay.

HTH,
Jean-Fran?ois Wauthy

From mlepinski at bbn.com  Wed Nov 22 09:47:23 2006
From: mlepinski at bbn.com (Matt Lepinski)
Date: Wed, 22 Nov 2006 09:47:23 -0500
Subject: [p2p-sip] Doubt in draft draft-bryan-sipping-p2p-03
In-Reply-To: <EXCHANGEVM2mQPoHEpx00000be3@exchangevm.ipunity.com>
References: <EXCHANGEVM2mQPoHEpx00000be3@exchangevm.ipunity.com>
Message-ID: <456462FB.2010507@bbn.com>

Darshan,

Perhaps this is best explained by way of an example.

For instance, let's say that I'm a peer and that I have four peers in my 
finger table. Each of these peers has a peer ID obtained by hashing the 
peer's IP address and port. For simplicity, in this example, I'm going 
to assume that the hash function (SHA-1) returns 24 bits (instead of 160 
bits) and so the output can be represented as 6 hexidecimal digits. So 
my finger table contains the following four entries.

Peer                   Peer-ID
------                   ---------
10.0.0.1:5678       12A9 
10.0.0.2:5678       2D8B
10.0.0.3:5678       409C
10.0.0.4:5678       8F25

Now let's say that I want to find the resource sip://bob at p2psip.example.net
I would obtain the Resource ID for this resource by applying the same 
hash (SHA-1) to the name of the resource (sip://bob at p2psip.example.net).
So let's say that I apply the hash and I get a Resource ID of 59FF. I 
would then compare the Resource ID to the Peer ID's in my finger table 
and determine that 409C is closest to the Resource ID that I'm looking 
for. Therefore, I would route request for information about 
sip://bob at p2psip.example.net to the peer at 10.0.0.3:5678.

Hope this helps,
- Matt Lepinski

Darshan Bildikar wrote:

>I've been going through the draft (draft-bryan-sipping-p2p) and there's one
>thing I don't understand. I'm hoping someone can help me understand it
>better. 
>
>Section 5.2 states
>
>"When locating a resource with a particular Resource-ID, the peer will send
>the request to the finger table entry with the Peer-ID closest to the
>desired Resource-ID"
>
>It also states earlier that the resource ID is a hash of user name whereas
>peer-id is a hash of IP and port. In this regard how will a peer determine
>closeness of resource id to peer id when the hash values are actually based
>on different parameters?
>
>I suspect the explanation lies in this line ("The Resource-IDs map to the
>same space as the Peer-IDs") that I don't really understand. Clarification
>would help. What is it about the relation between the peer-id and the
>resource-id that helps a peer identify where to send the request?
>
>Additionally, it would help if this section were to clearly define what a
>resource is and what a peer is. And the overlap between them. 
>
>BR,
>Darshan
>
>_______________________________________________
>p2p-sip mailing list
>p2p-sip at cs.columbia.edu
>https://lists.cs.columbia.edu/cucslists/listinfo/p2p-sip
>
>  
>



From hsinnrei at adobe.com  Wed Nov 22 09:53:05 2006
From: hsinnrei at adobe.com (Henry Sinnreich)
Date: Wed, 22 Nov 2006 06:53:05 -0800
Subject: [p2p-sip] Fwd:  what's wrong with DNS?
In-Reply-To: <038401c70dde$92b477f0$6b02a8c0@Quinthar>
Message-ID: <24CCCC428EFEA2469BF046DB3C7A8D22141350@namail5.corp.adobe.com>

>- Invent some P2P-DNS to handle the remaining 1%

I suppose we can also add Skype with its >100M users to the 1%  ;-)>

We do not seem to be able have to any better agreement than already in
the proposed charter:

"P2P SIP may use DNS but does not require it" or something like that.

So can we please move on and leave everyone to use DNS or NOT use DNS as
well?

I doubt taking this discussion any further on the list will change
anyone's mind. What was to be said has been said already. Maybe a
balanced group could make a summary offline, but that's beyond the scope
of this list.
Any good implementation will beat arguments anyway...

Thanks, Henry

-----Original Message-----
From: p2p-sip-bounces at cs.columbia.edu
[mailto:p2p-sip-bounces at cs.columbia.edu] On Behalf Of David Barrett
Sent: Tuesday, November 21, 2006 8:33 PM
To: 'Spencer Dawkins'; 'P2P-SIP'
Subject: Re: [p2p-sip] Fwd: what's wrong with DNS?

Excellent summary.  All but five of the use cases (in fact, all the
"big"
use cases affecting 99% of users) support access to the global internet
DNS.
Those that do not include: 

- Anonymous communication
- Ad-host and ephemeral groups
- Emergency first responder networks
- Extending the reach of mobile devices
- Deployments in the developing world

It seems that many or even all of these could support private DNS if we
set
our minds to it and made it easy.  The problem of resolving names to IPs
is
a big one -- it matters to much more than just VoIP.  Rather than
creating a
SIP-only solution that lets people talk but do nothing else, why not
just
tackle this at the DNS layer and thus make everything "just work"?

Anyway, the reason I liked Adam's solution so much is it seems like
*such a
cleaner interface*.  Rather than trying to recreate a new fancy P2P
naming
system *and* solve VoIP, why not just take each separately?  Dynamic DNS
is
a really clever solution to the naming problem that seems workable on
the
internet.  Furthermore, it doesn't seem too difficult (or, rather, any
more
difficult) to invent a P2P-DNS system that works for VoIP as well as
everything else, and that handles the five scenarios listed above.

So here's my proposal:
- Use DynamicDNS for 99% of users
- Invent some P2P-DNS to handle the remaining 1%
- Make VoIP work like every other internet application (use DNS)
- Make all internet applications "just work" with P2P-DNS in the cases
above

Sometimes circling the wagons differently can make a really hard problem
simple.  I like Alex's proposal because it might offer just that.

-david

> -----Original Message-----
> From: p2p-sip-bounces at cs.columbia.edu [mailto:p2p-sip-
> bounces at cs.columbia.edu] On Behalf Of Spencer Dawkins
> Sent: Tuesday, November 21, 2006 5:33 PM
> To: 'P2P-SIP'
> Subject: Re: [p2p-sip] Fwd: what's wrong with DNS?
> 
> And, on this note...
> 
> > If that's the case, then who *are* we targeting this to?
> 
> I can almost PROMISE that the authors of
> http://www.p2psip.org/drafts/draft-bryan-sipping-p2p-usecases-00.html
> would
> love to hear comments about their work to date, and I can GUARANTEE
that
> the
> editor of http://www.p2psip.org/P2PUseCases.php is waiting with bated
> breath
> for any thoughts that occur to people who are thinking about "then who
> *are*
> we targeting this to?"
> 
> Thanks,
> 
> Spencer
> 
> 
> _______________________________________________
> p2p-sip mailing list
> p2p-sip at cs.columbia.edu
> https://lists.cs.columbia.edu/cucslists/listinfo/p2p-sip

_______________________________________________
p2p-sip mailing list
p2p-sip at cs.columbia.edu
https://lists.cs.columbia.edu/cucslists/listinfo/p2p-sip


From slavitch at gmail.com  Wed Nov 22 10:25:27 2006
From: slavitch at gmail.com (Michael Slavitch)
Date: Wed, 22 Nov 2006 10:25:27 -0500
Subject: [p2p-sip] Fwd: what's wrong with DNS?
In-Reply-To: <24CCCC428EFEA2469BF046DB3C7A8D22141350@namail5.corp.adobe.com>
References: <038401c70dde$92b477f0$6b02a8c0@Quinthar>
	<24CCCC428EFEA2469BF046DB3C7A8D22141350@namail5.corp.adobe.com>
Message-ID: <1fc32c8f0611220725q1211de39sbd27367840c6baea@mail.gmail.com>

DDNS is fine as one of many possible methods, each method fully
described and fully enumerated in the standard  protocol, for protocol
discovery and for identification. It might even be the first one in
the list, and the one in use in most early-adopter implementations.

It is not fine as the one and only especially since P2P can exploits
alternatives as well.
The same is not true for C/S SIP so that is not an issue.

Regards

M

On 11/22/06, Henry Sinnreich <hsinnrei at adobe.com> wrote:
> >- Invent some P2P-DNS to handle the remaining 1%
>
> I suppose we can also add Skype with its >100M users to the 1%  ;-)>
>
> We do not seem to be able have to any better agreement than already in
> the proposed charter:
>
> "P2P SIP may use DNS but does not require it" or something like that.
>
> So can we please move on and leave everyone to use DNS or NOT use DNS as
> well?
>
> I doubt taking this discussion any further on the list will change
> anyone's mind. What was to be said has been said already. Maybe a
> balanced group could make a summary offline, but that's beyond the scope
> of this list.
> Any good implementation will beat arguments anyway...
>
> Thanks, Henry
>
> -----Original Message-----
> From: p2p-sip-bounces at cs.columbia.edu
> [mailto:p2p-sip-bounces at cs.columbia.edu] On Behalf Of David Barrett
> Sent: Tuesday, November 21, 2006 8:33 PM
> To: 'Spencer Dawkins'; 'P2P-SIP'
> Subject: Re: [p2p-sip] Fwd: what's wrong with DNS?
>
> Excellent summary.  All but five of the use cases (in fact, all the
> "big"
> use cases affecting 99% of users) support access to the global internet
> DNS.
> Those that do not include:
>
> - Anonymous communication
> - Ad-host and ephemeral groups
> - Emergency first responder networks
> - Extending the reach of mobile devices
> - Deployments in the developing world
>
> It seems that many or even all of these could support private DNS if we
> set
> our minds to it and made it easy.  The problem of resolving names to IPs
> is
> a big one -- it matters to much more than just VoIP.  Rather than
> creating a
> SIP-only solution that lets people talk but do nothing else, why not
> just
> tackle this at the DNS layer and thus make everything "just work"?
>
> Anyway, the reason I liked Adam's solution so much is it seems like
> *such a
> cleaner interface*.  Rather than trying to recreate a new fancy P2P
> naming
> system *and* solve VoIP, why not just take each separately?  Dynamic DNS
> is
> a really clever solution to the naming problem that seems workable on
> the
> internet.  Furthermore, it doesn't seem too difficult (or, rather, any
> more
> difficult) to invent a P2P-DNS system that works for VoIP as well as
> everything else, and that handles the five scenarios listed above.
>
> So here's my proposal:
> - Use DynamicDNS for 99% of users
> - Invent some P2P-DNS to handle the remaining 1%
> - Make VoIP work like every other internet application (use DNS)
> - Make all internet applications "just work" with P2P-DNS in the cases
> above
>
> Sometimes circling the wagons differently can make a really hard problem
> simple.  I like Alex's proposal because it might offer just that.
>
> -david
>
> > -----Original Message-----
> > From: p2p-sip-bounces at cs.columbia.edu [mailto:p2p-sip-
> > bounces at cs.columbia.edu] On Behalf Of Spencer Dawkins
> > Sent: Tuesday, November 21, 2006 5:33 PM
> > To: 'P2P-SIP'
> > Subject: Re: [p2p-sip] Fwd: what's wrong with DNS?
> >
> > And, on this note...
> >
> > > If that's the case, then who *are* we targeting this to?
> >
> > I can almost PROMISE that the authors of
> > http://www.p2psip.org/drafts/draft-bryan-sipping-p2p-usecases-00.html
> > would
> > love to hear comments about their work to date, and I can GUARANTEE
> that
> > the
> > editor of http://www.p2psip.org/P2PUseCases.php is waiting with bated
> > breath
> > for any thoughts that occur to people who are thinking about "then who
> > *are*
> > we targeting this to?"
> >
> > Thanks,
> >
> > Spencer
> >
> >
> > _______________________________________________
> > p2p-sip mailing list
> > p2p-sip at cs.columbia.edu
> > https://lists.cs.columbia.edu/cucslists/listinfo/p2p-sip
>
> _______________________________________________
> p2p-sip mailing list
> p2p-sip at cs.columbia.edu
> https://lists.cs.columbia.edu/cucslists/listinfo/p2p-sip
>
> _______________________________________________
> p2p-sip mailing list
> p2p-sip at cs.columbia.edu
> https://lists.cs.columbia.edu/cucslists/listinfo/p2p-sip
>

From adamfisk at gmail.com  Wed Nov 22 10:49:33 2006
From: adamfisk at gmail.com (Adam Fisk)
Date: Wed, 22 Nov 2006 10:49:33 -0500
Subject: [p2p-sip] Fwd: what's wrong with DNS?
In-Reply-To: <24CCCC428EFEA2469BF046DB3C7A8D22141350@namail5.corp.adobe.com>
References: <038401c70dde$92b477f0$6b02a8c0@Quinthar>
	<24CCCC428EFEA2469BF046DB3C7A8D22141350@namail5.corp.adobe.com>
Message-ID: <f40963db0611220749n73bf0473m9b86346122a02a0d@mail.gmail.com>

> >- Invent some P2P-DNS to handle the remaining 1%
>
> I suppose we can also add Skype with its >100M users to the 1%  ;-)>


I'd argue Skype is the perfect use case for the DDNS approach, actually.


> So can we please move on and leave everyone to use DNS or NOT use DNS as
> well?


Sounds good to me.  These conversations are exhausting!


> Any good implementation will beat arguments anyway...



Amen to that.

-Adam
-------------- next part --------------
An HTML attachment was scrubbed...
URL: https://lists.cs.columbia.edu/pipermail/p2p-sip/attachments/20061122/ccbf1c53/attachment.html

From dbarrett at quinthar.com  Wed Nov 22 18:57:08 2006
From: dbarrett at quinthar.com (David Barrett)
Date: Wed, 22 Nov 2006 15:57:08 -0800
Subject: [p2p-sip] Fwd: what's wrong with DNS?
In-Reply-To: <1fc32c8f0611220725q1211de39sbd27367840c6baea@mail.gmail.com>
Message-ID: <014201c70e91$eb8b2500$0202a8c0@Quinthar>

> On 11/22/06, Henry Sinnreich <hsinnrei at adobe.com> wrote:
> > >- Invent some P2P-DNS to handle the remaining 1%
> >
> > I suppose we can also add Skype with its >100M users to the 1%  ;-)>

Skype depends wholly and entirely on DNS for the first and most critical
function: authentication.

So, let's put those 100M users solidly back in the "utterly dependent upon
DNS" category.  Indeed, even if Skype didn't so completely rely on globally
hosted central servers, that's not my point: the point I'm trying to make is
all Skype users *do* have reliable access to DNS, and so will a
hypermajority of P2P-SIP users.

Again, the number of users who will not have access to DNS is infinitesimal.
Even an extreme "disaster" network will get DNS quickly, and an "ad hoc"
network of nodes 20 feet away from each other is better served by talking
REALLY LOUD.  Anybody sufficiently prepared to establish a long-range
fixed-wireless point-to-point network could easily set up a private DNS
server (and indeed, would probably have one already).

Thus rather than inventing a new DNS and making it specific only to VoIP,
why not just use the existing DNS, and then figure out how to extend DNS
into the incredibly rare networks where it isn't already available?

-david


From matthew at matthew.at  Wed Nov 22 19:56:44 2006
From: matthew at matthew.at (Matthew Kaufman)
Date: Wed, 22 Nov 2006 16:56:44 -0800
Subject: [p2p-sip] Fwd: what's wrong with DNS?
In-Reply-To: <014201c70e91$eb8b2500$0202a8c0@Quinthar>
Message-ID: <014c01c70e9a$3f2bb550$029b0a0a@matthewdeskxps>

David Barrett:
> Again, the number of users who will not have access to DNS is
infinitesimal.
> Even an extreme "disaster" network will get DNS quickly, and an "ad hoc"
> network of nodes 20 feet away from each other is better served by talking
> REALLY LOUD.  Anybody sufficiently prepared to establish a long-range
> fixed-wireless point-to-point network could easily set up a private DNS
> server (and indeed, would probably have one already).

And anywhere that you can set up an ad-hoc network you can use
zeroconf/mDNS. I'd love to see a non-null enumeration of the P2P-SIP use
cases where you don't have:
  1) access to the global DNS or,
  2) everyone on the same LAN such that mDNS is sufficient or,
  3) access to a network-local DNS set up by the local infrastructure
provider that you can see via DHCP or manual configuration

As David points out, #3 followed soon thereafter by #1 is likely in any real
disaster scenario, and #2 applies to most if not all ad-hoc wireless (or
wired) cases.

Matthew Kaufman
matthew at matthew.at




From huang-ming.pan at comcast.net  Wed Nov 22 21:16:35 2006
From: huang-ming.pan at comcast.net (Peter Pan)
Date: Wed, 22 Nov 2006 18:16:35 -0800
Subject: [p2p-sip] Fwd: what's wrong with DNS?
References: <014201c70e91$eb8b2500$0202a8c0@Quinthar>
Message-ID: <000901c70ea5$67209610$030aa8c0@comcast.net>

Cannot live with web browsing without DNS.

However, to decentralize p2p-sip with best efforts, is it possible that DNS
is used ONLY IN CASE for "evacuation" purpose?

BTW, it seems to me that Skype uses NO DNS on (super)node lookup.

Also, Skype could do authentication without DNS, IMHO.

Peter
----- Original Message -----
From: "David Barrett" <dbarrett at quinthar.com>
To: "'Michael Slavitch'" <slavitch at gmail.com>; "'Henry Sinnreich'"
<hsinnrei at adobe.com>; "'P2P-SIP'" <p2p-sip at cs.columbia.edu>
Sent: Wednesday, November 22, 2006 3:57 PM
Subject: Re: [p2p-sip] Fwd: what's wrong with DNS?


> > On 11/22/06, Henry Sinnreich <hsinnrei at adobe.com> wrote:
> > > >- Invent some P2P-DNS to handle the remaining 1%
> > >
> > > I suppose we can also add Skype with its >100M users to the 1%  ;-)>
>
> Skype depends wholly and entirely on DNS for the first and most critical
> function: authentication.
>
> So, let's put those 100M users solidly back in the "utterly dependent upon
> DNS" category.  Indeed, even if Skype didn't so completely rely on
globally
> hosted central servers, that's not my point: the point I'm trying to make
is
> all Skype users *do* have reliable access to DNS, and so will a
> hypermajority of P2P-SIP users.
>
> Again, the number of users who will not have access to DNS is
infinitesimal.
> Even an extreme "disaster" network will get DNS quickly, and an "ad hoc"
> network of nodes 20 feet away from each other is better served by talking
> REALLY LOUD.  Anybody sufficiently prepared to establish a long-range
> fixed-wireless point-to-point network could easily set up a private DNS
> server (and indeed, would probably have one already).
>
> Thus rather than inventing a new DNS and making it specific only to VoIP,
> why not just use the existing DNS, and then figure out how to extend DNS
> into the incredibly rare networks where it isn't already available?
>
> -david
>
> _______________________________________________
> p2p-sip mailing list
> p2p-sip at cs.columbia.edu
> https://lists.cs.columbia.edu/cucslists/listinfo/p2p-sip


From adamfisk at gmail.com  Wed Nov 22 21:59:42 2006
From: adamfisk at gmail.com (Adam Fisk)
Date: Wed, 22 Nov 2006 21:59:42 -0500
Subject: [p2p-sip] Fwd: what's wrong with DNS?
In-Reply-To: <000901c70ea5$67209610$030aa8c0@comcast.net>
References: <014201c70e91$eb8b2500$0202a8c0@Quinthar>
	<000901c70ea5$67209610$030aa8c0@comcast.net>
Message-ID: <f40963db0611221859h2d51321cw7f7f450d73ba88f3@mail.gmail.com>

>
> BTW, it seems to me that Skype uses NO DNS on (super)node lookup.



Right, but they could do it more easily *with DNS*.  Just a much simpler
architecture that takes out the bulk of the work.  It would also lower
latency in the average case.

-Adam
-------------- next part --------------
An HTML attachment was scrubbed...
URL: https://lists.cs.columbia.edu/pipermail/p2p-sip/attachments/20061122/07fbd08b/attachment-0001.html

From saikat at cs.cornell.edu  Wed Nov 22 23:01:45 2006
From: saikat at cs.cornell.edu (Saikat Guha)
Date: Wed, 22 Nov 2006 23:01:45 -0500
Subject: [p2p-sip] Fwd: what's wrong with DNS?
In-Reply-To: <000901c70ea5$67209610$030aa8c0@comcast.net>
References: <014201c70e91$eb8b2500$0202a8c0@Quinthar>
	<000901c70ea5$67209610$030aa8c0@comcast.net>
Message-ID: <1164254505.12268.7.camel@sioux.systems.cs.cornell.edu>

On Wed, 2006-11-22 at 18:16 -0800, Peter Pan wrote:
> BTW, it seems to me that Skype uses NO DNS on (super)node lookup.
> 
> Also, Skype could do authentication without DNS, IMHO.

Skype actually resorts to hard-coded IP addresses in the binary for
bootstrap; after the first successful bootstrap, it downloads additional
IPs. The point is moot IMHO, however, since the Skype model is different
from p2p-sip -- the Skype model is adversarial where the network
administrator is assumed to be actively trying to squelch Skype. As a
result Skype attempts to be as hard to detect and block as possible
including its use of random ports, encrypted packets, and a complete
lack of reliance on services provided/controlled by the network admin. 


On the other hand, the P2P primitive woven into Windows Vista is based
on the DNS interface -- each endpoint has a hostname, others can resolve
the IP for an endpoint by issuing a DNS query. That the backend
implementation of this interface (PNRP) is peer to peer is a separate
matter. The interface to name resolution chosen by Vista still is DNS.

-- 
Saikat
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 189 bytes
Desc: This is a digitally signed message part
Url : https://lists.cs.columbia.edu/pipermail/p2p-sip/attachments/20061122/4a6a7460/attachment.bin

From huang-ming.pan at comcast.net  Thu Nov 23 00:43:05 2006
From: huang-ming.pan at comcast.net (Peter Pan)
Date: Wed, 22 Nov 2006 21:43:05 -0800
Subject: [p2p-sip] Fwd: what's wrong with DNS?
References: <014201c70e91$eb8b2500$0202a8c0@Quinthar>
	<000901c70ea5$67209610$030aa8c0@comcast.net>
	<f40963db0611221859h2d51321cw7f7f450d73ba88f3@mail.gmail.com>
Message-ID: <000801c70ec2$402e9f30$030aa8c0@comcast.net>

"lower latency"? it seems opposite...

Peter
----- Original Message -----
From: Adam Fisk
To: Peter Pan
Cc: David Barrett ; Michael Slavitch ; Henry Sinnreich ; P2P-SIP
Sent: Wednesday, November 22, 2006 6:59 PM
Subject: Re: [p2p-sip] Fwd: what's wrong with DNS?





BTW, it seems to me that Skype uses NO DNS on (super)node lookup.


Right, but they could do it more easily *with DNS*.  Just a much simpler
architecture that takes out the bulk of the work.  It would also lower
latency in the average case.

-Adam


From dbarrett at quinthar.com  Thu Nov 23 00:58:37 2006
From: dbarrett at quinthar.com (David Barrett)
Date: Wed, 22 Nov 2006 21:58:37 -0800
Subject: [p2p-sip] Fwd: what's wrong with DNS?
In-Reply-To: <000801c70ec2$402e9f30$030aa8c0@comcast.net>
Message-ID: <01d601c70ec4$6f4ee9d0$0202a8c0@Quinthar>

Yes, *much* lower latency.  A simple lookup on a central server much faster
than a distributed lookup where you connect to not just one, but multiple
client computers hopping along a DHT.

-david

> -----Original Message-----
> From: Peter Pan [mailto:huang-ming.pan at comcast.net]
> Sent: Wednesday, November 22, 2006 9:43 PM
> To: Adam Fisk
> Cc: David Barrett; Michael Slavitch; Henry Sinnreich; P2P-SIP
> Subject: Re: [p2p-sip] Fwd: what's wrong with DNS?
> 
> "lower latency"? it seems opposite...
> 
> Peter
> ----- Original Message -----
> From: Adam Fisk
> To: Peter Pan
> Cc: David Barrett ; Michael Slavitch ; Henry Sinnreich ; P2P-SIP
> Sent: Wednesday, November 22, 2006 6:59 PM
> Subject: Re: [p2p-sip] Fwd: what's wrong with DNS?
> 
> 
> 
> 
> 
> BTW, it seems to me that Skype uses NO DNS on (super)node lookup.
> 
> 
> Right, but they could do it more easily *with DNS*.  Just a much simpler
> architecture that takes out the bulk of the work.  It would also lower
> latency in the average case.
> 
> -Adam


From huang-ming.pan at comcast.net  Thu Nov 23 01:10:06 2006
From: huang-ming.pan at comcast.net (Peter Pan)
Date: Wed, 22 Nov 2006 22:10:06 -0800
Subject: [p2p-sip] Fwd: what's wrong with DNS?
References: <01d601c70ec4$6f4ee9d0$0202a8c0@Quinthar>
Message-ID: <002101c70ec6$061f8170$030aa8c0@comcast.net>

Problem here is, if you want the central server why p2p for sip?

or, we can be dismissed now :)

Peter
----- Original Message -----
From: "David Barrett" <dbarrett at quinthar.com>
To: "'Peter Pan'" <huang-ming.pan at comcast.net>; "'Adam Fisk'"
<adamfisk at gmail.com>
Cc: "'Michael Slavitch'" <slavitch at gmail.com>; "'Henry Sinnreich'"
<hsinnrei at adobe.com>; "'P2P-SIP'" <p2p-sip at cs.columbia.edu>
Sent: Wednesday, November 22, 2006 9:58 PM
Subject: RE: [p2p-sip] Fwd: what's wrong with DNS?


> Yes, *much* lower latency.  A simple lookup on a central server much
faster
> than a distributed lookup where you connect to not just one, but multiple
> client computers hopping along a DHT.
>
> -david
>
> > -----Original Message-----
> > From: Peter Pan [mailto:huang-ming.pan at comcast.net]
> > Sent: Wednesday, November 22, 2006 9:43 PM
> > To: Adam Fisk
> > Cc: David Barrett; Michael Slavitch; Henry Sinnreich; P2P-SIP
> > Subject: Re: [p2p-sip] Fwd: what's wrong with DNS?
> >
> > "lower latency"? it seems opposite...
> >
> > Peter
> > ----- Original Message -----
> > From: Adam Fisk
> > To: Peter Pan
> > Cc: David Barrett ; Michael Slavitch ; Henry Sinnreich ; P2P-SIP
> > Sent: Wednesday, November 22, 2006 6:59 PM
> > Subject: Re: [p2p-sip] Fwd: what's wrong with DNS?
> >
> >
> >
> >
> >
> > BTW, it seems to me that Skype uses NO DNS on (super)node lookup.
> >
> >
> > Right, but they could do it more easily *with DNS*.  Just a much simpler
> > architecture that takes out the bulk of the work.  It would also lower
> > latency in the average case.
> >
> > -Adam
>


From adamfisk at gmail.com  Thu Nov 23 01:17:12 2006
From: adamfisk at gmail.com (Adam Fisk)
Date: Thu, 23 Nov 2006 01:17:12 -0500
Subject: [p2p-sip] Fwd: what's wrong with DNS?
In-Reply-To: <002101c70ec6$061f8170$030aa8c0@comcast.net>
References: <01d601c70ec4$6f4ee9d0$0202a8c0@Quinthar>
	<002101c70ec6$061f8170$030aa8c0@comcast.net>
Message-ID: <f40963db0611222217s2bf3ce4fp943f792e123661a5@mail.gmail.com>

On 11/23/06, Peter Pan <huang-ming.pan at comcast.net> wrote:
>
> Problem here is, if you want the central server why p2p for sip?


Right.  We're actually trying to solve real problems in the simplest way
rather than force a particular solution.

-Adam
-------------- next part --------------
An HTML attachment was scrubbed...
URL: https://lists.cs.columbia.edu/pipermail/p2p-sip/attachments/20061123/6bf14e11/attachment.html

From dbarrett at quinthar.com  Thu Nov 23 01:30:08 2006
From: dbarrett at quinthar.com (David Barrett)
Date: Wed, 22 Nov 2006 22:30:08 -0800
Subject: [p2p-sip] Fwd: what's wrong with DNS?
In-Reply-To: <002101c70ec6$061f8170$030aa8c0@comcast.net>
Message-ID: <01e701c70ec8$d39edb80$0202a8c0@Quinthar>

> -----Original Message-----
> From: Peter Pan [mailto:huang-ming.pan at comcast.net]
> Subject: Re: [p2p-sip] Fwd: what's wrong with DNS?
> 
> Problem here is, if you want the central server why p2p for sip?
> 
> or, we can be dismissed now :)

There are still plenty of P2P problems to be solved.  NAT traversal alone is
a huge one.  But we needn't make things harder than they already are by
embarking on a grand DHT experiment when there are easier solutions to the
name-resolution problem ready and waiting.

-davd



From philip_matthews at magma.ca  Thu Nov 23 15:42:44 2006
From: philip_matthews at magma.ca (Philip Matthews)
Date: Thu, 23 Nov 2006 15:42:44 -0500
Subject: [p2p-sip] Fwd:  what's wrong with DNS?
In-Reply-To: <035401c70dcc$9ac6ae70$6b02a8c0@Quinthar>
References: <035401c70dcc$9ac6ae70$6b02a8c0@Quinthar>
Message-ID: <47959272-70F4-40D0-A599-ED3EF02D9ED5@magma.ca>


On 21-Nov-06, at 19:24 , David Barrett wrote:

>
> If that's the case, then who *are* we targeting this to?  Is this  
> really
> being designed exclusively around the needs of the very first hours  
> after
> disaster strikes as an alternative to walkie talkies or yelling  
> really loud?


Our system has the requirement that it must work without Internet  
access,
and we have made a few sales where this is the case. We sell the  
phones in a bundle
with a PoE switch and a PSTN gateway, and the customer is happy.

Despite what people on this mailing list probably think, not EVERYONE
thinks internet access is a must-have. My next-door neighbors don't
even have a computer ...

Adam's idea sounds to me like a way to interface a P2PSIP system
with a CS SIP system. But I don't see it as the P2PSIP system itself.

- Philip


From matthew at matthew.at  Thu Nov 23 18:10:09 2006
From: matthew at matthew.at (Matthew Kaufman)
Date: Thu, 23 Nov 2006 15:10:09 -0800
Subject: [p2p-sip] Fwd:  what's wrong with DNS?
In-Reply-To: <47959272-70F4-40D0-A599-ED3EF02D9ED5@magma.ca>
Message-ID: <018301c70f54$857e87c0$029b0a0a@matthewdeskxps>

Philip Matthews:
>
> Our system has the requirement that it must work without Internet  
> access, and we have made a few sales where this is the case. We sell the  
> phones in a bundle with a PoE switch and a PSTN gateway, and the 
> customer is happy.

Sounds like mDNS on the devices, or even easier, a smart DNS server on the
PSTN gateway, and you'd be set, if DNS were the node-finding mechanism.

I'm still having trouble thinking of a use case where neither the global
DNS, nor a network-local DNS server, nor mDNS will work for that purpose.

At the same time, I find it hard to believe that the proposed Chord-like DHT
that's been tossed around as a "decentralized" solution to the node-lookup
problem on this list will actually provide useful answers with the sort of
churn one sees on a typical multi-thousand-user P2P overlay.

Personally, I see some real "P2P" things that SIP could use solved in a
standardized way, and while those are interesting problems, and the
solutions would help real users, a DHT isn't the solution, since the actual
problems aren't in the "node lookup" domain at all.

I agree that starting with something like DNS as a base for node lookup
means the actually useful work in the form of making sure that
P2P-SIP-enabled phones could establish direct registrar-less P2P connections
-- with NAT in the way (as it usually is) -- could get taken care of first.
Then, if anyone finds a real use case where DNS or mDNS is insufficient,
maybe we can all go back to experimenting with new signaling message formats
for several-year-old DHT technology.

Matthew Kaufman
matthew at matthew.at


From karstbj at yahoo.com  Fri Nov 24 03:13:33 2006
From: karstbj at yahoo.com (Karst Bjorgson)
Date: Fri, 24 Nov 2006 00:13:33 -0800 (PST)
Subject: [p2p-sip] High security use cases (was: Re: Fwd: what's wrong with
	DNS?)
In-Reply-To: <090601c70dd6$1fec97f0$6a01a8c0@china.huawei.com>
Message-ID: <461323.61755.qm@web53014.mail.yahoo.com>


> I can almost PROMISE that the authors of 
>
http://www.p2psip.org/drafts/draft-bryan-sipping-p2p-usecases-00.html
> would 
> love to hear comments about their work to date, and
> I can GUARANTEE that the 
> editor of http://www.p2psip.org/P2PUseCases.php is
> waiting with bated breath 
> for any thoughts that occur to people who are
> thinking about "then who *are* 
> we targeting this to?"

Section 3.2 of the use cases drafts reviews high
security environments. AFAIK, many of these
environments have an extra "logging" requirement. For
example, in financial institutions, SEC requires
logging of all communications to enforce compliance
with trading rules. I am not sure that logging is easy
to provie in a P2P system (P2P logging?), but I
believe this requirement should at least be
acknowledged in the draft. 

-- Karst


 
____________________________________________________________________________________
Do you Yahoo!?
Everyone is raving about the all-new Yahoo! Mail beta.
http://new.mail.yahoo.com

From karstbj at yahoo.com  Fri Nov 24 03:17:56 2006
From: karstbj at yahoo.com (Karst Bjorgson)
Date: Fri, 24 Nov 2006 00:17:56 -0800 (PST)
Subject: [p2p-sip] Fwd: what's wrong with DNS?
In-Reply-To: <01e701c70ec8$d39edb80$0202a8c0@Quinthar>
Message-ID: <398251.93684.qm@web53002.mail.yahoo.com>


--- David Barrett <dbarrett at quinthar.com> wrote:

> There are still plenty of P2P problems to be solved.
>  NAT traversal alone is
> a huge one.  But we needn't make things harder than
> they already are by
> embarking on a grand DHT experiment when there are
> easier solutions to the
> name-resolution problem ready and waiting.

Right. Besides, if someone wants to experiment with
DHT, they can use the DHT to provide an alternate DNS
service, resolving names in an ad hoc domain,
something like my-ua-name.dht-example.net.

The big advantage of this approach is risk management.
The P2PSIP can go on solving SIP level problems while
a separate group can experiment with P2P DNS. The two
groups can both succeed or fail without impacting each
other.

-- Karst


 
____________________________________________________________________________________
Cheap talk?
Check out Yahoo! Messenger's low PC-to-Phone call rates.
http://voice.yahoo.com

From Omar.El-yacoubi at insa-lyon.fr  Fri Nov 24 08:28:26 2006
From: Omar.El-yacoubi at insa-lyon.fr (el Yacoubi Omar)
Date: Fri, 24 Nov 2006 14:28:26 +0100
Subject: [p2p-sip] API JAVA of P2PSIP
Message-ID: <1164374906.4566f37ab756e@webmail.insa-lyon.fr>



Hello,
I need please some information about how the packages in the JAVA API of SIP
are
going to be modified?
I understand the fact that P2P is encoded in SIP but I don?t exactly see the
consequences on the SIP API and its packages? architecture.

Thanks for your answer,

Omar.


From br at brianrosen.net  Fri Nov 24 22:18:58 2006
From: br at brianrosen.net (Brian Rosen)
Date: Fri, 24 Nov 2006 22:18:58 -0500
Subject: [p2p-sip] API JAVA of P2PSIP
In-Reply-To: <1164374906.4566f37ab756e@webmail.insa-lyon.fr>
Message-ID: <008a01c71040$75e742b0$640fa8c0@cis.neustar.com>

> I need please some information about how the packages in the JAVA API of
> SIP
> are
> going to be modified?
> I understand the fact that P2P is encoded in SIP but I don't exactly see
> the
> consequences on the SIP API and its packages' architecture.


We don't know yet

Brian


From hsinnrei at adobe.com  Sat Nov 25 09:12:38 2006
From: hsinnrei at adobe.com (Henry Sinnreich)
Date: Sat, 25 Nov 2006 06:12:38 -0800
Subject: [p2p-sip] Fwd:  what's wrong with DNS?
In-Reply-To: <47959272-70F4-40D0-A599-ED3EF02D9ED5@magma.ca>
Message-ID: <24CCCC428EFEA2469BF046DB3C7A8D22141359@namail5.corp.adobe.com>

I would like to add the important application of home network for
consumer electronics that must also work even when disconnected form the
Internet.

Eunsoo Shim could probably expand.

Philip Mathews is right with the observation "Despite what people on
this mailing list probably think": 

Dependency on DNS is in last instance a business decision and all the
major P2P systems out there DO NOT rely on the DNS and will not be
influenced by these interminable discussion anyway. 

Live with it: P2P need not rely on DNS.

Thanks, Henry

-----Original Message-----
From: p2p-sip-bounces at cs.columbia.edu
[mailto:p2p-sip-bounces at cs.columbia.edu] On Behalf Of Philip Matthews
Sent: Thursday, November 23, 2006 2:43 PM
To: David Barrett; Adam Fisk
Cc: P2P-SIP
Subject: Re: [p2p-sip] Fwd: what's wrong with DNS?


On 21-Nov-06, at 19:24 , David Barrett wrote:

>
> If that's the case, then who *are* we targeting this to?  Is this  
> really
> being designed exclusively around the needs of the very first hours  
> after
> disaster strikes as an alternative to walkie talkies or yelling  
> really loud?


Our system has the requirement that it must work without Internet  
access,
and we have made a few sales where this is the case. We sell the  
phones in a bundle
with a PoE switch and a PSTN gateway, and the customer is happy.

Despite what people on this mailing list probably think, not EVERYONE
thinks internet access is a must-have. My next-door neighbors don't
even have a computer ...

Adam's idea sounds to me like a way to interface a P2PSIP system
with a CS SIP system. But I don't see it as the P2PSIP system itself.

- Philip

_______________________________________________
p2p-sip mailing list
p2p-sip at cs.columbia.edu
https://lists.cs.columbia.edu/cucslists/listinfo/p2p-sip


From victor.pascual.avila at gmail.com  Sat Nov 25 09:19:55 2006
From: victor.pascual.avila at gmail.com (=?ISO-8859-1?Q?Victor_Pascual_=C1vila?=)
Date: Sat, 25 Nov 2006 15:19:55 +0100
Subject: [p2p-sip] Fwd: what's wrong with DNS?
In-Reply-To: <24CCCC428EFEA2469BF046DB3C7A8D22141359@namail5.corp.adobe.com>
References: <47959272-70F4-40D0-A599-ED3EF02D9ED5@magma.ca>
	<24CCCC428EFEA2469BF046DB3C7A8D22141359@namail5.corp.adobe.com>
Message-ID: <618e24240611250619u4429c8cbw1a2666d898ae6af9@mail.gmail.com>

Hi,
2006/11/25, Henry Sinnreich <hsinnrei at adobe.com>:
> I would like to add the important application of home network for
> consumer electronics that must also work even when disconnected form the
> Internet.
>
> Eunsoo Shim could probably expand.
>
> Philip Mathews is right with the observation "Despite what people on
> this mailing list probably think":
>
> Dependency on DNS is in last instance a business decision and all the
> major P2P systems out there DO NOT rely on the DNS and will not be
> influenced by these interminable discussion anyway.
I agree with you but I'd like to add that, from my point of view, it
should _not_ rely on any Inet service... It should be a closed system
compatible with other networks but with no dependence...
This is the most important key if you are working in special
situations where infrastructure is not available...
>
> Live with it: P2P need not rely on DNS.
>
> Thanks, Henry
>
> -----Original Message-----
> From: p2p-sip-bounces at cs.columbia.edu
> [mailto:p2p-sip-bounces at cs.columbia.edu] On Behalf Of Philip Matthews
> Sent: Thursday, November 23, 2006 2:43 PM
> To: David Barrett; Adam Fisk
> Cc: P2P-SIP
> Subject: Re: [p2p-sip] Fwd: what's wrong with DNS?
>
>
> On 21-Nov-06, at 19:24 , David Barrett wrote:
>
> >
> > If that's the case, then who *are* we targeting this to?  Is this
> > really
> > being designed exclusively around the needs of the very first hours
> > after
> > disaster strikes as an alternative to walkie talkies or yelling
> > really loud?
>
>
> Our system has the requirement that it must work without Internet
> access,
> and we have made a few sales where this is the case. We sell the
> phones in a bundle
> with a PoE switch and a PSTN gateway, and the customer is happy.
>
> Despite what people on this mailing list probably think, not EVERYONE
> thinks internet access is a must-have. My next-door neighbors don't
> even have a computer ...
>
> Adam's idea sounds to me like a way to interface a P2PSIP system
> with a CS SIP system. But I don't see it as the P2PSIP system itself.
>
> - Philip
>
> _______________________________________________
> p2p-sip mailing list
> p2p-sip at cs.columbia.edu
> https://lists.cs.columbia.edu/cucslists/listinfo/p2p-sip
>
> _______________________________________________
> p2p-sip mailing list
> p2p-sip at cs.columbia.edu
> https://lists.cs.columbia.edu/cucslists/listinfo/p2p-sip
>


-- 
Victor Pascual ?vila

mail:victor.pascual.avila at gmail.com
sip:victor.pascual at iptel.org
linux user: 224443


From dbarrett at quinthar.com  Mon Nov 27 03:05:46 2006
From: dbarrett at quinthar.com (David Barrett)
Date: Mon, 27 Nov 2006 00:05:46 -0800
Subject: [p2p-sip] Fwd:  what's wrong with DNS?
In-Reply-To: <24CCCC428EFEA2469BF046DB3C7A8D22141359@namail5.corp.adobe.com>
Message-ID: <052c01c711fa$deb4b730$0202a8c0@Quinthar>

> -----Original Message-----
> From: Henry Sinnreich [mailto:hsinnrei at adobe.com]
> Subject: RE: [p2p-sip] Fwd: what's wrong with DNS?
> 
> I would like to add the important application of home network for
> consumer electronics that must also work even when disconnected form the
> Internet.

Seems like VoIP in a home network would have a hard time competing with
TALKING VERY LOUD.  Furthermore, anyone sufficiently geeky to have a fully
wired home would probably have a pretty good internet connection, or the
capacity to run a private DNS server -- especially if we made private,
disconnected DNS dead easy.


> Live with it: P2P need not rely on DNS.

I totally, 100% agree that P2P *need* not rely on DNS.  Indeed, it needn't
rely upon IP -- we could even standardize our own wireless protocol to avoid
the patent problems of 802.11.

But just like it's silly to recreate 802.11, IP, UDP, RTP and other
protocols, it seems a waste to rebuild name resolution.

Even though we *needn't* count on DNS, I don't see why we *couldn't* rely
upon it, or indeed why we *shouldn't*.  So far as I've heard, DNS (private
or public) will or could easily be available anywhere VoIP will be used.

Why is it worth the effort to go further?

-david


From dbarrett at quinthar.com  Mon Nov 27 03:39:29 2006
From: dbarrett at quinthar.com (David Barrett)
Date: Mon, 27 Nov 2006 00:39:29 -0800
Subject: [p2p-sip] Secure decentralized name resolution? (Was "what's wrong
	with DNS?")
In-Reply-To: <1164254505.12268.7.camel@sioux.systems.cs.cornell.edu>
Message-ID: <053601c711ff$95e60180$0202a8c0@Quinthar>

> -----Original Message-----
> From: Saikat Guha [mailto:saikat at cs.cornell.edu]
> Subject: Re: [p2p-sip] Fwd: what's wrong with DNS?
> 
> Skype actually resorts to hard-coded IP addresses in the binary for
> bootstrap; after the first successful bootstrap, it downloads additional
> IPs.

Thanks for the correction.  So Skype uses fixed, hard-coded IP addresses
rather than DNS to find the central server. [1]  In that sense, it doesn't
depend on DNS.  Rather, it's even worse: it's even more centralized and less
open than DNS.

Thus at the end of the day, the most successful P2P VoIP application has at
its very heart a completely closed client/server architecture.  All the
fancy P2P stuff is neat, but it can't exist without this big,
privately-owned central infrastructure.  If that central authentication
infrastructure goes down, nobody can login, and Skype goes offline.

So the question is: how can P2P-SIP not suffer the same failing?

Adam proposed that instead of trusting a single private company with the
world's directory, we leverage the existing global consortium of domain
registrars and adopt the existing international conflict resolution
framework of the IANA.  It's there, it works.  It has decades of experience
resolving all the problems that we're just looking at now.

It does what we want to done.  Why not "buy" rather than "build"?

Indeed, I still don't understand what has been proposed we "build" as a
replacement.  How is it possible to manage a unique namespace without
central authorities?  How, fundamentally, does it ensure that anyone in the
world who dials my number rings my phone?  And if someone somehow steals my
identity, how do I get it back?  How can it solve all this and still have
something human-memorable that I can print on my business card?

I've heard people talk about DHTs.  But how can I prevent you from hijacking
my entry?

I've heard people talk about PKI.  But where do you look up the public key
associated with my name, and how can I change my public key without letting
a hacker do the same?

This is much bigger than a technology problem; indeed, technology is an
afterthought.  Skype went the obvious and smart route: build a central
directory.  Adam proposed a decentralized and distributed solution: dynamic
DNS.

What is the competing proposal, and why is it better than DNS?

-david

----------------------------------------------------------------------------
[1] An Analysis of the Skype Peer-to-Peer Internet Telephony Protocol, Baset
& Schulzrinne, 2004

4.2.2 Login Server
After a SC is connected to a SN, the SC must authenticate the user
name and password with the Skype login server. The login server
is the only central component in the Skype network. It stores
Skype user names and passwords and ensures that Skype user
names are unique across the Skype name space. SC must
authenticate itself with login server for a successful login. During
our experiments we observed that SC always exchanged data over
TCP with a node whose IP address was 80.160.91.11. We believe
that this node is the login server. A reverse lookup of this IP
address retrieved NS records whose values are ns14.inet.tele.dk
and ns15.inet.tele.dk. It thus appears from the reverse lookup that
the login server is hosted by an ISP based in Denmark.


From billmcc64 at gmail.com  Mon Nov 27 10:15:16 2006
From: billmcc64 at gmail.com (Bill Mccormick)
Date: Mon, 27 Nov 2006 10:15:16 -0500
Subject: [p2p-sip] Secure decentralized name resolution? (Was "what's
	wrong with DNS?")
In-Reply-To: <053601c711ff$95e60180$0202a8c0@Quinthar>
References: <1164254505.12268.7.camel@sioux.systems.cs.cornell.edu>
	<053601c711ff$95e60180$0202a8c0@Quinthar>
Message-ID: <fb4306550611270715i4804f031r377e372ec5d5f18e@mail.gmail.com>

Hi David,

The technology difference that I see is that a DHT will push out a
change in URI/IP address binding, whereas DNS (including DDNS) will
rely on cache flushing to occur in a timely manner.   Maybe this is
good enough for some people, maybe not.   I know it's not fast enough
for the 3GPP IMS guys...

I firmly believe the devil is in the details in this stuff.   I find
it hard to believe that we
could use the existing DNS off the shelf.    David Bryan et al have
put together a pretty reasonable description detailing the P2PSIP
proposal - and the charter does specifically exclude the use of DDNS.

If you really want to convince everyone, how about you produce a
document equivalent to the 'A P2P Approach to SIP Registration and
Resource Location' document.   It's pretty hard to compare the 2
approaches based on the traffic on the mailing list.

Bill

On 11/27/06, David Barrett <dbarrett at quinthar.com> wrote:
> > -----Original Message-----
> > From: Saikat Guha [mailto:saikat at cs.cornell.edu]
> > Subject: Re: [p2p-sip] Fwd: what's wrong with DNS?
> >
> > Skype actually resorts to hard-coded IP addresses in the binary for
> > bootstrap; after the first successful bootstrap, it downloads additional
> > IPs.
>
> Thanks for the correction.  So Skype uses fixed, hard-coded IP addresses
> rather than DNS to find the central server. [1]  In that sense, it doesn't
> depend on DNS.  Rather, it's even worse: it's even more centralized and less
> open than DNS.
>
> Thus at the end of the day, the most successful P2P VoIP application has at
> its very heart a completely closed client/server architecture.  All the
> fancy P2P stuff is neat, but it can't exist without this big,
> privately-owned central infrastructure.  If that central authentication
> infrastructure goes down, nobody can login, and Skype goes offline.
>
> So the question is: how can P2P-SIP not suffer the same failing?
>
> Adam proposed that instead of trusting a single private company with the
> world's directory, we leverage the existing global consortium of domain
> registrars and adopt the existing international conflict resolution
> framework of the IANA.  It's there, it works.  It has decades of experience
> resolving all the problems that we're just looking at now.
>
> It does what we want to done.  Why not "buy" rather than "build"?
>
> Indeed, I still don't understand what has been proposed we "build" as a
> replacement.  How is it possible to manage a unique namespace without
> central authorities?  How, fundamentally, does it ensure that anyone in the
> world who dials my number rings my phone?  And if someone somehow steals my
> identity, how do I get it back?  How can it solve all this and still have
> something human-memorable that I can print on my business card?
>
> I've heard people talk about DHTs.  But how can I prevent you from hijacking
> my entry?
>
> I've heard people talk about PKI.  But where do you look up the public key
> associated with my name, and how can I change my public key without letting
> a hacker do the same?
>
> This is much bigger than a technology problem; indeed, technology is an
> afterthought.  Skype went the obvious and smart route: build a central
> directory.  Adam proposed a decentralized and distributed solution: dynamic
> DNS.
>
> What is the competing proposal, and why is it better than DNS?
>
> -david
>
> ----------------------------------------------------------------------------
> [1] An Analysis of the Skype Peer-to-Peer Internet Telephony Protocol, Baset
> & Schulzrinne, 2004
>
> 4.2.2 Login Server
> After a SC is connected to a SN, the SC must authenticate the user
> name and password with the Skype login server. The login server
> is the only central component in the Skype network. It stores
> Skype user names and passwords and ensures that Skype user
> names are unique across the Skype name space. SC must
> authenticate itself with login server for a successful login. During
> our experiments we observed that SC always exchanged data over
> TCP with a node whose IP address was 80.160.91.11. We believe
> that this node is the login server. A reverse lookup of this IP
> address retrieved NS records whose values are ns14.inet.tele.dk
> and ns15.inet.tele.dk. It thus appears from the reverse lookup that
> the login server is hosted by an ISP based in Denmark.
>
> _______________________________________________
> p2p-sip mailing list
> p2p-sip at cs.columbia.edu
> https://lists.cs.columbia.edu/cucslists/listinfo/p2p-sip
>


-- 
Bill McCormick

From hsinnrei at adobe.com  Mon Nov 27 11:08:00 2006
From: hsinnrei at adobe.com (Henry Sinnreich)
Date: Mon, 27 Nov 2006 08:08:00 -0800
Subject: [p2p-sip] Fwd:  what's wrong with DNS?
In-Reply-To: <052c01c711fa$deb4b730$0202a8c0@Quinthar>
Message-ID: <24CCCC428EFEA2469BF046DB3C7A8D2214135D@namail5.corp.adobe.com>

David,

> Seems like VoIP in a home network would have a hard time competing 
> with TALKING VERY LOUD.

It's not about VoIP only, but about all consumer electronic devices in
the home that include entertainment, appliances, cars, home security,
garage door, sprinkler system, etc.

You bring home a new camera and it finds the printers, etc. or you can
use any remote control as a phone as well...  

There are papers on this from Panasonic. See p2psip.org.

It is hard to predict what technology will eventually link our consumer
electronic devices, but DNS has not popped up anywhere except on this
list.

Henry


-----Original Message-----
From: David Barrett [mailto:dbarrett at quinthar.com] 
Sent: Monday, November 27, 2006 2:06 AM
To: Henry Sinnreich; 'Philip Matthews'; 'Adam Fisk'
Cc: 'P2P-SIP'; 'Eunsoo Shim'
Subject: RE: [p2p-sip] Fwd: what's wrong with DNS?

> -----Original Message-----
> From: Henry Sinnreich [mailto:hsinnrei at adobe.com]
> Subject: RE: [p2p-sip] Fwd: what's wrong with DNS?
> 
> I would like to add the important application of home network for
> consumer electronics that must also work even when disconnected form
the
> Internet.

Seems like VoIP in a home network would have a hard time competing with
TALKING VERY LOUD.  Furthermore, anyone sufficiently geeky to have a
fully
wired home would probably have a pretty good internet connection, or the
capacity to run a private DNS server -- especially if we made private,
disconnected DNS dead easy.


> Live with it: P2P need not rely on DNS.

I totally, 100% agree that P2P *need* not rely on DNS.  Indeed, it
needn't
rely upon IP -- we could even standardize our own wireless protocol to
avoid
the patent problems of 802.11.

But just like it's silly to recreate 802.11, IP, UDP, RTP and other
protocols, it seems a waste to rebuild name resolution.

Even though we *needn't* count on DNS, I don't see why we *couldn't*
rely
upon it, or indeed why we *shouldn't*.  So far as I've heard, DNS
(private
or public) will or could easily be available anywhere VoIP will be used.

Why is it worth the effort to go further?

-david



From huang-ming.pan at comcast.net  Mon Nov 27 11:23:53 2006
From: huang-ming.pan at comcast.net (Peter Pan)
Date: Mon, 27 Nov 2006 08:23:53 -0800
Subject: [p2p-sip] Fwd:  what's wrong with DNS?
References: <24CCCC428EFEA2469BF046DB3C7A8D2214135D@namail5.corp.adobe.com>
Message-ID: <001301c71240$6e76e100$030aa8c0@comcast.net>

> It is hard to predict what technology will eventually link our consumer
> electronic devices, but DNS has not popped up anywhere except on this
> list.

Within LAN/WLANs, uPnP will do it.

Peter
----- Original Message -----
From: "Henry Sinnreich" <hsinnrei at adobe.com>
To: "David Barrett" <dbarrett at quinthar.com>; "Philip Matthews"
<philip_matthews at magma.ca>; "Adam Fisk" <adamfisk at gmail.com>
Cc: "P2P-SIP" <p2p-sip at cs.columbia.edu>
Sent: Monday, November 27, 2006 8:08 AM
Subject: Re: [p2p-sip] Fwd: what's wrong with DNS?


> David,
>
> > Seems like VoIP in a home network would have a hard time competing
> > with TALKING VERY LOUD.
>
> It's not about VoIP only, but about all consumer electronic devices in
> the home that include entertainment, appliances, cars, home security,
> garage door, sprinkler system, etc.
>
> You bring home a new camera and it finds the printers, etc. or you can
> use any remote control as a phone as well...
>
> There are papers on this from Panasonic. See p2psip.org.
>
> It is hard to predict what technology will eventually link our consumer
> electronic devices, but DNS has not popped up anywhere except on this
> list.
>
> Henry
>
>
> -----Original Message-----
> From: David Barrett [mailto:dbarrett at quinthar.com]
> Sent: Monday, November 27, 2006 2:06 AM
> To: Henry Sinnreich; 'Philip Matthews'; 'Adam Fisk'
> Cc: 'P2P-SIP'; 'Eunsoo Shim'
> Subject: RE: [p2p-sip] Fwd: what's wrong with DNS?
>
> > -----Original Message-----
> > From: Henry Sinnreich [mailto:hsinnrei at adobe.com]
> > Subject: RE: [p2p-sip] Fwd: what's wrong with DNS?
> >
> > I would like to add the important application of home network for
> > consumer electronics that must also work even when disconnected form
> the
> > Internet.
>
> Seems like VoIP in a home network would have a hard time competing with
> TALKING VERY LOUD.  Furthermore, anyone sufficiently geeky to have a
> fully
> wired home would probably have a pretty good internet connection, or the
> capacity to run a private DNS server -- especially if we made private,
> disconnected DNS dead easy.
>
>
> > Live with it: P2P need not rely on DNS.
>
> I totally, 100% agree that P2P *need* not rely on DNS.  Indeed, it
> needn't
> rely upon IP -- we could even standardize our own wireless protocol to
> avoid
> the patent problems of 802.11.
>
> But just like it's silly to recreate 802.11, IP, UDP, RTP and other
> protocols, it seems a waste to rebuild name resolution.
>
> Even though we *needn't* count on DNS, I don't see why we *couldn't*
> rely
> upon it, or indeed why we *shouldn't*.  So far as I've heard, DNS
> (private
> or public) will or could easily be available anywhere VoIP will be used.
>
> Why is it worth the effort to go further?
>
> -david
>
>
> _______________________________________________
> p2p-sip mailing list
> p2p-sip at cs.columbia.edu
> https://lists.cs.columbia.edu/cucslists/listinfo/p2p-sip


From slavitch at gmail.com  Mon Nov 27 11:20:44 2006
From: slavitch at gmail.com (Michael Slavitch)
Date: Mon, 27 Nov 2006 11:20:44 -0500
Subject: [p2p-sip] Fwd: what's wrong with DNS?
In-Reply-To: <24CCCC428EFEA2469BF046DB3C7A8D2214135D@namail5.corp.adobe.com>
References: <052c01c711fa$deb4b730$0202a8c0@Quinthar>
	<24CCCC428EFEA2469BF046DB3C7A8D2214135D@namail5.corp.adobe.com>
Message-ID: <1fc32c8f0611270820m7623a139k23f0a10167dde940@mail.gmail.com>

As a form of assessing identity DNS is particularly weak and requires
considerable care and expert feeding as provisioning is always
ultimately human-based requiring some kind of authority.  A
self-organizing automated system needs no such crutches.

A key provision of P2PSIP is to eliminate the need for such
authorities in the first place because they are superfluous for the
application.  An extended social network needs no central domain
administrator nor a central certificate authority.   It needs at most
a human facilitator not an engineer.

Such social networks are not trite. For example social networks are
the basis of all peer reviews for medical journals.  Shouting doesn't
work when 500 doctors are to be consulted, on several continents, with
varied access to the Internet, and over time.  The person running the
peer review is a journal editor, a person who uses end-user tools, not
a network engineer.  DNS isn't the selection criteria here, reputation
is.

M



On 11/27/06, Henry Sinnreich <hsinnrei at adobe.com> wrote:
> David,
>
> > Seems like VoIP in a home network would have a hard time competing
> > with TALKING VERY LOUD.
>
> It's not about VoIP only, but about all consumer electronic devices in
> the home that include entertainment, appliances, cars, home security,
> garage door, sprinkler system, etc.
>
> You bring home a new camera and it finds the printers, etc. or you can
> use any remote control as a phone as well...
>
> There are papers on this from Panasonic. See p2psip.org.
>
> It is hard to predict what technology will eventually link our consumer
> electronic devices, but DNS has not popped up anywhere except on this
> list.
>
> Henry
>
>
> -----Original Message-----
> From: David Barrett [mailto:dbarrett at quinthar.com]
> Sent: Monday, November 27, 2006 2:06 AM
> To: Henry Sinnreich; 'Philip Matthews'; 'Adam Fisk'
> Cc: 'P2P-SIP'; 'Eunsoo Shim'
> Subject: RE: [p2p-sip] Fwd: what's wrong with DNS?
>
> > -----Original Message-----
> > From: Henry Sinnreich [mailto:hsinnrei at adobe.com]
> > Subject: RE: [p2p-sip] Fwd: what's wrong with DNS?
> >
> > I would like to add the important application of home network for
> > consumer electronics that must also work even when disconnected form
> the
> > Internet.
>
> Seems like VoIP in a home network would have a hard time competing with
> TALKING VERY LOUD.  Furthermore, anyone sufficiently geeky to have a
> fully
> wired home would probably have a pretty good internet connection, or the
> capacity to run a private DNS server -- especially if we made private,
> disconnected DNS dead easy.
>
>
> > Live with it: P2P need not rely on DNS.
>
> I totally, 100% agree that P2P *need* not rely on DNS.  Indeed, it
> needn't
> rely upon IP -- we could even standardize our own wireless protocol to
> avoid
> the patent problems of 802.11.
>
> But just like it's silly to recreate 802.11, IP, UDP, RTP and other
> protocols, it seems a waste to rebuild name resolution.
>
> Even though we *needn't* count on DNS, I don't see why we *couldn't*
> rely
> upon it, or indeed why we *shouldn't*.  So far as I've heard, DNS
> (private
> or public) will or could easily be available anywhere VoIP will be used.
>
> Why is it worth the effort to go further?
>
> -david
>
>
> _______________________________________________
> p2p-sip mailing list
> p2p-sip at cs.columbia.edu
> https://lists.cs.columbia.edu/cucslists/listinfo/p2p-sip
>

From br at brianrosen.net  Mon Nov 27 11:25:52 2006
From: br at brianrosen.net (Brian Rosen)
Date: Mon, 27 Nov 2006 11:25:52 -0500
Subject: [p2p-sip] Secure decentralized name resolution? (Was "what's
	wrongwith DNS?")
In-Reply-To: <053601c711ff$95e60180$0202a8c0@Quinthar>
Message-ID: <052801c71240$c16a13f0$640fa8c0@cis.neustar.com>

Well, for one thing, there are "credentials".  Think of this as a username,
within some domain.  Might also have some other things related to security.
The domain ("overlay network") is the domain of users that are allowed into
the domain.  

I think we want controls on that.

We have decided, quite a while ago, that a "centralized" credentialing
service was a useful thing to have.

DNS would leave everyone to themselves on SPIT.  I think we now understand
that such an environment is not a good idea.  

We could try leaveraging this high "risk" project on other high risk
projects like DKIM.  I think, at least for now, that we shouldn't do that,
and our current notion of overlay networks with a credentialing service is
the right model.

Brian

> -----Original Message-----
> From: p2p-sip-bounces at cs.columbia.edu [mailto:p2p-sip-
> bounces at cs.columbia.edu] On Behalf Of David Barrett
> Sent: Monday, November 27, 2006 3:39 AM
> To: 'P2P-SIP'
> Subject: [p2p-sip] Secure decentralized name resolution? (Was "what's
> wrongwith DNS?")
> 
> > -----Original Message-----
> > From: Saikat Guha [mailto:saikat at cs.cornell.edu]
> > Subject: Re: [p2p-sip] Fwd: what's wrong with DNS?
> >
> > Skype actually resorts to hard-coded IP addresses in the binary for
> > bootstrap; after the first successful bootstrap, it downloads additional
> > IPs.
> 
> Thanks for the correction.  So Skype uses fixed, hard-coded IP addresses
> rather than DNS to find the central server. [1]  In that sense, it doesn't
> depend on DNS.  Rather, it's even worse: it's even more centralized and
> less
> open than DNS.
> 
> Thus at the end of the day, the most successful P2P VoIP application has
> at
> its very heart a completely closed client/server architecture.  All the
> fancy P2P stuff is neat, but it can't exist without this big,
> privately-owned central infrastructure.  If that central authentication
> infrastructure goes down, nobody can login, and Skype goes offline.
> 
> So the question is: how can P2P-SIP not suffer the same failing?
> 
> Adam proposed that instead of trusting a single private company with the
> world's directory, we leverage the existing global consortium of domain
> registrars and adopt the existing international conflict resolution
> framework of the IANA.  It's there, it works.  It has decades of
> experience
> resolving all the problems that we're just looking at now.
> 
> It does what we want to done.  Why not "buy" rather than "build"?
> 
> Indeed, I still don't understand what has been proposed we "build" as a
> replacement.  How is it possible to manage a unique namespace without
> central authorities?  How, fundamentally, does it ensure that anyone in
> the
> world who dials my number rings my phone?  And if someone somehow steals
> my
> identity, how do I get it back?  How can it solve all this and still have
> something human-memorable that I can print on my business card?
> 
> I've heard people talk about DHTs.  But how can I prevent you from
> hijacking
> my entry?
> 
> I've heard people talk about PKI.  But where do you look up the public key
> associated with my name, and how can I change my public key without
> letting
> a hacker do the same?
> 
> This is much bigger than a technology problem; indeed, technology is an
> afterthought.  Skype went the obvious and smart route: build a central
> directory.  Adam proposed a decentralized and distributed solution:
> dynamic
> DNS.
> 
> What is the competing proposal, and why is it better than DNS?
> 
> -david
> 
> --------------------------------------------------------------------------
> --
> [1] An Analysis of the Skype Peer-to-Peer Internet Telephony Protocol,
> Baset
> & Schulzrinne, 2004
> 
> 4.2.2 Login Server
> After a SC is connected to a SN, the SC must authenticate the user
> name and password with the Skype login server. The login server
> is the only central component in the Skype network. It stores
> Skype user names and passwords and ensures that Skype user
> names are unique across the Skype name space. SC must
> authenticate itself with login server for a successful login. During
> our experiments we observed that SC always exchanged data over
> TCP with a node whose IP address was 80.160.91.11. We believe
> that this node is the login server. A reverse lookup of this IP
> address retrieved NS records whose values are ns14.inet.tele.dk
> and ns15.inet.tele.dk. It thus appears from the reverse lookup that
> the login server is hosted by an ISP based in Denmark.
> 
> _______________________________________________
> p2p-sip mailing list
> p2p-sip at cs.columbia.edu
> https://lists.cs.columbia.edu/cucslists/listinfo/p2p-sip


From adamfisk at gmail.com  Mon Nov 27 11:49:42 2006
From: adamfisk at gmail.com (Adam Fisk)
Date: Mon, 27 Nov 2006 11:49:42 -0500
Subject: [p2p-sip] Fwd: what's wrong with DNS?
In-Reply-To: <001301c71240$6e76e100$030aa8c0@comcast.net>
References: <24CCCC428EFEA2469BF046DB3C7A8D2214135D@namail5.corp.adobe.com>
	<001301c71240$6e76e100$030aa8c0@comcast.net>
Message-ID: <f40963db0611270849o15919b4bl8ffdf561a24f700@mail.gmail.com>

On 11/27/06, Peter Pan <huang-ming.pan at comcast.net> wrote:
>
> > It is hard to predict what technology will eventually link our consumer
> > electronic devices, but DNS has not popped up anywhere except on this
> > list.
>
> Within LAN/WLANs, uPnP will do it.



Exactly.  It would be quite a shock for p2p sip to win out in that battle.
I worry this list is trying to create some sort of panacea for all the
problems of the Internet instead of sticking to solving discrete problems
well.  Don't get me wrong, the technology's as cool as it gets, but I worry
it's seductive and becoming a solution in search of a problem.

We've addressed the entire spectrum just on this thread from social networks
to consumer electronic devices to completely disconnected networks to car
based networks, and undoubtedly more that aren't coming to mind at the
moment.  P2P SIP is really the answer for all of these?  Does that scope
strike anyone else as way too ambitious?  Seems cool for a research problem,
but we really think we're going to deploy standards to address all of those
use cases?

-Adam
-------------- next part --------------
An HTML attachment was scrubbed...
URL: https://lists.cs.columbia.edu/pipermail/p2p-sip/attachments/20061127/99fe47fc/attachment.html

From fwmiller at cornfed.com  Mon Nov 27 12:02:19 2006
From: fwmiller at cornfed.com (Frank W. Miller)
Date: Mon, 27 Nov 2006 10:02:19 -0700
Subject: [p2p-sip] Secure decentralized name resolution? (Was "what's
	wrongwith DNS?")
In-Reply-To: <053601c711ff$95e60180$0202a8c0@Quinthar>
Message-ID: <200611271702.kARH2LL10302@celine.siteprotect.com>



-----Original Message-----
From: p2p-sip-bounces at cs.columbia.edu
[mailto:p2p-sip-bounces at cs.columbia.edu] On Behalf Of David Barrett
Sent: Monday, November 27, 2006 1:39 AM
To: 'P2P-SIP'
Subject: [p2p-sip] Secure decentralized name resolution? (Was "what's
wrongwith DNS?")

> -----Original Message-----
> From: Saikat Guha [mailto:saikat at cs.cornell.edu]
> Subject: Re: [p2p-sip] Fwd: what's wrong with DNS?
> 
> Skype actually resorts to hard-coded IP addresses in the binary for
> bootstrap; after the first successful bootstrap, it downloads additional
> IPs.

Thanks for the correction.  So Skype uses fixed, hard-coded IP addresses
rather than DNS to find the central server. [1]  In that sense, it doesn't
depend on DNS.  Rather, it's even worse: it's even more centralized and less
open than DNS.


FM: I'd like to draw attention to this.  Whether or not there is any kind of
central server function that is part of this effort is a (pardon the pun)
central question.

At the very least, it is likely that some kind of centralized, or at least
authoritative service will be necessary for authorization and
authentication.  So, while all this discussion about DNS is interesting, I
think it may be secondary to some larger architectural considerations.
Perhaps the discussion should take up at least this larger question prior to
worrying about naming.

So, will the protocol disallow any kind of centralized functionality?

If not, what specific functions will allowed to be centralized and which
will not?  Here's a list:

1) Authorization: i.e. I call myself something and the P2P cloud authorized
me and me alone to call myself that

2) Authentication: i.e. I tell somebody who I am and that somebody has the
ability to determine whether I am who I say I am


3) Location discovery: i.e. If I want to find "somebody", figuring out where
that somebody is in the overlay

4) Signaling: i.e. telling somebody that I want to setup a session and
carrying that out

5) Media: i.e. the session media streams

IMHO, the first two items require some sort of centralized (or at least
authoritative) function to keep all the user client software honest.
Thoughts?

FM



From hsinnrei at adobe.com  Mon Nov 27 12:50:17 2006
From: hsinnrei at adobe.com (Henry Sinnreich)
Date: Mon, 27 Nov 2006 09:50:17 -0800
Subject: [p2p-sip] Fwd:  what's wrong with DNS?
In-Reply-To: <001301c71240$6e76e100$030aa8c0@comcast.net>
Message-ID: <24CCCC428EFEA2469BF046DB3C7A8D22141367@namail5.corp.adobe.com>

> Within LAN/WLANs, uPnP will do it.

I am not sure if UPnP meets all the security requirements for the
control of home systems. Besides, it is not a given that all consumer
electronic devices will be developed along a Wintel specification.

Is there a UPnP expert on the list to brief us on the status of UPnP
security and acceptance for home electronics?

Thanks, Henry

-----Original Message-----
From: Peter Pan [mailto:huang-ming.pan at comcast.net] 
Sent: Monday, November 27, 2006 10:24 AM
To: Henry Sinnreich; David Barrett; Philip Matthews; Adam Fisk
Cc: P2P-SIP
Subject: Re: [p2p-sip] Fwd: what's wrong with DNS?

> It is hard to predict what technology will eventually link our
consumer
> electronic devices, but DNS has not popped up anywhere except on this
> list.

Within LAN/WLANs, uPnP will do it.

Peter
----- Original Message -----
From: "Henry Sinnreich" <hsinnrei at adobe.com>
To: "David Barrett" <dbarrett at quinthar.com>; "Philip Matthews"
<philip_matthews at magma.ca>; "Adam Fisk" <adamfisk at gmail.com>
Cc: "P2P-SIP" <p2p-sip at cs.columbia.edu>
Sent: Monday, November 27, 2006 8:08 AM
Subject: Re: [p2p-sip] Fwd: what's wrong with DNS?


> David,
>
> > Seems like VoIP in a home network would have a hard time competing
> > with TALKING VERY LOUD.
>
> It's not about VoIP only, but about all consumer electronic devices in
> the home that include entertainment, appliances, cars, home security,
> garage door, sprinkler system, etc.
>
> You bring home a new camera and it finds the printers, etc. or you can
> use any remote control as a phone as well...
>
> There are papers on this from Panasonic. See p2psip.org.
>
> It is hard to predict what technology will eventually link our
consumer
> electronic devices, but DNS has not popped up anywhere except on this
> list.
>
> Henry
>
>
> -----Original Message-----
> From: David Barrett [mailto:dbarrett at quinthar.com]
> Sent: Monday, November 27, 2006 2:06 AM
> To: Henry Sinnreich; 'Philip Matthews'; 'Adam Fisk'
> Cc: 'P2P-SIP'; 'Eunsoo Shim'
> Subject: RE: [p2p-sip] Fwd: what's wrong with DNS?
>
> > -----Original Message-----
> > From: Henry Sinnreich [mailto:hsinnrei at adobe.com]
> > Subject: RE: [p2p-sip] Fwd: what's wrong with DNS?
> >
> > I would like to add the important application of home network for
> > consumer electronics that must also work even when disconnected form
> the
> > Internet.
>
> Seems like VoIP in a home network would have a hard time competing
with
> TALKING VERY LOUD.  Furthermore, anyone sufficiently geeky to have a
> fully
> wired home would probably have a pretty good internet connection, or
the
> capacity to run a private DNS server -- especially if we made private,
> disconnected DNS dead easy.
>
>
> > Live with it: P2P need not rely on DNS.
>
> I totally, 100% agree that P2P *need* not rely on DNS.  Indeed, it
> needn't
> rely upon IP -- we could even standardize our own wireless protocol to
> avoid
> the patent problems of 802.11.
>
> But just like it's silly to recreate 802.11, IP, UDP, RTP and other
> protocols, it seems a waste to rebuild name resolution.
>
> Even though we *needn't* count on DNS, I don't see why we *couldn't*
> rely
> upon it, or indeed why we *shouldn't*.  So far as I've heard, DNS
> (private
> or public) will or could easily be available anywhere VoIP will be
used.
>
> Why is it worth the effort to go further?
>
> -david
>
>
> _______________________________________________
> p2p-sip mailing list
> p2p-sip at cs.columbia.edu
> https://lists.cs.columbia.edu/cucslists/listinfo/p2p-sip



From br at brianrosen.net  Mon Nov 27 13:05:59 2006
From: br at brianrosen.net (Brian Rosen)
Date: Mon, 27 Nov 2006 13:05:59 -0500
Subject: [p2p-sip] Secure decentralized name resolution? (Was
	"what'swrongwith DNS?")
In-Reply-To: <200611271702.kARH2LL10302@celine.siteprotect.com>
Message-ID: <05d801c7124e$b5696200$640fa8c0@cis.neustar.com>

I think we have agreed that issuing a credential is centralized.  Checking
it is probably not.  Depending on exactly how you want to define
"authorization", it may be centralized or not.

All the rest are decentralized among the peers.  We haven't really decided
how bootstrapping works.  There may be some kind of distinguished entities
that assist bootstrapping.

Brian

> -----Original Message-----
> From: p2p-sip-bounces at cs.columbia.edu [mailto:p2p-sip-
> bounces at cs.columbia.edu] On Behalf Of Frank W. Miller
> Sent: Monday, November 27, 2006 12:02 PM
> To: 'David Barrett'; 'P2P-SIP'
> Subject: Re: [p2p-sip] Secure decentralized name resolution? (Was
> "what'swrongwith DNS?")
> 
> 
> 
> -----Original Message-----
> From: p2p-sip-bounces at cs.columbia.edu
> [mailto:p2p-sip-bounces at cs.columbia.edu] On Behalf Of David Barrett
> Sent: Monday, November 27, 2006 1:39 AM
> To: 'P2P-SIP'
> Subject: [p2p-sip] Secure decentralized name resolution? (Was "what's
> wrongwith DNS?")
> 
> > -----Original Message-----
> > From: Saikat Guha [mailto:saikat at cs.cornell.edu]
> > Subject: Re: [p2p-sip] Fwd: what's wrong with DNS?
> >
> > Skype actually resorts to hard-coded IP addresses in the binary for
> > bootstrap; after the first successful bootstrap, it downloads additional
> > IPs.
> 
> Thanks for the correction.  So Skype uses fixed, hard-coded IP addresses
> rather than DNS to find the central server. [1]  In that sense, it doesn't
> depend on DNS.  Rather, it's even worse: it's even more centralized and
> less
> open than DNS.
> 
> 
> FM: I'd like to draw attention to this.  Whether or not there is any kind
> of
> central server function that is part of this effort is a (pardon the pun)
> central question.
> 
> At the very least, it is likely that some kind of centralized, or at least
> authoritative service will be necessary for authorization and
> authentication.  So, while all this discussion about DNS is interesting, I
> think it may be secondary to some larger architectural considerations.
> Perhaps the discussion should take up at least this larger question prior
> to
> worrying about naming.
> 
> So, will the protocol disallow any kind of centralized functionality?
> 
> If not, what specific functions will allowed to be centralized and which
> will not?  Here's a list:
> 
> 1) Authorization: i.e. I call myself something and the P2P cloud
> authorized
> me and me alone to call myself that
> 
> 2) Authentication: i.e. I tell somebody who I am and that somebody has the
> ability to determine whether I am who I say I am
> 
> 
> 3) Location discovery: i.e. If I want to find "somebody", figuring out
> where
> that somebody is in the overlay
> 
> 4) Signaling: i.e. telling somebody that I want to setup a session and
> carrying that out
> 
> 5) Media: i.e. the session media streams
> 
> IMHO, the first two items require some sort of centralized (or at least
> authoritative) function to keep all the user client software honest.
> Thoughts?
> 
> FM
> 
> 
> _______________________________________________
> p2p-sip mailing list
> p2p-sip at cs.columbia.edu
> https://lists.cs.columbia.edu/cucslists/listinfo/p2p-sip


From matthew at matthew.at  Mon Nov 27 13:10:43 2006
From: matthew at matthew.at (Matthew Kaufman)
Date: Mon, 27 Nov 2006 10:10:43 -0800 (PST)
Subject: [p2p-sip] Fwd:  what's wrong with DNS?
In-Reply-To: <001301c71240$6e76e100$030aa8c0@comcast.net>
Message-ID: <200611271810.kARIAhc91931@where.matthew.at>

Peter Pan:
>> It is hard to predict what technology will eventually link our consumer
>> electronic devices, but DNS has not popped up anywhere except on this
>> list.
>
> Within LAN/WLANs, uPnP will do it.

Along with zeroconf using mDNS (already being used to find printers, webcams, etc.
in the home or SOHO network environment)

Which gets back to my previous question about use cases that can't use any of
global DNS, locally-administered private DNS, or mDNS.

Matthew Kaufman
matthew at matthew.at


From matthew at matthew.at  Mon Nov 27 13:18:00 2006
From: matthew at matthew.at (Matthew Kaufman)
Date: Mon, 27 Nov 2006 10:18:00 -0800 (PST)
Subject: [p2p-sip] Fwd: what's wrong with DNS?
In-Reply-To: <1fc32c8f0611270820m7623a139k23f0a10167dde940@mail.gmail.com>
Message-ID: <200611271818.kARII0191955@where.matthew.at>

Michael Slavitch:
> As a form of assessing identity DNS is particularly weak and requires
> considerable care and expert feeding as provisioning is always
> ultimately human-based requiring some kind of authority. 

Agreed. Though in small disconnected networks, zeroconf using mDNS avoids
much of this, and in large networks the IT guys or your ISP manage the
complexity for you (though they haven't solved the security problem yet)

> A self-organizing automated system needs no such crutches.

Except for bootstrapping to get the first two nodes talking. Which, if on
the Internet, means a way to find IP addresses of those nodes. Which, on
today's Internet, could use DNS. Or, as the Skype folks do, hardcoded bootstrap
addresses. But either way, bootstrap is such a crutch.

> A key provision of P2PSIP is to eliminate the need for such
> authorities in the first place because they are superfluous for the
> application.  An extended social network needs no central domain
> administrator nor a central certificate authority.   It needs at most
> a human facilitator not an engineer.

So P2PSIP includes a web-of-trust security model? If so, why not leverage
existing solutions for that as well (like PGP)?

> Such social networks are not trite. For example social networks are
> the basis of all peer reviews for medical journals.  Shouting doesn't
> work when 500 doctors are to be consulted, on several continents, with
> varied access to the Internet, and over time.  The person running the
> peer review is a journal editor, a person who uses end-user tools, not
> a network engineer.  DNS isn't the selection criteria here, reputation
> is.

But DNS is how the journal's mailing list manager finds the mail hosts
for each of the 500 doctors. Delivery of the article and reply comments
across several continents over time works today just fine without P2P
or DHTs or reputation-based web-of-trust security models. See this mailing
list as an existence proof.

Matthew Kaufman
matthew at matthew.at


From dbarrett at quinthar.com  Tue Nov 28 01:49:34 2006
From: dbarrett at quinthar.com (David Barrett)
Date: Mon, 27 Nov 2006 22:49:34 -0800
Subject: [p2p-sip] Fwd: what's wrong with DNS?
In-Reply-To: <200611271818.kARII0191955@where.matthew.at>
Message-ID: <015301c712b9$5daeb950$0200a8c0@Quinthar>

I think Matthew highlights the Achilles heel of so-called "pure" P2P: it
never is.  In the real world, there's always a central service behind the
curtain:

- In Skype's case, it's a central directory at a hard-coded IP address.

- In P2P-SIP's case (if Frank and Brian are right) is some kind of
centralized authorization/authentication/bootstrapping service.

So granted, there's the question of "how do you find this service?"  The
obvious choice favored by every other application on the internet (and most
private networks) is "DNS".

But ignoring that, there's an even more important question:

	"Who controls this service?"

If we build some fancy DHT solution we imply "some unnamed benevolent
dictator yet to be appointed" will rule us all.  We've sorta assumed this
(or similarly vague explanations) for the entire course of the P2P-SIP
discussion.  For a long while I was not only on board, I was promoting this
"let's solve security later" approach.

But then Adam proposed a different owner, the IANA.  And the way to do that
is through the existing, tried-and-true interface of DNS (specifically,
dynamic DNS).  I think this is obvious and brilliant.

Building a nonpartial, internationally represented, neutral global third
party isn't easy, and while the IANA hasn't wildly succeeded in this,
they're closer than anybody else (and decades ahead of anything we'd invent
while searching for our benevolent dictator).


Now I'll grant DNS is limited to only, say, 99.999% of users.  And I
understand that there's a lot of interest in how the remaining 0.001% of
users will work without it.  But I'm having a hard time understanding why
we're even discussing them.  Indeed, Matthew's repeated calls for a
real-world use case of a network that doesn't already have, won't quickly
get, or couldn't easily be configured to have DNS has remained unanswered.
(And I'd add to that filter "outside of loud talking range.")
 
Furthermore, I'll grant you could build a DHT layer that serves as a
read-through cache of the central security mechanism that provides a faster
response and propagation time than DNS.  But Akamai, UltraDNS, and other
massive, global organizations have proven that DNS can propagate pretty dang
quick.

And while I think it's an excellent, open question on how to improve DNS to
make it even faster (perhaps with some kind of P2P layer), or how to make it
work better in partially or totally disconnected scenarios, or how to make
it more resistant to spamming/phishing and so forth, that seems like a much
bigger problem than SIP alone and thus better suited to its own group.

And finally, I agree there are open questions as to how my car-of-the-future
will talk to my garage door and sprinkler system.  But why is this relevant?
I share Adam's concern that we're dreaming up a panacea to every conceivable
problem rather than standardizing a solution to problems that actually exist
with technologies that are already at hand.


I apologize for being obtuse, but dynamic DNS (even with today's
limitations, and perhaps with future improvements) seems like such an
obvious way to solve the distributed name resolution problem that I'm
astonished it's being discounted so easily in favor of complex solutions to
nebulous problems experienced by nonexistent userbases.

I'll admit the conclusion is anticlimactic -- I'm a P2P developer who likes
high tech as much as anybody.  But where has my reasoning fallen off track?

-david

> -----Original Message-----
> From: p2p-sip-bounces at cs.columbia.edu [mailto:p2p-sip-
> bounces at cs.columbia.edu] On Behalf Of Matthew Kaufman
> Sent: Monday, November 27, 2006 10:18 AM
> To: hsinnrei at adobe.com; slavitch at gmail.com
> Cc: p2p-sip at cs.columbia.edu
> Subject: Re: [p2p-sip] Fwd: what's wrong with DNS?
> 
> Michael Slavitch:
> > As a form of assessing identity DNS is particularly weak and requires
> > considerable care and expert feeding as provisioning is always
> > ultimately human-based requiring some kind of authority.
> 
> Agreed. Though in small disconnected networks, zeroconf using mDNS avoids
> much of this, and in large networks the IT guys or your ISP manage the
> complexity for you (though they haven't solved the security problem yet)
> 
> > A self-organizing automated system needs no such crutches.
> 
> Except for bootstrapping to get the first two nodes talking. Which, if on
> the Internet, means a way to find IP addresses of those nodes. Which, on
> today's Internet, could use DNS. Or, as the Skype folks do, hardcoded
> bootstrap
> addresses. But either way, bootstrap is such a crutch.
> 
> > A key provision of P2PSIP is to eliminate the need for such
> > authorities in the first place because they are superfluous for the
> > application.  An extended social network needs no central domain
> > administrator nor a central certificate authority.   It needs at most
> > a human facilitator not an engineer.
> 
> So P2PSIP includes a web-of-trust security model? If so, why not leverage
> existing solutions for that as well (like PGP)?
> 
> > Such social networks are not trite. For example social networks are
> > the basis of all peer reviews for medical journals.  Shouting doesn't
> > work when 500 doctors are to be consulted, on several continents, with
> > varied access to the Internet, and over time.  The person running the
> > peer review is a journal editor, a person who uses end-user tools, not
> > a network engineer.  DNS isn't the selection criteria here, reputation
> > is.
> 
> But DNS is how the journal's mailing list manager finds the mail hosts
> for each of the 500 doctors. Delivery of the article and reply comments
> across several continents over time works today just fine without P2P
> or DHTs or reputation-based web-of-trust security models. See this mailing
> list as an existence proof.
> 
> Matthew Kaufman
> matthew at matthew.at
> 
> _______________________________________________
> p2p-sip mailing list
> p2p-sip at cs.columbia.edu
> https://lists.cs.columbia.edu/cucslists/listinfo/p2p-sip


From br at brianrosen.net  Tue Nov 28 09:28:19 2006
From: br at brianrosen.net (Brian Rosen)
Date: Tue, 28 Nov 2006 09:28:19 -0500
Subject: [p2p-sip] Fwd: what's wrong with DNS?
In-Reply-To: <015301c712b9$5daeb950$0200a8c0@Quinthar>
Message-ID: <083001c712f9$79b91d80$640fa8c0@cis.neustar.com>

The point of a "centralized" credentials service is to limit who can call
you to someone who has credentials you accept.

Is any userid in any domain in the DNS acceptable to you for a credential?
It's not to me.  I think one of the reasons our charter will be accepted is
that we have allowed for a plausible credential issuing system.  If we had
said that we don't plan to have one, credentials are not needed, and all you
need is any domain to give you user at domain identity, then a lot of tough
questions would have been asked.  By saying you need credentials, and
credentials can be issued by some kind of authority, we create a feasible
way to control SPIT.

Skype works pretty much the way we see p2psip working.  It has a centralized
credential service; that's the website at skype.com where you sign up to get
a SkypeId.  It also has a set of hardwired addresses of peers that lets you
bootstrap your way to the cloud.  The addresses are not those of centralized
directory servers.  They are just known to be available peers.  Any peer
will do, and they start with the well known ones.  Once you get going and
are attached to the cloud, you don't need those peers any more, but they are
your way to get started.  We'll need a bootstrap mechanism, and we haven't
had any serious discussion on what that would be.

Brian

> -----Original Message-----
> From: p2p-sip-bounces at cs.columbia.edu [mailto:p2p-sip-
> bounces at cs.columbia.edu] On Behalf Of David Barrett
> Sent: Tuesday, November 28, 2006 1:50 AM
> To: 'Matthew Kaufman'; hsinnrei at adobe.com; slavitch at gmail.com
> Cc: p2p-sip at cs.columbia.edu
> Subject: Re: [p2p-sip] Fwd: what's wrong with DNS?
> 
> I think Matthew highlights the Achilles heel of so-called "pure" P2P: it
> never is.  In the real world, there's always a central service behind the
> curtain:
> 
> - In Skype's case, it's a central directory at a hard-coded IP address.
> 
> - In P2P-SIP's case (if Frank and Brian are right) is some kind of
> centralized authorization/authentication/bootstrapping service.
> 
> So granted, there's the question of "how do you find this service?"  The
> obvious choice favored by every other application on the internet (and
> most
> private networks) is "DNS".
> 
> But ignoring that, there's an even more important question:
> 
> 	"Who controls this service?"
> 
> If we build some fancy DHT solution we imply "some unnamed benevolent
> dictator yet to be appointed" will rule us all.  We've sorta assumed this
> (or similarly vague explanations) for the entire course of the P2P-SIP
> discussion.  For a long while I was not only on board, I was promoting
> this
> "let's solve security later" approach.
> 
> But then Adam proposed a different owner, the IANA.  And the way to do
> that
> is through the existing, tried-and-true interface of DNS (specifically,
> dynamic DNS).  I think this is obvious and brilliant.
> 
> Building a nonpartial, internationally represented, neutral global third
> party isn't easy, and while the IANA hasn't wildly succeeded in this,
> they're closer than anybody else (and decades ahead of anything we'd
> invent
> while searching for our benevolent dictator).
> 
> 
> Now I'll grant DNS is limited to only, say, 99.999% of users.  And I
> understand that there's a lot of interest in how the remaining 0.001% of
> users will work without it.  But I'm having a hard time understanding why
> we're even discussing them.  Indeed, Matthew's repeated calls for a
> real-world use case of a network that doesn't already have, won't quickly
> get, or couldn't easily be configured to have DNS has remained unanswered.
> (And I'd add to that filter "outside of loud talking range.")
> 
> Furthermore, I'll grant you could build a DHT layer that serves as a
> read-through cache of the central security mechanism that provides a
> faster
> response and propagation time than DNS.  But Akamai, UltraDNS, and other
> massive, global organizations have proven that DNS can propagate pretty
> dang
> quick.
> 
> And while I think it's an excellent, open question on how to improve DNS
> to
> make it even faster (perhaps with some kind of P2P layer), or how to make
> it
> work better in partially or totally disconnected scenarios, or how to make
> it more resistant to spamming/phishing and so forth, that seems like a
> much
> bigger problem than SIP alone and thus better suited to its own group.
> 
> And finally, I agree there are open questions as to how my car-of-the-
> future
> will talk to my garage door and sprinkler system.  But why is this
> relevant?
> I share Adam's concern that we're dreaming up a panacea to every
> conceivable
> problem rather than standardizing a solution to problems that actually
> exist
> with technologies that are already at hand.
> 
> 
> I apologize for being obtuse, but dynamic DNS (even with today's
> limitations, and perhaps with future improvements) seems like such an
> obvious way to solve the distributed name resolution problem that I'm
> astonished it's being discounted so easily in favor of complex solutions
> to
> nebulous problems experienced by nonexistent userbases.
> 
> I'll admit the conclusion is anticlimactic -- I'm a P2P developer who
> likes
> high tech as much as anybody.  But where has my reasoning fallen off
> track?
> 
> -david
> 
> > -----Original Message-----
> > From: p2p-sip-bounces at cs.columbia.edu [mailto:p2p-sip-
> > bounces at cs.columbia.edu] On Behalf Of Matthew Kaufman
> > Sent: Monday, November 27, 2006 10:18 AM
> > To: hsinnrei at adobe.com; slavitch at gmail.com
> > Cc: p2p-sip at cs.columbia.edu
> > Subject: Re: [p2p-sip] Fwd: what's wrong with DNS?
> >
> > Michael Slavitch:
> > > As a form of assessing identity DNS is particularly weak and requires
> > > considerable care and expert feeding as provisioning is always
> > > ultimately human-based requiring some kind of authority.
> >
> > Agreed. Though in small disconnected networks, zeroconf using mDNS
> avoids
> > much of this, and in large networks the IT guys or your ISP manage the
> > complexity for you (though they haven't solved the security problem yet)
> >
> > > A self-organizing automated system needs no such crutches.
> >
> > Except for bootstrapping to get the first two nodes talking. Which, if
> on
> > the Internet, means a way to find IP addresses of those nodes. Which, on
> > today's Internet, could use DNS. Or, as the Skype folks do, hardcoded
> > bootstrap
> > addresses. But either way, bootstrap is such a crutch.
> >
> > > A key provision of P2PSIP is to eliminate the need for such
> > > authorities in the first place because they are superfluous for the
> > > application.  An extended social network needs no central domain
> > > administrator nor a central certificate authority.   It needs at most
> > > a human facilitator not an engineer.
> >
> > So P2PSIP includes a web-of-trust security model? If so, why not
> leverage
> > existing solutions for that as well (like PGP)?
> >
> > > Such social networks are not trite. For example social networks are
> > > the basis of all peer reviews for medical journals.  Shouting doesn't
> > > work when 500 doctors are to be consulted, on several continents, with
> > > varied access to the Internet, and over time.  The person running the
> > > peer review is a journal editor, a person who uses end-user tools, not
> > > a network engineer.  DNS isn't the selection criteria here, reputation
> > > is.
> >
> > But DNS is how the journal's mailing list manager finds the mail hosts
> > for each of the 500 doctors. Delivery of the article and reply comments
> > across several continents over time works today just fine without P2P
> > or DHTs or reputation-based web-of-trust security models. See this
> mailing
> > list as an existence proof.
> >
> > Matthew Kaufman
> > matthew at matthew.at
> >
> > _______________________________________________
> > p2p-sip mailing list
> > p2p-sip at cs.columbia.edu
> > https://lists.cs.columbia.edu/cucslists/listinfo/p2p-sip
> 
> _______________________________________________
> p2p-sip mailing list
> p2p-sip at cs.columbia.edu
> https://lists.cs.columbia.edu/cucslists/listinfo/p2p-sip


From eunsooshim at gmail.com  Tue Nov 28 09:30:17 2006
From: eunsooshim at gmail.com (Eunsoo Shim)
Date: Tue, 28 Nov 2006 09:30:17 -0500
Subject: [p2p-sip] Fwd: what's wrong with DNS?
In-Reply-To: <015301c712b9$5daeb950$0200a8c0@Quinthar>
References: <200611271818.kARII0191955@where.matthew.at>
	<015301c712b9$5daeb950$0200a8c0@Quinthar>
Message-ID: <7b683f3f0611280630t369ea8aesc2ceb1a1efea110c@mail.gmail.com>

I am not sure what your proposal is.
Could you provide more specifics on your proposal?
For example, are you suggesting that we use existing DNS servers to store
whatever we want to store at the P2P(SIP) network?
Or are you suggesting that we use the DNS PROTOCOL as (part of) the protocol
for P2P SIP and build a separate (global or ad-hoc) DNS-like infrastructure?
Thanks.

Eunsoo

On 11/28/06, David Barrett <dbarrett at quinthar.com> wrote:
>
> I think Matthew highlights the Achilles heel of so-called "pure" P2P: it
> never is.  In the real world, there's always a central service behind the
> curtain:
>
> - In Skype's case, it's a central directory at a hard-coded IP address.
>
> - In P2P-SIP's case (if Frank and Brian are right) is some kind of
> centralized authorization/authentication/bootstrapping service.
>
> So granted, there's the question of "how do you find this service?"  The
> obvious choice favored by every other application on the internet (and
> most
> private networks) is "DNS".
>
> But ignoring that, there's an even more important question:
>
>         "Who controls this service?"
>
> If we build some fancy DHT solution we imply "some unnamed benevolent
> dictator yet to be appointed" will rule us all.  We've sorta assumed this
> (or similarly vague explanations) for the entire course of the P2P-SIP
> discussion.  For a long while I was not only on board, I was promoting
> this
> "let's solve security later" approach.
>
> But then Adam proposed a different owner, the IANA.  And the way to do
> that
> is through the existing, tried-and-true interface of DNS (specifically,
> dynamic DNS).  I think this is obvious and brilliant.
>
> Building a nonpartial, internationally represented, neutral global third
> party isn't easy, and while the IANA hasn't wildly succeeded in this,
> they're closer than anybody else (and decades ahead of anything we'd
> invent
> while searching for our benevolent dictator).
>
>
> Now I'll grant DNS is limited to only, say, 99.999% of users.  And I
> understand that there's a lot of interest in how the remaining 0.001% of
> users will work without it.  But I'm having a hard time understanding why
> we're even discussing them.  Indeed, Matthew's repeated calls for a
> real-world use case of a network that doesn't already have, won't quickly
> get, or couldn't easily be configured to have DNS has remained unanswered.
> (And I'd add to that filter "outside of loud talking range.")
>
> Furthermore, I'll grant you could build a DHT layer that serves as a
> read-through cache of the central security mechanism that provides a
> faster
> response and propagation time than DNS.  But Akamai, UltraDNS, and other
> massive, global organizations have proven that DNS can propagate pretty
> dang
> quick.
>
> And while I think it's an excellent, open question on how to improve DNS
> to
> make it even faster (perhaps with some kind of P2P layer), or how to make
> it
> work better in partially or totally disconnected scenarios, or how to make
> it more resistant to spamming/phishing and so forth, that seems like a
> much
> bigger problem than SIP alone and thus better suited to its own group.
>
> And finally, I agree there are open questions as to how my
> car-of-the-future
> will talk to my garage door and sprinkler system.  But why is this
> relevant?
> I share Adam's concern that we're dreaming up a panacea to every
> conceivable
> problem rather than standardizing a solution to problems that actually
> exist
> with technologies that are already at hand.
>
>
> I apologize for being obtuse, but dynamic DNS (even with today's
> limitations, and perhaps with future improvements) seems like such an
> obvious way to solve the distributed name resolution problem that I'm
> astonished it's being discounted so easily in favor of complex solutions
> to
> nebulous problems experienced by nonexistent userbases.
>
> I'll admit the conclusion is anticlimactic -- I'm a P2P developer who
> likes
> high tech as much as anybody.  But where has my reasoning fallen off
> track?
>
> -david
>
> > -----Original Message-----
> > From: p2p-sip-bounces at cs.columbia.edu [mailto:p2p-sip-
> > bounces at cs.columbia.edu] On Behalf Of Matthew Kaufman
> > Sent: Monday, November 27, 2006 10:18 AM
> > To: hsinnrei at adobe.com; slavitch at gmail.com
> > Cc: p2p-sip at cs.columbia.edu
> > Subject: Re: [p2p-sip] Fwd: what's wrong with DNS?
> >
> > Michael Slavitch:
> > > As a form of assessing identity DNS is particularly weak and requires
> > > considerable care and expert feeding as provisioning is always
> > > ultimately human-based requiring some kind of authority.
> >
> > Agreed. Though in small disconnected networks, zeroconf using mDNS
> avoids
> > much of this, and in large networks the IT guys or your ISP manage the
> > complexity for you (though they haven't solved the security problem yet)
> >
> > > A self-organizing automated system needs no such crutches.
> >
> > Except for bootstrapping to get the first two nodes talking. Which, if
> on
> > the Internet, means a way to find IP addresses of those nodes. Which, on
> > today's Internet, could use DNS. Or, as the Skype folks do, hardcoded
> > bootstrap
> > addresses. But either way, bootstrap is such a crutch.
> >
> > > A key provision of P2PSIP is to eliminate the need for such
> > > authorities in the first place because they are superfluous for the
> > > application.  An extended social network needs no central domain
> > > administrator nor a central certificate authority.   It needs at most
> > > a human facilitator not an engineer.
> >
> > So P2PSIP includes a web-of-trust security model? If so, why not
> leverage
> > existing solutions for that as well (like PGP)?
> >
> > > Such social networks are not trite. For example social networks are
> > > the basis of all peer reviews for medical journals.  Shouting doesn't
> > > work when 500 doctors are to be consulted, on several continents, with
> > > varied access to the Internet, and over time.  The person running the
> > > peer review is a journal editor, a person who uses end-user tools, not
> > > a network engineer.  DNS isn't the selection criteria here, reputation
> > > is.
> >
> > But DNS is how the journal's mailing list manager finds the mail hosts
> > for each of the 500 doctors. Delivery of the article and reply comments
> > across several continents over time works today just fine without P2P
> > or DHTs or reputation-based web-of-trust security models. See this
> mailing
> > list as an existence proof.
> >
> > Matthew Kaufman
> > matthew at matthew.at
> >
> > _______________________________________________
> > p2p-sip mailing list
> > p2p-sip at cs.columbia.edu
> > https://lists.cs.columbia.edu/cucslists/listinfo/p2p-sip
>
> _______________________________________________
> p2p-sip mailing list
> p2p-sip at cs.columbia.edu
> https://lists.cs.columbia.edu/cucslists/listinfo/p2p-sip
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: https://lists.cs.columbia.edu/pipermail/p2p-sip/attachments/20061128/eae1d548/attachment-0001.html

From fwmiller at cornfed.com  Tue Nov 28 10:15:51 2006
From: fwmiller at cornfed.com (Frank W. Miller)
Date: Tue, 28 Nov 2006 08:15:51 -0700
Subject: [p2p-sip] Fwd: what's wrong with DNS?
In-Reply-To: <7b683f3f0611280630t369ea8aesc2ceb1a1efea110c@mail.gmail.com>
References: <200611271818.kARII0191955@where.matthew.at>
	<015301c712b9$5daeb950$0200a8c0@Quinthar>
	<7b683f3f0611280630t369ea8aesc2ceb1a1efea110c@mail.gmail.com>
Message-ID: <1164726951.2622.19.camel@localhost.localdomain>



I would propose something like this:

There are two elements, User Agents and the Identity Service. Identity
refers to messages that are sent between an Internet host and the
Identity Service to authorize and authenticate UAIs.  The Identity
Service is assumed to be authoritative and trusted, and very likely a
centralized service.  Identity operations include the following
functions:

   1.  Authorization

   2.  Authentication

Authorization

I call the user portion of a SIP URI the User Agent Instance (UAI).
The UAI is allocated to the UA by some means outside of the scope of this
protocol, i.e. by visiting a website or something.  Authorization refers
to the registration of a UA instance for a UAI with an Identity Service.
UAs are generally authorized when a user wants to begin interacting with
other UAs, e.g. when their UA client program is started.  (Note that these
messages are notional.  They don't have to even be SIP.  The idea of
these templates is to list the information moving between the elements)


   Frank            Identity Service
     |      AUTHORIZE      |
     |-------------------->|
     |       200 OK        |
     |<--------------------|
     |                     |


   AUTHORIZE
   UAI: Frank
   Password: ... Frank's password ...
   CSeq: 2 AUTHORIZE

Frank's UA requests authorization of the UAI.  The request includes a
shared secret that was setup when the UAI was issued.  The shared secret must
be sent encrypted.


   200 OK
   UAI: Frank
   Authorization: ... Cookie sent to anyone authenticating Frank ...
   CSeq: 2 AUTHORIZE

Frank's UA is returned a Cookie.  This cookie will be sent to other UAs
for subsequent interactions.  Other UAs can use the cookie for authenticating
Frank's UA.  Note that this assumes the Authorization cookie is encrypted.


Authentication

Authentication refers to the process of querying the Identity Service
regarding a UAI.  The procedure described herein is usually attempted
when a UA receives its first request for interaction from another UA.


    Bob            Identity Service
     |     AUTHENTICATE    |
     |-------------------->|
     |       200 OK        |
     |<--------------------|
     |                     |



   AUTHENTICATE
   UAI: Frank
   Password: ... Bob's password ...
   CSeq: 2 AUTHORIZE

Bob send's an authentication request to the Identity Service.  Bob is
only allowed to authenticate Frank if his UA is authorized using the
shared secret for Bob's UAI.  The shared secret must be encrypted.


   MSP 200 OK
   UAI: Frank
   Authentication: ... Cookie sent to Frank during authorization ...
   CSeq: 2 AUTHORIZE

The identity service returns the cookie that was sent by Frank with
the initial interaction.  Bob's UAI can compare the two and determine
if Frank is really Frank.  Again, the cookie must be encrypted.

This proposal requires some level of encryption in all these
messages.  I am actually in favor of SSL/TLS being required for
transport of these messages, whether that is for this separate
protocol blurb or for SIP in general.

The management of UAIs for each p2p cloud becomes the authoritative
information for that cloud.  I can imagine services like FWD providing
it for free and I can also imagine commercial services charging for
the service.

FM



On Tue, 2006-11-28 at 09:30 -0500, Eunsoo Shim wrote:
> I am not sure what your proposal is.
> Could you provide more specifics on your proposal?
> For example, are you suggesting that we use existing DNS servers to
> store whatever we want to store at the P2P(SIP) network?
> Or are you suggesting that we use the DNS PROTOCOL as (part of) the
> protocol for P2P SIP and build a separate (global or ad-hoc) DNS-like
> infrastructure?
> Thanks.
> 
> Eunsoo
> 
> On 11/28/06, David Barrett <dbarrett at quinthar.com> wrote:
>         I think Matthew highlights the Achilles heel of so-called
>         "pure" P2P: it
>         never is.  In the real world, there's always a central service
>         behind the
>         curtain:
>         
>         - In Skype's case, it's a central directory at a hard-coded IP
>         address. 
>         
>         - In P2P-SIP's case (if Frank and Brian are right) is some
>         kind of
>         centralized authorization/authentication/bootstrapping
>         service.
>         
>         So granted, there's the question of "how do you find this
>         service?"  The 
>         obvious choice favored by every other application on the
>         internet (and most
>         private networks) is "DNS".
>         
>         But ignoring that, there's an even more important question:
>         
>                 "Who controls this service?" 
>         
>         If we build some fancy DHT solution we imply "some unnamed
>         benevolent
>         dictator yet to be appointed" will rule us all.  We've sorta
>         assumed this
>         (or similarly vague explanations) for the entire course of the
>         P2P-SIP 
>         discussion.  For a long while I was not only on board, I was
>         promoting this
>         "let's solve security later" approach.
>         
>         But then Adam proposed a different owner, the IANA.  And the
>         way to do that
>         is through the existing, tried-and-true interface of DNS
>         (specifically, 
>         dynamic DNS).  I think this is obvious and brilliant.
>         
>         Building a nonpartial, internationally represented, neutral
>         global third
>         party isn't easy, and while the IANA hasn't wildly succeeded
>         in this,
>         they're closer than anybody else (and decades ahead of
>         anything we'd invent 
>         while searching for our benevolent dictator).
>         
>         
>         Now I'll grant DNS is limited to only, say, 99.999% of
>         users.  And I
>         understand that there's a lot of interest in how the remaining
>         0.001% of
>         users will work without it.  But I'm having a hard time
>         understanding why 
>         we're even discussing them.  Indeed, Matthew's repeated calls
>         for a
>         real-world use case of a network that doesn't already have,
>         won't quickly
>         get, or couldn't easily be configured to have DNS has remained
>         unanswered. 
>         (And I'd add to that filter "outside of loud talking range.")
>         
>         Furthermore, I'll grant you could build a DHT layer that
>         serves as a
>         read-through cache of the central security mechanism that
>         provides a faster 
>         response and propagation time than DNS.  But Akamai, UltraDNS,
>         and other
>         massive, global organizations have proven that DNS can
>         propagate pretty dang
>         quick.
>         
>         And while I think it's an excellent, open question on how to
>         improve DNS to 
>         make it even faster (perhaps with some kind of P2P layer), or
>         how to make it
>         work better in partially or totally disconnected scenarios, or
>         how to make
>         it more resistant to spamming/phishing and so forth, that
>         seems like a much 
>         bigger problem than SIP alone and thus better suited to its
>         own group.
>         
>         And finally, I agree there are open questions as to how my
>         car-of-the-future
>         will talk to my garage door and sprinkler system.  But why is
>         this relevant? 
>         I share Adam's concern that we're dreaming up a panacea to
>         every conceivable
>         problem rather than standardizing a solution to problems that
>         actually exist
>         with technologies that are already at hand.
>         
>         
>         I apologize for being obtuse, but dynamic DNS (even with
>         today's 
>         limitations, and perhaps with future improvements) seems like
>         such an
>         obvious way to solve the distributed name resolution problem
>         that I'm
>         astonished it's being discounted so easily in favor of complex
>         solutions to 
>         nebulous problems experienced by nonexistent userbases.
>         
>         I'll admit the conclusion is anticlimactic -- I'm a P2P
>         developer who likes
>         high tech as much as anybody.  But where has my reasoning
>         fallen off track? 
>         
>         -david
>         
>         > -----Original Message-----
>         > From: p2p-sip-bounces at cs.columbia.edu [mailto:p2p-sip-
>         > bounces at cs.columbia.edu] On Behalf Of Matthew Kaufman
>         > Sent: Monday, November 27, 2006 10:18 AM
>         > To: hsinnrei at adobe.com; slavitch at gmail.com
>         > Cc: p2p-sip at cs.columbia.edu
>         > Subject: Re: [p2p-sip] Fwd: what's wrong with DNS?
>         >
>         > Michael Slavitch:
>         > > As a form of assessing identity DNS is particularly weak
>         and requires 
>         > > considerable care and expert feeding as provisioning is
>         always
>         > > ultimately human-based requiring some kind of authority.
>         >
>         > Agreed. Though in small disconnected networks, zeroconf
>         using mDNS avoids 
>         > much of this, and in large networks the IT guys or your ISP
>         manage the
>         > complexity for you (though they haven't solved the security
>         problem yet)
>         >
>         > > A self-organizing automated system needs no such
>         crutches. 
>         >
>         > Except for bootstrapping to get the first two nodes talking.
>         Which, if on
>         > the Internet, means a way to find IP addresses of those
>         nodes. Which, on
>         > today's Internet, could use DNS. Or, as the Skype folks do,
>         hardcoded 
>         > bootstrap
>         > addresses. But either way, bootstrap is such a crutch.
>         >
>         > > A key provision of P2PSIP is to eliminate the need for
>         such
>         > > authorities in the first place because they are
>         superfluous for the 
>         > > application.  An extended social network needs no central
>         domain
>         > > administrator nor a central certificate authority.   It
>         needs at most
>         > > a human facilitator not an engineer.
>         >
>         > So P2PSIP includes a web-of-trust security model? If so, why
>         not leverage
>         > existing solutions for that as well (like PGP)?
>         >
>         > > Such social networks are not trite. For example social
>         networks are 
>         > > the basis of all peer reviews for medical
>         journals.  Shouting doesn't
>         > > work when 500 doctors are to be consulted, on several
>         continents, with
>         > > varied access to the Internet, and over time.  The person
>         running the 
>         > > peer review is a journal editor, a person who uses
>         end-user tools, not
>         > > a network engineer.  DNS isn't the selection criteria
>         here, reputation
>         > > is.
>         >
>         > But DNS is how the journal's mailing list manager finds the
>         mail hosts 
>         > for each of the 500 doctors. Delivery of the article and
>         reply comments
>         > across several continents over time works today just fine
>         without P2P
>         > or DHTs or reputation-based web-of-trust security models.
>         See this mailing 
>         > list as an existence proof.
>         >
>         > Matthew Kaufman
>         > matthew at matthew.at
>         >
>         > _______________________________________________
>         > p2p-sip mailing list 
>         > p2p-sip at cs.columbia.edu
>         > https://lists.cs.columbia.edu/cucslists/listinfo/p2p-sip
>         
>         _______________________________________________
>         p2p-sip mailing list
>         p2p-sip at cs.columbia.edu
>         https://lists.cs.columbia.edu/cucslists/listinfo/p2p-sip
> 
> _______________________________________________
> p2p-sip mailing list
> p2p-sip at cs.columbia.edu
> https://lists.cs.columbia.edu/cucslists/listinfo/p2p-sip


From br at brianrosen.net  Tue Nov 28 10:56:12 2006
From: br at brianrosen.net (Brian Rosen)
Date: Tue, 28 Nov 2006 10:56:12 -0500
Subject: [p2p-sip] Fwd: what's wrong with DNS?
In-Reply-To: <1164726951.2622.19.camel@localhost.localdomain>
Message-ID: <085001c71305$be4f70f0$640fa8c0@cis.neustar.com>

Wow.

You understand that you are proposing a centralized service that
authenticates you every time you boot, and a centralized authorization
service that authorizes you AT CALL TIME, at least the first time per boot?

The direction we are going is that you use the centralized credentialing
service to get credentials effectively once in a lifetime of the credential,
which might last at least months, and maybe years.  Thereafter, the
distributed hash table takes care of authentication and authorization peer
to peer without any central service.

I would definitely not call what you are proposing a peer to peer service;
it relies on centralized authentication and centralized authorization
services in real time.

Brian

> -----Original Message-----
> From: p2p-sip-bounces at cs.columbia.edu [mailto:p2p-sip-
> bounces at cs.columbia.edu] On Behalf Of Frank W. Miller
> Sent: Tuesday, November 28, 2006 10:16 AM
> To: Eunsoo Shim
> Cc: p2p-sip at cs.columbia.edu
> Subject: Re: [p2p-sip] Fwd: what's wrong with DNS?
> 
> 
> 
> I would propose something like this:
> 
> There are two elements, User Agents and the Identity Service. Identity
> refers to messages that are sent between an Internet host and the
> Identity Service to authorize and authenticate UAIs.  The Identity
> Service is assumed to be authoritative and trusted, and very likely a
> centralized service.  Identity operations include the following
> functions:
> 
>    1.  Authorization
> 
>    2.  Authentication
> 
> Authorization
> 
> I call the user portion of a SIP URI the User Agent Instance (UAI).
> The UAI is allocated to the UA by some means outside of the scope of this
> protocol, i.e. by visiting a website or something.  Authorization refers
> to the registration of a UA instance for a UAI with an Identity Service.
> UAs are generally authorized when a user wants to begin interacting with
> other UAs, e.g. when their UA client program is started.  (Note that these
> messages are notional.  They don't have to even be SIP.  The idea of
> these templates is to list the information moving between the elements)
> 
> 
>    Frank            Identity Service
>      |      AUTHORIZE      |
>      |-------------------->|
>      |       200 OK        |
>      |<--------------------|
>      |                     |
> 
> 
>    AUTHORIZE
>    UAI: Frank
>    Password: ... Frank's password ...
>    CSeq: 2 AUTHORIZE
> 
> Frank's UA requests authorization of the UAI.  The request includes a
> shared secret that was setup when the UAI was issued.  The shared secret
> must
> be sent encrypted.
> 
> 
>    200 OK
>    UAI: Frank
>    Authorization: ... Cookie sent to anyone authenticating Frank ...
>    CSeq: 2 AUTHORIZE
> 
> Frank's UA is returned a Cookie.  This cookie will be sent to other UAs
> for subsequent interactions.  Other UAs can use the cookie for
> authenticating
> Frank's UA.  Note that this assumes the Authorization cookie is encrypted.
> 
> 
> Authentication
> 
> Authentication refers to the process of querying the Identity Service
> regarding a UAI.  The procedure described herein is usually attempted
> when a UA receives its first request for interaction from another UA.
> 
> 
>     Bob            Identity Service
>      |     AUTHENTICATE    |
>      |-------------------->|
>      |       200 OK        |
>      |<--------------------|
>      |                     |
> 
> 
> 
>    AUTHENTICATE
>    UAI: Frank
>    Password: ... Bob's password ...
>    CSeq: 2 AUTHORIZE
> 
> Bob send's an authentication request to the Identity Service.  Bob is
> only allowed to authenticate Frank if his UA is authorized using the
> shared secret for Bob's UAI.  The shared secret must be encrypted.
> 
> 
>    MSP 200 OK
>    UAI: Frank
>    Authentication: ... Cookie sent to Frank during authorization ...
>    CSeq: 2 AUTHORIZE
> 
> The identity service returns the cookie that was sent by Frank with
> the initial interaction.  Bob's UAI can compare the two and determine
> if Frank is really Frank.  Again, the cookie must be encrypted.
> 
> This proposal requires some level of encryption in all these
> messages.  I am actually in favor of SSL/TLS being required for
> transport of these messages, whether that is for this separate
> protocol blurb or for SIP in general.
> 
> The management of UAIs for each p2p cloud becomes the authoritative
> information for that cloud.  I can imagine services like FWD providing
> it for free and I can also imagine commercial services charging for
> the service.
> 
> FM
> 
> 
> 
> On Tue, 2006-11-28 at 09:30 -0500, Eunsoo Shim wrote:
> > I am not sure what your proposal is.
> > Could you provide more specifics on your proposal?
> > For example, are you suggesting that we use existing DNS servers to
> > store whatever we want to store at the P2P(SIP) network?
> > Or are you suggesting that we use the DNS PROTOCOL as (part of) the
> > protocol for P2P SIP and build a separate (global or ad-hoc) DNS-like
> > infrastructure?
> > Thanks.
> >
> > Eunsoo
> >
> > On 11/28/06, David Barrett <dbarrett at quinthar.com> wrote:
> >         I think Matthew highlights the Achilles heel of so-called
> >         "pure" P2P: it
> >         never is.  In the real world, there's always a central service
> >         behind the
> >         curtain:
> >
> >         - In Skype's case, it's a central directory at a hard-coded IP
> >         address.
> >
> >         - In P2P-SIP's case (if Frank and Brian are right) is some
> >         kind of
> >         centralized authorization/authentication/bootstrapping
> >         service.
> >
> >         So granted, there's the question of "how do you find this
> >         service?"  The
> >         obvious choice favored by every other application on the
> >         internet (and most
> >         private networks) is "DNS".
> >
> >         But ignoring that, there's an even more important question:
> >
> >                 "Who controls this service?"
> >
> >         If we build some fancy DHT solution we imply "some unnamed
> >         benevolent
> >         dictator yet to be appointed" will rule us all.  We've sorta
> >         assumed this
> >         (or similarly vague explanations) for the entire course of the
> >         P2P-SIP
> >         discussion.  For a long while I was not only on board, I was
> >         promoting this
> >         "let's solve security later" approach.
> >
> >         But then Adam proposed a different owner, the IANA.  And the
> >         way to do that
> >         is through the existing, tried-and-true interface of DNS
> >         (specifically,
> >         dynamic DNS).  I think this is obvious and brilliant.
> >
> >         Building a nonpartial, internationally represented, neutral
> >         global third
> >         party isn't easy, and while the IANA hasn't wildly succeeded
> >         in this,
> >         they're closer than anybody else (and decades ahead of
> >         anything we'd invent
> >         while searching for our benevolent dictator).
> >
> >
> >         Now I'll grant DNS is limited to only, say, 99.999% of
> >         users.  And I
> >         understand that there's a lot of interest in how the remaining
> >         0.001% of
> >         users will work without it.  But I'm having a hard time
> >         understanding why
> >         we're even discussing them.  Indeed, Matthew's repeated calls
> >         for a
> >         real-world use case of a network that doesn't already have,
> >         won't quickly
> >         get, or couldn't easily be configured to have DNS has remained
> >         unanswered.
> >         (And I'd add to that filter "outside of loud talking range.")
> >
> >         Furthermore, I'll grant you could build a DHT layer that
> >         serves as a
> >         read-through cache of the central security mechanism that
> >         provides a faster
> >         response and propagation time than DNS.  But Akamai, UltraDNS,
> >         and other
> >         massive, global organizations have proven that DNS can
> >         propagate pretty dang
> >         quick.
> >
> >         And while I think it's an excellent, open question on how to
> >         improve DNS to
> >         make it even faster (perhaps with some kind of P2P layer), or
> >         how to make it
> >         work better in partially or totally disconnected scenarios, or
> >         how to make
> >         it more resistant to spamming/phishing and so forth, that
> >         seems like a much
> >         bigger problem than SIP alone and thus better suited to its
> >         own group.
> >
> >         And finally, I agree there are open questions as to how my
> >         car-of-the-future
> >         will talk to my garage door and sprinkler system.  But why is
> >         this relevant?
> >         I share Adam's concern that we're dreaming up a panacea to
> >         every conceivable
> >         problem rather than standardizing a solution to problems that
> >         actually exist
> >         with technologies that are already at hand.
> >
> >
> >         I apologize for being obtuse, but dynamic DNS (even with
> >         today's
> >         limitations, and perhaps with future improvements) seems like
> >         such an
> >         obvious way to solve the distributed name resolution problem
> >         that I'm
> >         astonished it's being discounted so easily in favor of complex
> >         solutions to
> >         nebulous problems experienced by nonexistent userbases.
> >
> >         I'll admit the conclusion is anticlimactic -- I'm a P2P
> >         developer who likes
> >         high tech as much as anybody.  But where has my reasoning
> >         fallen off track?
> >
> >         -david
> >
> >         > -----Original Message-----
> >         > From: p2p-sip-bounces at cs.columbia.edu [mailto:p2p-sip-
> >         > bounces at cs.columbia.edu] On Behalf Of Matthew Kaufman
> >         > Sent: Monday, November 27, 2006 10:18 AM
> >         > To: hsinnrei at adobe.com; slavitch at gmail.com
> >         > Cc: p2p-sip at cs.columbia.edu
> >         > Subject: Re: [p2p-sip] Fwd: what's wrong with DNS?
> >         >
> >         > Michael Slavitch:
> >         > > As a form of assessing identity DNS is particularly weak
> >         and requires
> >         > > considerable care and expert feeding as provisioning is
> >         always
> >         > > ultimately human-based requiring some kind of authority.
> >         >
> >         > Agreed. Though in small disconnected networks, zeroconf
> >         using mDNS avoids
> >         > much of this, and in large networks the IT guys or your ISP
> >         manage the
> >         > complexity for you (though they haven't solved the security
> >         problem yet)
> >         >
> >         > > A self-organizing automated system needs no such
> >         crutches.
> >         >
> >         > Except for bootstrapping to get the first two nodes talking.
> >         Which, if on
> >         > the Internet, means a way to find IP addresses of those
> >         nodes. Which, on
> >         > today's Internet, could use DNS. Or, as the Skype folks do,
> >         hardcoded
> >         > bootstrap
> >         > addresses. But either way, bootstrap is such a crutch.
> >         >
> >         > > A key provision of P2PSIP is to eliminate the need for
> >         such
> >         > > authorities in the first place because they are
> >         superfluous for the
> >         > > application.  An extended social network needs no central
> >         domain
> >         > > administrator nor a central certificate authority.   It
> >         needs at most
> >         > > a human facilitator not an engineer.
> >         >
> >         > So P2PSIP includes a web-of-trust security model? If so, why
> >         not leverage
> >         > existing solutions for that as well (like PGP)?
> >         >
> >         > > Such social networks are not trite. For example social
> >         networks are
> >         > > the basis of all peer reviews for medical
> >         journals.  Shouting doesn't
> >         > > work when 500 doctors are to be consulted, on several
> >         continents, with
> >         > > varied access to the Internet, and over time.  The person
> >         running the
> >         > > peer review is a journal editor, a person who uses
> >         end-user tools, not
> >         > > a network engineer.  DNS isn't the selection criteria
> >         here, reputation
> >         > > is.
> >         >
> >         > But DNS is how the journal's mailing list manager finds the
> >         mail hosts
> >         > for each of the 500 doctors. Delivery of the article and
> >         reply comments
> >         > across several continents over time works today just fine
> >         without P2P
> >         > or DHTs or reputation-based web-of-trust security models.
> >         See this mailing
> >         > list as an existence proof.
> >         >
> >         > Matthew Kaufman
> >         > matthew at matthew.at
> >         >
> >         > _______________________________________________
> >         > p2p-sip mailing list
> >         > p2p-sip at cs.columbia.edu
> >         > https://lists.cs.columbia.edu/cucslists/listinfo/p2p-sip
> >
> >         _______________________________________________
> >         p2p-sip mailing list
> >         p2p-sip at cs.columbia.edu
> >         https://lists.cs.columbia.edu/cucslists/listinfo/p2p-sip
> >
> > _______________________________________________
> > p2p-sip mailing list
> > p2p-sip at cs.columbia.edu
> > https://lists.cs.columbia.edu/cucslists/listinfo/p2p-sip
> 
> _______________________________________________
> p2p-sip mailing list
> p2p-sip at cs.columbia.edu
> https://lists.cs.columbia.edu/cucslists/listinfo/p2p-sip


From slavitch at gmail.com  Tue Nov 28 10:58:41 2006
From: slavitch at gmail.com (Michael Slavitch)
Date: Tue, 28 Nov 2006 10:58:41 -0500
Subject: [p2p-sip] Fwd: what's wrong with DNS?
In-Reply-To: <085001c71305$be4f70f0$640fa8c0@cis.neustar.com>
References: <1164726951.2622.19.camel@localhost.localdomain>
	<085001c71305$be4f70f0$640fa8c0@cis.neustar.com>
Message-ID: <1fc32c8f0611280758v1c2e660bvc0c97029add640a5@mail.gmail.com>

"The direction we are going is that you use the centralized credentialing
service to get credentials effectively once in a lifetime of the credential."

Indeed. That is the whole point of a the DHT store.


On 11/28/06, Brian Rosen <br at brianrosen.net> wrote:
> Wow.
>
> You understand that you are proposing a centralized service that
> authenticates you every time you boot, and a centralized authorization
> service that authorizes you AT CALL TIME, at least the first time per boot?
>
> The direction we are going is that you use the centralized credentialing
> service to get credentials effectively once in a lifetime of the credential,
> which might last at least months, and maybe years.  Thereafter, the
> distributed hash table takes care of authentication and authorization peer
> to peer without any central service.
>
> I would definitely not call what you are proposing a peer to peer service;
> it relies on centralized authentication and centralized authorization
> services in real time.
>
> Brian
>
> > -----Original Message-----
> > From: p2p-sip-bounces at cs.columbia.edu [mailto:p2p-sip-
> > bounces at cs.columbia.edu] On Behalf Of Frank W. Miller
> > Sent: Tuesday, November 28, 2006 10:16 AM
> > To: Eunsoo Shim
> > Cc: p2p-sip at cs.columbia.edu
> > Subject: Re: [p2p-sip] Fwd: what's wrong with DNS?
> >
> >
> >
> > I would propose something like this:
> >
> > There are two elements, User Agents and the Identity Service. Identity
> > refers to messages that are sent between an Internet host and the
> > Identity Service to authorize and authenticate UAIs.  The Identity
> > Service is assumed to be authoritative and trusted, and very likely a
> > centralized service.  Identity operations include the following
> > functions:
> >
> >    1.  Authorization
> >
> >    2.  Authentication
> >
> > Authorization
> >
> > I call the user portion of a SIP URI the User Agent Instance (UAI).
> > The UAI is allocated to the UA by some means outside of the scope of this
> > protocol, i.e. by visiting a website or something.  Authorization refers
> > to the registration of a UA instance for a UAI with an Identity Service.
> > UAs are generally authorized when a user wants to begin interacting with
> > other UAs, e.g. when their UA client program is started.  (Note that these
> > messages are notional.  They don't have to even be SIP.  The idea of
> > these templates is to list the information moving between the elements)
> >
> >
> >    Frank            Identity Service
> >      |      AUTHORIZE      |
> >      |-------------------->|
> >      |       200 OK        |
> >      |<--------------------|
> >      |                     |
> >
> >
> >    AUTHORIZE
> >    UAI: Frank
> >    Password: ... Frank's password ...
> >    CSeq: 2 AUTHORIZE
> >
> > Frank's UA requests authorization of the UAI.  The request includes a
> > shared secret that was setup when the UAI was issued.  The shared secret
> > must
> > be sent encrypted.
> >
> >
> >    200 OK
> >    UAI: Frank
> >    Authorization: ... Cookie sent to anyone authenticating Frank ...
> >    CSeq: 2 AUTHORIZE
> >
> > Frank's UA is returned a Cookie.  This cookie will be sent to other UAs
> > for subsequent interactions.  Other UAs can use the cookie for
> > authenticating
> > Frank's UA.  Note that this assumes the Authorization cookie is encrypted.
> >
> >
> > Authentication
> >
> > Authentication refers to the process of querying the Identity Service
> > regarding a UAI.  The procedure described herein is usually attempted
> > when a UA receives its first request for interaction from another UA.
> >
> >
> >     Bob            Identity Service
> >      |     AUTHENTICATE    |
> >      |-------------------->|
> >      |       200 OK        |
> >      |<--------------------|
> >      |                     |
> >
> >
> >
> >    AUTHENTICATE
> >    UAI: Frank
> >    Password: ... Bob's password ...
> >    CSeq: 2 AUTHORIZE
> >
> > Bob send's an authentication request to the Identity Service.  Bob is
> > only allowed to authenticate Frank if his UA is authorized using the
> > shared secret for Bob's UAI.  The shared secret must be encrypted.
> >
> >
> >    MSP 200 OK
> >    UAI: Frank
> >    Authentication: ... Cookie sent to Frank during authorization ...
> >    CSeq: 2 AUTHORIZE
> >
> > The identity service returns the cookie that was sent by Frank with
> > the initial interaction.  Bob's UAI can compare the two and determine
> > if Frank is really Frank.  Again, the cookie must be encrypted.
> >
> > This proposal requires some level of encryption in all these
> > messages.  I am actually in favor of SSL/TLS being required for
> > transport of these messages, whether that is for this separate
> > protocol blurb or for SIP in general.
> >
> > The management of UAIs for each p2p cloud becomes the authoritative
> > information for that cloud.  I can imagine services like FWD providing
> > it for free and I can also imagine commercial services charging for
> > the service.
> >
> > FM
> >
> >
> >
> > On Tue, 2006-11-28 at 09:30 -0500, Eunsoo Shim wrote:
> > > I am not sure what your proposal is.
> > > Could you provide more specifics on your proposal?
> > > For example, are you suggesting that we use existing DNS servers to
> > > store whatever we want to store at the P2P(SIP) network?
> > > Or are you suggesting that we use the DNS PROTOCOL as (part of) the
> > > protocol for P2P SIP and build a separate (global or ad-hoc) DNS-like
> > > infrastructure?
> > > Thanks.
> > >
> > > Eunsoo
> > >
> > > On 11/28/06, David Barrett <dbarrett at quinthar.com> wrote:
> > >         I think Matthew highlights the Achilles heel of so-called
> > >         "pure" P2P: it
> > >         never is.  In the real world, there's always a central service
> > >         behind the
> > >         curtain:
> > >
> > >         - In Skype's case, it's a central directory at a hard-coded IP
> > >         address.
> > >
> > >         - In P2P-SIP's case (if Frank and Brian are right) is some
> > >         kind of
> > >         centralized authorization/authentication/bootstrapping
> > >         service.
> > >
> > >         So granted, there's the question of "how do you find this
> > >         service?"  The
> > >         obvious choice favored by every other application on the
> > >         internet (and most
> > >         private networks) is "DNS".
> > >
> > >         But ignoring that, there's an even more important question:
> > >
> > >                 "Who controls this service?"
> > >
> > >         If we build some fancy DHT solution we imply "some unnamed
> > >         benevolent
> > >         dictator yet to be appointed" will rule us all.  We've sorta
> > >         assumed this
> > >         (or similarly vague explanations) for the entire course of the
> > >         P2P-SIP
> > >         discussion.  For a long while I was not only on board, I was
> > >         promoting this
> > >         "let's solve security later" approach.
> > >
> > >         But then Adam proposed a different owner, the IANA.  And the
> > >         way to do that
> > >         is through the existing, tried-and-true interface of DNS
> > >         (specifically,
> > >         dynamic DNS).  I think this is obvious and brilliant.
> > >
> > >         Building a nonpartial, internationally represented, neutral
> > >         global third
> > >         party isn't easy, and while the IANA hasn't wildly succeeded
> > >         in this,
> > >         they're closer than anybody else (and decades ahead of
> > >         anything we'd invent
> > >         while searching for our benevolent dictator).
> > >
> > >
> > >         Now I'll grant DNS is limited to only, say, 99.999% of
> > >         users.  And I
> > >         understand that there's a lot of interest in how the remaining
> > >         0.001% of
> > >         users will work without it.  But I'm having a hard time
> > >         understanding why
> > >         we're even discussing them.  Indeed, Matthew's repeated calls
> > >         for a
> > >         real-world use case of a network that doesn't already have,
> > >         won't quickly
> > >         get, or couldn't easily be configured to have DNS has remained
> > >         unanswered.
> > >         (And I'd add to that filter "outside of loud talking range.")
> > >
> > >         Furthermore, I'll grant you could build a DHT layer that
> > >         serves as a
> > >         read-through cache of the central security mechanism that
> > >         provides a faster
> > >         response and propagation time than DNS.  But Akamai, UltraDNS,
> > >         and other
> > >         massive, global organizations have proven that DNS can
> > >         propagate pretty dang
> > >         quick.
> > >
> > >         And while I think it's an excellent, open question on how to
> > >         improve DNS to
> > >         make it even faster (perhaps with some kind of P2P layer), or
> > >         how to make it
> > >         work better in partially or totally disconnected scenarios, or
> > >         how to make
> > >         it more resistant to spamming/phishing and so forth, that
> > >         seems like a much
> > >         bigger problem than SIP alone and thus better suited to its
> > >         own group.
> > >
> > >         And finally, I agree there are open questions as to how my
> > >         car-of-the-future
> > >         will talk to my garage door and sprinkler system.  But why is
> > >         this relevant?
> > >         I share Adam's concern that we're dreaming up a panacea to
> > >         every conceivable
> > >         problem rather than standardizing a solution to problems that
> > >         actually exist
> > >         with technologies that are already at hand.
> > >
> > >
> > >         I apologize for being obtuse, but dynamic DNS (even with
> > >         today's
> > >         limitations, and perhaps with future improvements) seems like
> > >         such an
> > >         obvious way to solve the distributed name resolution problem
> > >         that I'm
> > >         astonished it's being discounted so easily in favor of complex
> > >         solutions to
> > >         nebulous problems experienced by nonexistent userbases.
> > >
> > >         I'll admit the conclusion is anticlimactic -- I'm a P2P
> > >         developer who likes
> > >         high tech as much as anybody.  But where has my reasoning
> > >         fallen off track?
> > >
> > >         -david
> > >
> > >         > -----Original Message-----
> > >         > From: p2p-sip-bounces at cs.columbia.edu [mailto:p2p-sip-
> > >         > bounces at cs.columbia.edu] On Behalf Of Matthew Kaufman
> > >         > Sent: Monday, November 27, 2006 10:18 AM
> > >         > To: hsinnrei at adobe.com; slavitch at gmail.com
> > >         > Cc: p2p-sip at cs.columbia.edu
> > >         > Subject: Re: [p2p-sip] Fwd: what's wrong with DNS?
> > >         >
> > >         > Michael Slavitch:
> > >         > > As a form of assessing identity DNS is particularly weak
> > >         and requires
> > >         > > considerable care and expert feeding as provisioning is
> > >         always
> > >         > > ultimately human-based requiring some kind of authority.
> > >         >
> > >         > Agreed. Though in small disconnected networks, zeroconf
> > >         using mDNS avoids
> > >         > much of this, and in large networks the IT guys or your ISP
> > >         manage the
> > >         > complexity for you (though they haven't solved the security
> > >         problem yet)
> > >         >
> > >         > > A self-organizing automated system needs no such
> > >         crutches.
> > >         >
> > >         > Except for bootstrapping to get the first two nodes talking.
> > >         Which, if on
> > >         > the Internet, means a way to find IP addresses of those
> > >         nodes. Which, on
> > >         > today's Internet, could use DNS. Or, as the Skype folks do,
> > >         hardcoded
> > >         > bootstrap
> > >         > addresses. But either way, bootstrap is such a crutch.
> > >         >
> > >         > > A key provision of P2PSIP is to eliminate the need for
> > >         such
> > >         > > authorities in the first place because they are
> > >         superfluous for the
> > >         > > application.  An extended social network needs no central
> > >         domain
> > >         > > administrator nor a central certificate authority.   It
> > >         needs at most
> > >         > > a human facilitator not an engineer.
> > >         >
> > >         > So P2PSIP includes a web-of-trust security model? If so, why
> > >         not leverage
> > >         > existing solutions for that as well (like PGP)?
> > >         >
> > >         > > Such social networks are not trite. For example social
> > >         networks are
> > >         > > the basis of all peer reviews for medical
> > >         journals.  Shouting doesn't
> > >         > > work when 500 doctors are to be consulted, on several
> > >         continents, with
> > >         > > varied access to the Internet, and over time.  The person
> > >         running the
> > >         > > peer review is a journal editor, a person who uses
> > >         end-user tools, not
> > >         > > a network engineer.  DNS isn't the selection criteria
> > >         here, reputation
> > >         > > is.
> > >         >
> > >         > But DNS is how the journal's mailing list manager finds the
> > >         mail hosts
> > >         > for each of the 500 doctors. Delivery of the article and
> > >         reply comments
> > >         > across several continents over time works today just fine
> > >         without P2P
> > >         > or DHTs or reputation-based web-of-trust security models.
> > >         See this mailing
> > >         > list as an existence proof.
> > >         >
> > >         > Matthew Kaufman
> > >         > matthew at matthew.at
> > >         >
> > >         > _______________________________________________
> > >         > p2p-sip mailing list
> > >         > p2p-sip at cs.columbia.edu
> > >         > https://lists.cs.columbia.edu/cucslists/listinfo/p2p-sip
> > >
> > >         _______________________________________________
> > >         p2p-sip mailing list
> > >         p2p-sip at cs.columbia.edu
> > >         https://lists.cs.columbia.edu/cucslists/listinfo/p2p-sip
> > >
> > > _______________________________________________
> > > p2p-sip mailing list
> > > p2p-sip at cs.columbia.edu
> > > https://lists.cs.columbia.edu/cucslists/listinfo/p2p-sip
> >
> > _______________________________________________
> > p2p-sip mailing list
> > p2p-sip at cs.columbia.edu
> > https://lists.cs.columbia.edu/cucslists/listinfo/p2p-sip
>
> _______________________________________________
> p2p-sip mailing list
> p2p-sip at cs.columbia.edu
> https://lists.cs.columbia.edu/cucslists/listinfo/p2p-sip
>

From hsinnrei at adobe.com  Tue Nov 28 10:59:18 2006
From: hsinnrei at adobe.com (Henry Sinnreich)
Date: Tue, 28 Nov 2006 07:59:18 -0800
Subject: [p2p-sip] Fwd: what's wrong with DNS?
In-Reply-To: <083001c712f9$79b91d80$640fa8c0@cis.neustar.com>
Message-ID: <24CCCC428EFEA2469BF046DB3C7A8D22141377@namail5.corp.adobe.com>

Brian's comments and the analogy with Skype sum it up quite well, though
P2P in the real world does not end with Skype. 

BTW: Existing P2P traffic is independent of the DNS and it dominates the
Internet traffic. Numbers between 60% and 80% have been published.

Also to consider is the use of a commercialized openDHT as has been
proposed (maybe already well funded by VC) that can not only act to
bootstrap new P2P systems but also as an _"Overlay Internet"_ in its own
right. 

See the papers on the openDHT web site. They are certainly a must read
for folks interested in this discussion.

As pointed out in reports on the Columbia University P2P system, the DHT
can be used (1) in the service mode and (2) in the data mode.

Thanks, Henry

-----Original Message-----
From: Brian Rosen [mailto:br at brianrosen.net] 
Sent: Tuesday, November 28, 2006 8:28 AM
To: 'David Barrett'; 'Matthew Kaufman'; Henry Sinnreich;
slavitch at gmail.com
Cc: p2p-sip at cs.columbia.edu
Subject: RE: [p2p-sip] Fwd: what's wrong with DNS?

The point of a "centralized" credentials service is to limit who can
call
you to someone who has credentials you accept.

Is any userid in any domain in the DNS acceptable to you for a
credential?
It's not to me.  I think one of the reasons our charter will be accepted
is
that we have allowed for a plausible credential issuing system.  If we
had
said that we don't plan to have one, credentials are not needed, and all
you
need is any domain to give you user at domain identity, then a lot of tough
questions would have been asked.  By saying you need credentials, and
credentials can be issued by some kind of authority, we create a
feasible
way to control SPIT.

Skype works pretty much the way we see p2psip working.  It has a
centralized
credential service; that's the website at skype.com where you sign up to
get
a SkypeId.  It also has a set of hardwired addresses of peers that lets
you
bootstrap your way to the cloud.  The addresses are not those of
centralized
directory servers.  They are just known to be available peers.  Any peer
will do, and they start with the well known ones.  Once you get going
and
are attached to the cloud, you don't need those peers any more, but they
are
your way to get started.  We'll need a bootstrap mechanism, and we
haven't
had any serious discussion on what that would be.

Brian

> -----Original Message-----
> From: p2p-sip-bounces at cs.columbia.edu [mailto:p2p-sip-
> bounces at cs.columbia.edu] On Behalf Of David Barrett
> Sent: Tuesday, November 28, 2006 1:50 AM
> To: 'Matthew Kaufman'; hsinnrei at adobe.com; slavitch at gmail.com
> Cc: p2p-sip at cs.columbia.edu
> Subject: Re: [p2p-sip] Fwd: what's wrong with DNS?
> 
> I think Matthew highlights the Achilles heel of so-called "pure" P2P:
it
> never is.  In the real world, there's always a central service behind
the
> curtain:
> 
> - In Skype's case, it's a central directory at a hard-coded IP
address.
> 
> - In P2P-SIP's case (if Frank and Brian are right) is some kind of
> centralized authorization/authentication/bootstrapping service.
> 
> So granted, there's the question of "how do you find this service?"
The
> obvious choice favored by every other application on the internet (and
> most
> private networks) is "DNS".
> 
> But ignoring that, there's an even more important question:
> 
> 	"Who controls this service?"
> 
> If we build some fancy DHT solution we imply "some unnamed benevolent
> dictator yet to be appointed" will rule us all.  We've sorta assumed
this
> (or similarly vague explanations) for the entire course of the P2P-SIP
> discussion.  For a long while I was not only on board, I was promoting
> this
> "let's solve security later" approach.
> 
> But then Adam proposed a different owner, the IANA.  And the way to do
> that
> is through the existing, tried-and-true interface of DNS
(specifically,
> dynamic DNS).  I think this is obvious and brilliant.
> 
> Building a nonpartial, internationally represented, neutral global
third
> party isn't easy, and while the IANA hasn't wildly succeeded in this,
> they're closer than anybody else (and decades ahead of anything we'd
> invent
> while searching for our benevolent dictator).
> 
> 
> Now I'll grant DNS is limited to only, say, 99.999% of users.  And I
> understand that there's a lot of interest in how the remaining 0.001%
of
> users will work without it.  But I'm having a hard time understanding
why
> we're even discussing them.  Indeed, Matthew's repeated calls for a
> real-world use case of a network that doesn't already have, won't
quickly
> get, or couldn't easily be configured to have DNS has remained
unanswered.
> (And I'd add to that filter "outside of loud talking range.")
> 
> Furthermore, I'll grant you could build a DHT layer that serves as a
> read-through cache of the central security mechanism that provides a
> faster
> response and propagation time than DNS.  But Akamai, UltraDNS, and
other
> massive, global organizations have proven that DNS can propagate
pretty
> dang
> quick.
> 
> And while I think it's an excellent, open question on how to improve
DNS
> to
> make it even faster (perhaps with some kind of P2P layer), or how to
make
> it
> work better in partially or totally disconnected scenarios, or how to
make
> it more resistant to spamming/phishing and so forth, that seems like a
> much
> bigger problem than SIP alone and thus better suited to its own group.
> 
> And finally, I agree there are open questions as to how my car-of-the-
> future
> will talk to my garage door and sprinkler system.  But why is this
> relevant?
> I share Adam's concern that we're dreaming up a panacea to every
> conceivable
> problem rather than standardizing a solution to problems that actually
> exist
> with technologies that are already at hand.
> 
> 
> I apologize for being obtuse, but dynamic DNS (even with today's
> limitations, and perhaps with future improvements) seems like such an
> obvious way to solve the distributed name resolution problem that I'm
> astonished it's being discounted so easily in favor of complex
solutions
> to
> nebulous problems experienced by nonexistent userbases.
> 
> I'll admit the conclusion is anticlimactic -- I'm a P2P developer who
> likes
> high tech as much as anybody.  But where has my reasoning fallen off
> track?
> 
> -david
> 
> > -----Original Message-----
> > From: p2p-sip-bounces at cs.columbia.edu [mailto:p2p-sip-
> > bounces at cs.columbia.edu] On Behalf Of Matthew Kaufman
> > Sent: Monday, November 27, 2006 10:18 AM
> > To: hsinnrei at adobe.com; slavitch at gmail.com
> > Cc: p2p-sip at cs.columbia.edu
> > Subject: Re: [p2p-sip] Fwd: what's wrong with DNS?
> >
> > Michael Slavitch:
> > > As a form of assessing identity DNS is particularly weak and
requires
> > > considerable care and expert feeding as provisioning is always
> > > ultimately human-based requiring some kind of authority.
> >
> > Agreed. Though in small disconnected networks, zeroconf using mDNS
> avoids
> > much of this, and in large networks the IT guys or your ISP manage
the
> > complexity for you (though they haven't solved the security problem
yet)
> >
> > > A self-organizing automated system needs no such crutches.
> >
> > Except for bootstrapping to get the first two nodes talking. Which,
if
> on
> > the Internet, means a way to find IP addresses of those nodes.
Which, on
> > today's Internet, could use DNS. Or, as the Skype folks do,
hardcoded
> > bootstrap
> > addresses. But either way, bootstrap is such a crutch.
> >
> > > A key provision of P2PSIP is to eliminate the need for such
> > > authorities in the first place because they are superfluous for
the
> > > application.  An extended social network needs no central domain
> > > administrator nor a central certificate authority.   It needs at
most
> > > a human facilitator not an engineer.
> >
> > So P2PSIP includes a web-of-trust security model? If so, why not
> leverage
> > existing solutions for that as well (like PGP)?
> >
> > > Such social networks are not trite. For example social networks
are
> > > the basis of all peer reviews for medical journals.  Shouting
doesn't
> > > work when 500 doctors are to be consulted, on several continents,
with
> > > varied access to the Internet, and over time.  The person running
the
> > > peer review is a journal editor, a person who uses end-user tools,
not
> > > a network engineer.  DNS isn't the selection criteria here,
reputation
> > > is.
> >
> > But DNS is how the journal's mailing list manager finds the mail
hosts
> > for each of the 500 doctors. Delivery of the article and reply
comments
> > across several continents over time works today just fine without
P2P
> > or DHTs or reputation-based web-of-trust security models. See this
> mailing
> > list as an existence proof.
> >
> > Matthew Kaufman
> > matthew at matthew.at
> >
> > _______________________________________________
> > p2p-sip mailing list
> > p2p-sip at cs.columbia.edu
> > https://lists.cs.columbia.edu/cucslists/listinfo/p2p-sip
> 
> _______________________________________________
> p2p-sip mailing list
> p2p-sip at cs.columbia.edu
> https://lists.cs.columbia.edu/cucslists/listinfo/p2p-sip



From slavitch at gmail.com  Tue Nov 28 11:14:59 2006
From: slavitch at gmail.com (Michael Slavitch)
Date: Tue, 28 Nov 2006 11:14:59 -0500
Subject: [p2p-sip] Fwd: what's wrong with DNS?
In-Reply-To: <24CCCC428EFEA2469BF046DB3C7A8D22141377@namail5.corp.adobe.com>
References: <083001c712f9$79b91d80$640fa8c0@cis.neustar.com>
	<24CCCC428EFEA2469BF046DB3C7A8D22141377@namail5.corp.adobe.com>
Message-ID: <1fc32c8f0611280814g5e80f6e4x2a6ec3e0d632d49c@mail.gmail.com>

Google's "I'm feeling lucky" on "Overlay Internet" is a good one.

http://tinyurl.com/y4rmms

The Overlay Internet cat is no longer in any bag.  It is mainstream thinking.

M

On 11/28/06, Henry Sinnreich <hsinnrei at adobe.com> wrote:
> Brian's comments and the analogy with Skype sum it up quite well, though
> P2P in the real world does not end with Skype.
>
> BTW: Existing P2P traffic is independent of the DNS and it dominates the
> Internet traffic. Numbers between 60% and 80% have been published.
>
> Also to consider is the use of a commercialized openDHT as has been
> proposed (maybe already well funded by VC) that can not only act to
> bootstrap new P2P systems but also as an _"Overlay Internet"_ in its own
> right.
>
> See the papers on the openDHT web site. They are certainly a must read
> for folks interested in this discussion.
>
> As pointed out in reports on the Columbia University P2P system, the DHT
> can be used (1) in the service mode and (2) in the data mode.
>
> Thanks, Henry
>
> -----Original Message-----
> From: Brian Rosen [mailto:br at brianrosen.net]
> Sent: Tuesday, November 28, 2006 8:28 AM
> To: 'David Barrett'; 'Matthew Kaufman'; Henry Sinnreich;
> slavitch at gmail.com
> Cc: p2p-sip at cs.columbia.edu
> Subject: RE: [p2p-sip] Fwd: what's wrong with DNS?
>
> The point of a "centralized" credentials service is to limit who can
> call
> you to someone who has credentials you accept.
>
> Is any userid in any domain in the DNS acceptable to you for a
> credential?
> It's not to me.  I think one of the reasons our charter will be accepted
> is
> that we have allowed for a plausible credential issuing system.  If we
> had
> said that we don't plan to have one, credentials are not needed, and all
> you
> need is any domain to give you user at domain identity, then a lot of tough
> questions would have been asked.  By saying you need credentials, and
> credentials can be issued by some kind of authority, we create a
> feasible
> way to control SPIT.
>
> Skype works pretty much the way we see p2psip working.  It has a
> centralized
> credential service; that's the website at skype.com where you sign up to
> get
> a SkypeId.  It also has a set of hardwired addresses of peers that lets
> you
> bootstrap your way to the cloud.  The addresses are not those of
> centralized
> directory servers.  They are just known to be available peers.  Any peer
> will do, and they start with the well known ones.  Once you get going
> and
> are attached to the cloud, you don't need those peers any more, but they
> are
> your way to get started.  We'll need a bootstrap mechanism, and we
> haven't
> had any serious discussion on what that would be.
>
> Brian
>
> > -----Original Message-----
> > From: p2p-sip-bounces at cs.columbia.edu [mailto:p2p-sip-
> > bounces at cs.columbia.edu] On Behalf Of David Barrett
> > Sent: Tuesday, November 28, 2006 1:50 AM
> > To: 'Matthew Kaufman'; hsinnrei at adobe.com; slavitch at gmail.com
> > Cc: p2p-sip at cs.columbia.edu
> > Subject: Re: [p2p-sip] Fwd: what's wrong with DNS?
> >
> > I think Matthew highlights the Achilles heel of so-called "pure" P2P:
> it
> > never is.  In the real world, there's always a central service behind
> the
> > curtain:
> >
> > - In Skype's case, it's a central directory at a hard-coded IP
> address.
> >
> > - In P2P-SIP's case (if Frank and Brian are right) is some kind of
> > centralized authorization/authentication/bootstrapping service.
> >
> > So granted, there's the question of "how do you find this service?"
> The
> > obvious choice favored by every other application on the internet (and
> > most
> > private networks) is "DNS".
> >
> > But ignoring that, there's an even more important question:
> >
> >       "Who controls this service?"
> >
> > If we build some fancy DHT solution we imply "some unnamed benevolent
> > dictator yet to be appointed" will rule us all.  We've sorta assumed
> this
> > (or similarly vague explanations) for the entire course of the P2P-SIP
> > discussion.  For a long while I was not only on board, I was promoting
> > this
> > "let's solve security later" approach.
> >
> > But then Adam proposed a different owner, the IANA.  And the way to do
> > that
> > is through the existing, tried-and-true interface of DNS
> (specifically,
> > dynamic DNS).  I think this is obvious and brilliant.
> >
> > Building a nonpartial, internationally represented, neutral global
> third
> > party isn't easy, and while the IANA hasn't wildly succeeded in this,
> > they're closer than anybody else (and decades ahead of anything we'd
> > invent
> > while searching for our benevolent dictator).
> >
> >
> > Now I'll grant DNS is limited to only, say, 99.999% of users.  And I
> > understand that there's a lot of interest in how the remaining 0.001%
> of
> > users will work without it.  But I'm having a hard time understanding
> why
> > we're even discussing them.  Indeed, Matthew's repeated calls for a
> > real-world use case of a network that doesn't already have, won't
> quickly
> > get, or couldn't easily be configured to have DNS has remained
> unanswered.
> > (And I'd add to that filter "outside of loud talking range.")
> >
> > Furthermore, I'll grant you could build a DHT layer that serves as a
> > read-through cache of the central security mechanism that provides a
> > faster
> > response and propagation time than DNS.  But Akamai, UltraDNS, and
> other
> > massive, global organizations have proven that DNS can propagate
> pretty
> > dang
> > quick.
> >
> > And while I think it's an excellent, open question on how to improve
> DNS
> > to
> > make it even faster (perhaps with some kind of P2P layer), or how to
> make
> > it
> > work better in partially or totally disconnected scenarios, or how to
> make
> > it more resistant to spamming/phishing and so forth, that seems like a
> > much
> > bigger problem than SIP alone and thus better suited to its own group.
> >
> > And finally, I agree there are open questions as to how my car-of-the-
> > future
> > will talk to my garage door and sprinkler system.  But why is this
> > relevant?
> > I share Adam's concern that we're dreaming up a panacea to every
> > conceivable
> > problem rather than standardizing a solution to problems that actually
> > exist
> > with technologies that are already at hand.
> >
> >
> > I apologize for being obtuse, but dynamic DNS (even with today's
> > limitations, and perhaps with future improvements) seems like such an
> > obvious way to solve the distributed name resolution problem that I'm
> > astonished it's being discounted so easily in favor of complex
> solutions
> > to
> > nebulous problems experienced by nonexistent userbases.
> >
> > I'll admit the conclusion is anticlimactic -- I'm a P2P developer who
> > likes
> > high tech as much as anybody.  But where has my reasoning fallen off
> > track?
> >
> > -david
> >
> > > -----Original Message-----
> > > From: p2p-sip-bounces at cs.columbia.edu [mailto:p2p-sip-
> > > bounces at cs.columbia.edu] On Behalf Of Matthew Kaufman
> > > Sent: Monday, November 27, 2006 10:18 AM
> > > To: hsinnrei at adobe.com; slavitch at gmail.com
> > > Cc: p2p-sip at cs.columbia.edu
> > > Subject: Re: [p2p-sip] Fwd: what's wrong with DNS?
> > >
> > > Michael Slavitch:
> > > > As a form of assessing identity DNS is particularly weak and
> requires
> > > > considerable care and expert feeding as provisioning is always
> > > > ultimately human-based requiring some kind of authority.
> > >
> > > Agreed. Though in small disconnected networks, zeroconf using mDNS
> > avoids
> > > much of this, and in large networks the IT guys or your ISP manage
> the
> > > complexity for you (though they haven't solved the security problem
> yet)
> > >
> > > > A self-organizing automated system needs no such crutches.
> > >
> > > Except for bootstrapping to get the first two nodes talking. Which,
> if
> > on
> > > the Internet, means a way to find IP addresses of those nodes.
> Which, on
> > > today's Internet, could use DNS. Or, as the Skype folks do,
> hardcoded
> > > bootstrap
> > > addresses. But either way, bootstrap is such a crutch.
> > >
> > > > A key provision of P2PSIP is to eliminate the need for such
> > > > authorities in the first place because they are superfluous for
> the
> > > > application.  An extended social network needs no central domain
> > > > administrator nor a central certificate authority.   It needs at
> most
> > > > a human facilitator not an engineer.
> > >
> > > So P2PSIP includes a web-of-trust security model? If so, why not
> > leverage
> > > existing solutions for that as well (like PGP)?
> > >
> > > > Such social networks are not trite. For example social networks
> are
> > > > the basis of all peer reviews for medical journals.  Shouting
> doesn't
> > > > work when 500 doctors are to be consulted, on several continents,
> with
> > > > varied access to the Internet, and over time.  The person running
> the
> > > > peer review is a journal editor, a person who uses end-user tools,
> not
> > > > a network engineer.  DNS isn't the selection criteria here,
> reputation
> > > > is.
> > >
> > > But DNS is how the journal's mailing list manager finds the mail
> hosts
> > > for each of the 500 doctors. Delivery of the article and reply
> comments
> > > across several continents over time works today just fine without
> P2P
> > > or DHTs or reputation-based web-of-trust security models. See this
> > mailing
> > > list as an existence proof.
> > >
> > > Matthew Kaufman
> > > matthew at matthew.at
> > >
> > > _______________________________________________
> > > p2p-sip mailing list
> > > p2p-sip at cs.columbia.edu
> > > https://lists.cs.columbia.edu/cucslists/listinfo/p2p-sip
> >
> > _______________________________________________
> > p2p-sip mailing list
> > p2p-sip at cs.columbia.edu
> > https://lists.cs.columbia.edu/cucslists/listinfo/p2p-sip
>
>

From fwmiller at cornfed.com  Tue Nov 28 11:32:36 2006
From: fwmiller at cornfed.com (Frank W. Miller)
Date: Tue, 28 Nov 2006 09:32:36 -0700
Subject: [p2p-sip] Fwd: what's wrong with DNS?
In-Reply-To: <085001c71305$be4f70f0$640fa8c0@cis.neustar.com>
References: <085001c71305$be4f70f0$640fa8c0@cis.neustar.com>
Message-ID: <1164731556.2622.31.camel@localhost.localdomain>

On Tue, 2006-11-28 at 10:56 -0500, Brian Rosen wrote:
> You understand that you are proposing a centralized service that
> authenticates you every time you boot, and a centralized authorization
> service that authorizes you AT CALL TIME, at least the first time per boot?
> 

Not exactly.  Authorization happens at boot, i.e when your client starts
up.  Authentication happens at the first interaction between two UAs.


> The direction we are going is that you use the centralized credentialing
> service to get credentials effectively once in a lifetime of the credential,
> which might last at least months, and maybe years.  Thereafter, the
> distributed hash table takes care of authentication and authorization peer
> to peer without any central service.
> 

Understood.  I question the practicality of this for at least a couple
of reasons.

First, my view is that credentials should be changed periodically.  The
longer a credential is available, the more likely it is that it can be
hacked.

Second, the only real difference is how often the credentials are issued
and checked.  The DHT approach would do it "once" for all time, my
proposal does it "once" at client boot time and at the beginning of a
UA->UA interaction.  The need or lack of need for a centralized
authority does not go away because of the use of a DHT to lengthen the
time that credentials are valid.


> I would definitely not call what you are proposing a peer to peer service;
> it relies on centralized authentication and centralized authorization
> services in real time.
> 

Well, its P2P except for initial authorization and subsequent initial
authentications.


FM




From br at brianrosen.net  Tue Nov 28 12:36:10 2006
From: br at brianrosen.net (Brian Rosen)
Date: Tue, 28 Nov 2006 12:36:10 -0500
Subject: [p2p-sip] Fwd: what's wrong with DNS?
In-Reply-To: <1164731556.2622.31.camel@localhost.localdomain>
Message-ID: <088101c71313$b5be1c30$640fa8c0@cis.neustar.com>


I think you are confusing a credential validation from a credential
issuance.

I think what you have in mind is that you get a credential that changes
every few months: it's the password you use to authenticate.  That
credential is validated every time you boot.

The plan I think most of us have in mind is that you get a credential,
perhaps something like an x.509 cert, good for a couple of months or years,
that is validated every time you boot (e.g. connect to the overlay).

The difference is you want a centralized service to do the authentication,
and we want a p2p distributed service.

Then, you want a centralized authorization service, where we think the DHT
provides that also as a p2p distributed service.

Are there other people on the list who think what Frank is advocating is a
good idea?

Brian


> -----Original Message-----
> From: Frank W. Miller [mailto:fwmiller at cornfed.com]
> Sent: Tuesday, November 28, 2006 11:33 AM
> To: Brian Rosen
> Cc: 'Eunsoo Shim'; p2p-sip at cs.columbia.edu
> Subject: RE: [p2p-sip] Fwd: what's wrong with DNS?
> 
> On Tue, 2006-11-28 at 10:56 -0500, Brian Rosen wrote:
> > You understand that you are proposing a centralized service that
> > authenticates you every time you boot, and a centralized authorization
> > service that authorizes you AT CALL TIME, at least the first time per
> boot?
> >
> 
> Not exactly.  Authorization happens at boot, i.e when your client starts
> up.  Authentication happens at the first interaction between two UAs.
> 
> 
> > The direction we are going is that you use the centralized credentialing
> > service to get credentials effectively once in a lifetime of the
> credential,
> > which might last at least months, and maybe years.  Thereafter, the
> > distributed hash table takes care of authentication and authorization
> peer
> > to peer without any central service.
> >
> 
> Understood.  I question the practicality of this for at least a couple
> of reasons.
> 
> First, my view is that credentials should be changed periodically.  The
> longer a credential is available, the more likely it is that it can be
> hacked.
> 
> Second, the only real difference is how often the credentials are issued
> and checked.  The DHT approach would do it "once" for all time, my
> proposal does it "once" at client boot time and at the beginning of a
> UA->UA interaction.  The need or lack of need for a centralized
> authority does not go away because of the use of a DHT to lengthen the
> time that credentials are valid.
> 
> 
> > I would definitely not call what you are proposing a peer to peer
> service;
> > it relies on centralized authentication and centralized authorization
> > services in real time.
> >
> 
> Well, its P2P except for initial authorization and subsequent initial
> authentications.
> 
> 
> FM
> 



From slavitch at gmail.com  Tue Nov 28 12:53:03 2006
From: slavitch at gmail.com (Michael Slavitch)
Date: Tue, 28 Nov 2006 12:53:03 -0500
Subject: [p2p-sip] Fwd: what's wrong with DNS?
In-Reply-To: <088101c71313$b5be1c30$640fa8c0@cis.neustar.com>
References: <1164731556.2622.31.camel@localhost.localdomain>
	<088101c71313$b5be1c30$640fa8c0@cis.neustar.com>
Message-ID: <1fc32c8f0611280953i7d302628i2775ce0405febda0@mail.gmail.com>

Anyone using a Windows notebook that authenticates to a domain is
using this technology right now.  This is not open research.

On 11/28/06, Brian Rosen <br at brianrosen.net> wrote:
>
> I think you are confusing a credential validation from a credential
> issuance.
>
> I think what you have in mind is that you get a credential that changes
> every few months: it's the password you use to authenticate.  That
> credential is validated every time you boot.
>
> The plan I think most of us have in mind is that you get a credential,
> perhaps something like an x.509 cert, good for a couple of months or years,
> that is validated every time you boot (e.g. connect to the overlay).
>
> The difference is you want a centralized service to do the authentication,
> and we want a p2p distributed service.
>
> Then, you want a centralized authorization service, where we think the DHT
> provides that also as a p2p distributed service.
>
> Are there other people on the list who think what Frank is advocating is a
> good idea?
>
> Brian
>
>
> > -----Original Message-----
> > From: Frank W. Miller [mailto:fwmiller at cornfed.com]
> > Sent: Tuesday, November 28, 2006 11:33 AM
> > To: Brian Rosen
> > Cc: 'Eunsoo Shim'; p2p-sip at cs.columbia.edu
> > Subject: RE: [p2p-sip] Fwd: what's wrong with DNS?
> >
> > On Tue, 2006-11-28 at 10:56 -0500, Brian Rosen wrote:
> > > You understand that you are proposing a centralized service that
> > > authenticates you every time you boot, and a centralized authorization
> > > service that authorizes you AT CALL TIME, at least the first time per
> > boot?
> > >
> >
> > Not exactly.  Authorization happens at boot, i.e when your client starts
> > up.  Authentication happens at the first interaction between two UAs.
> >
> >
> > > The direction we are going is that you use the centralized credentialing
> > > service to get credentials effectively once in a lifetime of the
> > credential,
> > > which might last at least months, and maybe years.  Thereafter, the
> > > distributed hash table takes care of authentication and authorization
> > peer
> > > to peer without any central service.
> > >
> >
> > Understood.  I question the practicality of this for at least a couple
> > of reasons.
> >
> > First, my view is that credentials should be changed periodically.  The
> > longer a credential is available, the more likely it is that it can be
> > hacked.
> >
> > Second, the only real difference is how often the credentials are issued
> > and checked.  The DHT approach would do it "once" for all time, my
> > proposal does it "once" at client boot time and at the beginning of a
> > UA->UA interaction.  The need or lack of need for a centralized
> > authority does not go away because of the use of a DHT to lengthen the
> > time that credentials are valid.
> >
> >
> > > I would definitely not call what you are proposing a peer to peer
> > service;
> > > it relies on centralized authentication and centralized authorization
> > > services in real time.
> > >
> >
> > Well, its P2P except for initial authorization and subsequent initial
> > authentications.
> >
> >
> > FM
> >
>
>
> _______________________________________________
> p2p-sip mailing list
> p2p-sip at cs.columbia.edu
> https://lists.cs.columbia.edu/cucslists/listinfo/p2p-sip
>

From fwmiller at cornfed.com  Tue Nov 28 13:24:17 2006
From: fwmiller at cornfed.com (Frank W. Miller)
Date: Tue, 28 Nov 2006 11:24:17 -0700
Subject: [p2p-sip] Fwd: what's wrong with DNS?
In-Reply-To: <088101c71313$b5be1c30$640fa8c0@cis.neustar.com>
References: <088101c71313$b5be1c30$640fa8c0@cis.neustar.com>
Message-ID: <1164738258.2622.51.camel@localhost.localdomain>

On Tue, 2006-11-28 at 12:36 -0500, Brian Rosen wrote:
> I think what you have in mind is that you get a credential that changes
> every few months: it's the password you use to authenticate.  That
> credential is validated every time you boot.
> 

Well, in the proposal I made, there are actually two "credentials."
There is the password that is issued however you like, its outside the
scope of the proposal other than the Author/Authen authority being aware
of it.  Then there is the "cookie."  This is the thing that changes
frequently.  The password is long lived but the cookie is regenerated
when the client attaches to the overlay.  Then other UAs use the cookie
to do authentication.  This proposal isn't all that different than
current SIP, i.e. A registrar challenges the UA when it registers and
certainly anyone can challenge a call setup when that happens.


> The plan I think most of us have in mind is that you get a credential,
> perhaps something like an x.509 cert, good for a couple of months or years,
> that is validated every time you boot (e.g. connect to the overlay).
> 

K, so the x.509 cert would be the "password", and the difference would
be that the "validation" would yield value, i.e. the "cookie", that
would be used for authentication.  So, you're thinking that you just
pass the 509 cert around?


> The difference is you want a centralized service to do the authentication,
> and we want a p2p distributed service.
> 

Yes, I am a firm believer in a trusted third party acting as an
authority.


> Then, you want a centralized authorization service, where we think the DHT
> provides that also as a p2p distributed service.
> 

I'm leery of this.  At the bottom of that is some peer vouching for some
other peer.  I don't think thats good enough in many situations.  The
central authority doesn't have to be Big Brother, it just has to be
something trusted.  Who knows, that may be one or a group of peers
acting as the trusted authority in some deployment situations, but it
could be Big Brother too if thats required.

Its interesting, you've mentioned the DHT a number of times.  I've been
following this list for quite some time.  Maybe I missed it but is the
protocol being done such that the use of a DHT is assumed?


FM




From spencer at mcsr-labs.org  Tue Nov 28 13:30:25 2006
From: spencer at mcsr-labs.org (Spencer Dawkins)
Date: Tue, 28 Nov 2006 12:30:25 -0600
Subject: [p2p-sip] Fwd: what's wrong with DNS?
References: <200611271818.kARII0191955@where.matthew.at><015301c712b9$5daeb950$0200a8c0@Quinthar><7b683f3f0611280630t369ea8aesc2ceb1a1efea110c@mail.gmail.com>
	<1164726951.2622.19.camel@localhost.localdomain>
Message-ID: <08fe01c7131b$45f46dc0$63e06841@china.huawei.com>

Without reference to the specific proposal, this is a good example of the 
level of detail we need in order to have discussions that go someplace and 
terminate in finite time - we'll be chartered to produce specifications, not 
to fill up working group mailing list archives!

Frank, thanks for being concrete. The followups show that this posting 
helped people understand what was being proposed.

Spencer

> I would propose something like this:
>
> There are two elements, User Agents and the Identity Service. Identity
> refers to messages that are sent between an Internet host and the
> Identity Service to authorize and authenticate UAIs.  The Identity
> Service is assumed to be authoritative and trusted, and very likely a
> centralized service.  Identity operations include the following
> functions:
>
>   1.  Authorization
>
>   2.  Authentication
>
> Authorization
>
> I call the user portion of a SIP URI the User Agent Instance (UAI).
> The UAI is allocated to the UA by some means outside of the scope of this
> protocol, i.e. by visiting a website or something.  Authorization refers
> to the registration of a UA instance for a UAI with an Identity Service.
> UAs are generally authorized when a user wants to begin interacting with
> other UAs, e.g. when their UA client program is started.  (Note that these
> messages are notional.  They don't have to even be SIP.  The idea of
> these templates is to list the information moving between the elements)
>
>
>   Frank            Identity Service
>     |      AUTHORIZE      |
>     |-------------------->|
>     |       200 OK        |
>     |<--------------------|
>     |                     |
>
>
>   AUTHORIZE
>   UAI: Frank
>   Password: ... Frank's password ...
>   CSeq: 2 AUTHORIZE
>
> Frank's UA requests authorization of the UAI.  The request includes a
> shared secret that was setup when the UAI was issued.  The shared secret 
> must
> be sent encrypted.
>
>
>   200 OK
>   UAI: Frank
>   Authorization: ... Cookie sent to anyone authenticating Frank ...
>   CSeq: 2 AUTHORIZE
>
> Frank's UA is returned a Cookie.  This cookie will be sent to other UAs
> for subsequent interactions.  Other UAs can use the cookie for 
> authenticating
> Frank's UA.  Note that this assumes the Authorization cookie is encrypted.
>
>
> Authentication
>
> Authentication refers to the process of querying the Identity Service
> regarding a UAI.  The procedure described herein is usually attempted
> when a UA receives its first request for interaction from another UA.
>
>
>    Bob            Identity Service
>     |     AUTHENTICATE    |
>     |-------------------->|
>     |       200 OK        |
>     |<--------------------|
>     |                     |
>
>
>
>   AUTHENTICATE
>   UAI: Frank
>   Password: ... Bob's password ...
>   CSeq: 2 AUTHORIZE
>
> Bob send's an authentication request to the Identity Service.  Bob is
> only allowed to authenticate Frank if his UA is authorized using the
> shared secret for Bob's UAI.  The shared secret must be encrypted.
>
>
>   MSP 200 OK
>   UAI: Frank
>   Authentication: ... Cookie sent to Frank during authorization ...
>   CSeq: 2 AUTHORIZE
>
> The identity service returns the cookie that was sent by Frank with
> the initial interaction.  Bob's UAI can compare the two and determine
> if Frank is really Frank.  Again, the cookie must be encrypted.
>
> This proposal requires some level of encryption in all these
> messages.  I am actually in favor of SSL/TLS being required for
> transport of these messages, whether that is for this separate
> protocol blurb or for SIP in general.
>
> The management of UAIs for each p2p cloud becomes the authoritative
> information for that cloud.  I can imagine services like FWD providing
> it for free and I can also imagine commercial services charging for
> the service.
>
> FM
>
>
>
> On Tue, 2006-11-28 at 09:30 -0500, Eunsoo Shim wrote:
>> I am not sure what your proposal is.
>> Could you provide more specifics on your proposal?
>> For example, are you suggesting that we use existing DNS servers to
>> store whatever we want to store at the P2P(SIP) network?
>> Or are you suggesting that we use the DNS PROTOCOL as (part of) the
>> protocol for P2P SIP and build a separate (global or ad-hoc) DNS-like
>> infrastructure?
>> Thanks.
>>
>> Eunsoo
>>
>> On 11/28/06, David Barrett <dbarrett at quinthar.com> wrote:
>>         I think Matthew highlights the Achilles heel of so-called
>>         "pure" P2P: it
>>         never is.  In the real world, there's always a central service
>>         behind the
>>         curtain:
>>
>>         - In Skype's case, it's a central directory at a hard-coded IP
>>         address.
>>
>>         - In P2P-SIP's case (if Frank and Brian are right) is some
>>         kind of
>>         centralized authorization/authentication/bootstrapping
>>         service.
>>
>>         So granted, there's the question of "how do you find this
>>         service?"  The
>>         obvious choice favored by every other application on the
>>         internet (and most
>>         private networks) is "DNS".
>>
>>         But ignoring that, there's an even more important question:
>>
>>                 "Who controls this service?"
>>
>>         If we build some fancy DHT solution we imply "some unnamed
>>         benevolent
>>         dictator yet to be appointed" will rule us all.  We've sorta
>>         assumed this
>>         (or similarly vague explanations) for the entire course of the
>>         P2P-SIP
>>         discussion.  For a long while I was not only on board, I was
>>         promoting this
>>         "let's solve security later" approach.
>>
>>         But then Adam proposed a different owner, the IANA.  And the
>>         way to do that
>>         is through the existing, tried-and-true interface of DNS
>>         (specifically,
>>         dynamic DNS).  I think this is obvious and brilliant.
>>
>>         Building a nonpartial, internationally represented, neutral
>>         global third
>>         party isn't easy, and while the IANA hasn't wildly succeeded
>>         in this,
>>         they're closer than anybody else (and decades ahead of
>>         anything we'd invent
>>         while searching for our benevolent dictator).
>>
>>
>>         Now I'll grant DNS is limited to only, say, 99.999% of
>>         users.  And I
>>         understand that there's a lot of interest in how the remaining
>>         0.001% of
>>         users will work without it.  But I'm having a hard time
>>         understanding why
>>         we're even discussing them.  Indeed, Matthew's repeated calls
>>         for a
>>         real-world use case of a network that doesn't already have,
>>         won't quickly
>>         get, or couldn't easily be configured to have DNS has remained
>>         unanswered.
>>         (And I'd add to that filter "outside of loud talking range.")
>>
>>         Furthermore, I'll grant you could build a DHT layer that
>>         serves as a
>>         read-through cache of the central security mechanism that
>>         provides a faster
>>         response and propagation time than DNS.  But Akamai, UltraDNS,
>>         and other
>>         massive, global organizations have proven that DNS can
>>         propagate pretty dang
>>         quick.
>>
>>         And while I think it's an excellent, open question on how to
>>         improve DNS to
>>         make it even faster (perhaps with some kind of P2P layer), or
>>         how to make it
>>         work better in partially or totally disconnected scenarios, or
>>         how to make
>>         it more resistant to spamming/phishing and so forth, that
>>         seems like a much
>>         bigger problem than SIP alone and thus better suited to its
>>         own group.
>>
>>         And finally, I agree there are open questions as to how my
>>         car-of-the-future
>>         will talk to my garage door and sprinkler system.  But why is
>>         this relevant?
>>         I share Adam's concern that we're dreaming up a panacea to
>>         every conceivable
>>         problem rather than standardizing a solution to problems that
>>         actually exist
>>         with technologies that are already at hand.
>>
>>
>>         I apologize for being obtuse, but dynamic DNS (even with
>>         today's
>>         limitations, and perhaps with future improvements) seems like
>>         such an
>>         obvious way to solve the distributed name resolution problem
>>         that I'm
>>         astonished it's being discounted so easily in favor of complex
>>         solutions to
>>         nebulous problems experienced by nonexistent userbases.
>>
>>         I'll admit the conclusion is anticlimactic -- I'm a P2P
>>         developer who likes
>>         high tech as much as anybody.  But where has my reasoning
>>         fallen off track?
>>
>>         -david
>>
>>         > -----Original Message-----
>>         > From: p2p-sip-bounces at cs.columbia.edu [mailto:p2p-sip-
>>         > bounces at cs.columbia.edu] On Behalf Of Matthew Kaufman
>>         > Sent: Monday, November 27, 2006 10:18 AM
>>         > To: hsinnrei at adobe.com; slavitch at gmail.com
>>         > Cc: p2p-sip at cs.columbia.edu
>>         > Subject: Re: [p2p-sip] Fwd: what's wrong with DNS?
>>         >
>>         > Michael Slavitch:
>>         > > As a form of assessing identity DNS is particularly weak
>>         and requires
>>         > > considerable care and expert feeding as provisioning is
>>         always
>>         > > ultimately human-based requiring some kind of authority.
>>         >
>>         > Agreed. Though in small disconnected networks, zeroconf
>>         using mDNS avoids
>>         > much of this, and in large networks the IT guys or your ISP
>>         manage the
>>         > complexity for you (though they haven't solved the security
>>         problem yet)
>>         >
>>         > > A self-organizing automated system needs no such
>>         crutches.
>>         >
>>         > Except for bootstrapping to get the first two nodes talking.
>>         Which, if on
>>         > the Internet, means a way to find IP addresses of those
>>         nodes. Which, on
>>         > today's Internet, could use DNS. Or, as the Skype folks do,
>>         hardcoded
>>         > bootstrap
>>         > addresses. But either way, bootstrap is such a crutch.
>>         >
>>         > > A key provision of P2PSIP is to eliminate the need for
>>         such
>>         > > authorities in the first place because they are
>>         superfluous for the
>>         > > application.  An extended social network needs no central
>>         domain
>>         > > administrator nor a central certificate authority.   It
>>         needs at most
>>         > > a human facilitator not an engineer.
>>         >
>>         > So P2PSIP includes a web-of-trust security model? If so, why
>>         not leverage
>>         > existing solutions for that as well (like PGP)?
>>         >
>>         > > Such social networks are not trite. For example social
>>         networks are
>>         > > the basis of all peer reviews for medical
>>         journals.  Shouting doesn't
>>         > > work when 500 doctors are to be consulted, on several
>>         continents, with
>>         > > varied access to the Internet, and over time.  The person
>>         running the
>>         > > peer review is a journal editor, a person who uses
>>         end-user tools, not
>>         > > a network engineer.  DNS isn't the selection criteria
>>         here, reputation
>>         > > is.
>>         >
>>         > But DNS is how the journal's mailing list manager finds the
>>         mail hosts
>>         > for each of the 500 doctors. Delivery of the article and
>>         reply comments
>>         > across several continents over time works today just fine
>>         without P2P
>>         > or DHTs or reputation-based web-of-trust security models.
>>         See this mailing
>>         > list as an existence proof.
>>         >
>>         > Matthew Kaufman
>>         > matthew at matthew.at
>>         >
>>         > _______________________________________________
>>         > p2p-sip mailing list
>>         > p2p-sip at cs.columbia.edu
>>         > https://lists.cs.columbia.edu/cucslists/listinfo/p2p-sip
>>
>>         _______________________________________________
>>         p2p-sip mailing list
>>         p2p-sip at cs.columbia.edu
>>         https://lists.cs.columbia.edu/cucslists/listinfo/p2p-sip
>>
>> _______________________________________________
>> p2p-sip mailing list
>> p2p-sip at cs.columbia.edu
>> https://lists.cs.columbia.edu/cucslists/listinfo/p2p-sip
>
> _______________________________________________
> p2p-sip mailing list
> p2p-sip at cs.columbia.edu
> https://lists.cs.columbia.edu/cucslists/listinfo/p2p-sip
> 



From br at brianrosen.net  Tue Nov 28 14:15:49 2006
From: br at brianrosen.net (Brian Rosen)
Date: Tue, 28 Nov 2006 14:15:49 -0500
Subject: [p2p-sip] Fwd: what's wrong with DNS?
In-Reply-To: <1164738258.2622.51.camel@localhost.localdomain>
Message-ID: <08b301c71321$9fd25a40$640fa8c0@cis.neustar.com>

We haven't designed the whole mechanism for authenticating to the overlay
network yet, so it's hard to really compare ideas.  Cookies, as a way to
shortcut more expensive authentication/authorization mechanisms can be
useful.  Don't know yet whether we'll be using such ideas.

I do think that one way this could work is that the centralized credential
authority issues a cert which contains the identity in the domain.
Authentication would be a typical RSA operation using the cert to prove that
the peer is the identity in the cert.

We'll all agree we need a trusted entity to issue credentials.  The question
is do we need one to validate the credential (that is, provide
authentication services)?  I'd say, no, we don't.  If you get a credential
from a trusted source, we can use p2p mechanisms to validate that
credential.  We don't have to trust another peer; each peer can do the
authentication itself.  In the current environment, the idea of doing a full
credential check with RSA-1024 seems quite reasonable, done at a peer, per
call, as long as it only takes ONE hunk of crypto.  We don't need to
optimize that.  Similarly, we can do that kind of validation before we start
accepting the DHT data exchange messages that manage the distributed hash
table from a peer.  There doesn't seem to be a need for a centralized
mechanism for that either.  

I'm sure there are other mechanisms that would have distributed credential
checking, and they may have advantages over the above.  I don't see any
substantial advantage of your idea over mine, and mine is p2p past the
issuing of the cert, which is a substantial advantage I think.

Yes, I think we have agreed we will use a DHT.  Our charter actually says we
choose one mandatory-to-implement DHT.  I don't think the text actually has
to limit the overlay mechanism to using a DHT, but we'll see how torturous
the text gets if we make it neutral as to mechanism.

Brian

> -----Original Message-----
> From: Frank W. Miller [mailto:fwmiller at cornfed.com]
> Sent: Tuesday, November 28, 2006 1:24 PM
> To: Brian Rosen
> Cc: 'Eunsoo Shim'; p2p-sip at cs.columbia.edu
> Subject: RE: [p2p-sip] Fwd: what's wrong with DNS?
> 
> On Tue, 2006-11-28 at 12:36 -0500, Brian Rosen wrote:
> > I think what you have in mind is that you get a credential that changes
> > every few months: it's the password you use to authenticate.  That
> > credential is validated every time you boot.
> >
> 
> Well, in the proposal I made, there are actually two "credentials."
> There is the password that is issued however you like, its outside the
> scope of the proposal other than the Author/Authen authority being aware
> of it.  Then there is the "cookie."  This is the thing that changes
> frequently.  The password is long lived but the cookie is regenerated
> when the client attaches to the overlay.  Then other UAs use the cookie
> to do authentication.  This proposal isn't all that different than
> current SIP, i.e. A registrar challenges the UA when it registers and
> certainly anyone can challenge a call setup when that happens.
> 
> 
> > The plan I think most of us have in mind is that you get a credential,
> > perhaps something like an x.509 cert, good for a couple of months or
> years,
> > that is validated every time you boot (e.g. connect to the overlay).
> >
> 
> K, so the x.509 cert would be the "password", and the difference would
> be that the "validation" would yield value, i.e. the "cookie", that
> would be used for authentication.  So, you're thinking that you just
> pass the 509 cert around?
> 
> 
> > The difference is you want a centralized service to do the
> authentication,
> > and we want a p2p distributed service.
> >
> 
> Yes, I am a firm believer in a trusted third party acting as an
> authority.
> 
> 
> > Then, you want a centralized authorization service, where we think the
> DHT
> > provides that also as a p2p distributed service.
> >
> 
> I'm leery of this.  At the bottom of that is some peer vouching for some
> other peer.  I don't think thats good enough in many situations.  The
> central authority doesn't have to be Big Brother, it just has to be
> something trusted.  Who knows, that may be one or a group of peers
> acting as the trusted authority in some deployment situations, but it
> could be Big Brother too if thats required.
> 
> Its interesting, you've mentioned the DHT a number of times.  I've been
> following this list for quite some time.  Maybe I missed it but is the
> protocol being done such that the use of a DHT is assumed?
> 
> 
> FM
> 



From slavitch at gmail.com  Tue Nov 28 15:05:22 2006
From: slavitch at gmail.com (Michael Slavitch)
Date: Tue, 28 Nov 2006 15:05:22 -0500
Subject: [p2p-sip] Fwd: what's wrong with DNS?
In-Reply-To: <4C1596FBF66C67478BCFE7B3F81FC1E01751D402DD@DF-MASTIFF-MSG.exchange.corp.microsoft.com>
References: <1164731556.2622.31.camel@localhost.localdomain>
	<088101c71313$b5be1c30$640fa8c0@cis.neustar.com>
	<1fc32c8f0611280953i7d302628i2775ce0405febda0@mail.gmail.com>
	<4C1596FBF66C67478BCFE7B3F81FC1E01751D402DD@DF-MASTIFF-MSG.exchange.corp.microsoft.com>
Message-ID: <1fc32c8f0611281205v4a2bec5fy72d7d50b5abad3d6@mail.gmail.com>

User auth is cached locally on a PC , and can be cached for a long time
without access to the DC, the ticket duration can be tuned.  On connection
to the DC via a VPN or local connection the ticket is refreshed, which
allows a single user login session to be indefinitely long, which is why
login scripts fail for many users: they log in once per reboot.

>From a user perspective they logged in once many months ago with sessions
that are stored via hibernation.

For service tickets, the valid range is from 0 to 99,999 minutes. For user
tickets, the valid range is from 0 to 99,999 hours. A value of zero
effectively turns off expiration. Any other value sets a specific ticket
lifetime. Under the hood tickets get renewed, DC's get contacted, but only
once in a while.  A roaming user can be away for months without a DC
login.

A ticket value of zero is the common ticket length for road warriors for
just this purpose.

The same can be true for P2PSIP of course,  the tickets can be renewed on
contact with supernodes, which are identical to domain controllers at the
data level, with the provisio that their databases are cached.

A domain controller is nothing but a DHT with a hash ratio of 100%, a full
table.

In fact, in my analysis, ongoing moves and changes that are inherent in
Domain Controllers reduce the hash value to 98-99%.  In effect Active
Directory is already a DHT, and a good proof of viability.

Regards

M

On 11/28/06, Sean Olson <seanol at exchange.microsoft.com> wrote:
> A key difference with domain auth is that you do need access to the DC for
most auth needs. Kerb tickets last hours not months. Cert based auth can
give you those longer durations at a cost. We've also implemented other auth
techniques with P2P in Vista.
>
> -----Original Message-----
> From: p2p-sip-bounces at cs.columbia.edu [mailto:
p2p-sip-bounces at cs.columbia.edu] On Behalf Of Michael Slavitch
> Sent: Tuesday, November 28, 2006 9:53 AM
> To: Brian Rosen
> Cc: p2p-sip at cs.columbia.edu
> Subject: Re: [p2p-sip] Fwd: what's wrong with DNS?
>
> Anyone using a Windows notebook that authenticates to a domain is
> using this technology right now.  This is not open research.
>
> On 11/28/06, Brian Rosen <br at brianrosen.net> wrote:
> >
> > I think you are confusing a credential validation from a credential
> > issuance.
> >
> > I think what you have in mind is that you get a credential that changes
> > every few months: it's the password you use to authenticate.  That
> > credential is validated every time you boot.
> >
> > The plan I think most of us have in mind is that you get a credential,
> > perhaps something like an x.509 cert, good for a couple of months or
years,
> > that is validated every time you boot (e.g. connect to the overlay).
> >
> > The difference is you want a centralized service to do the
authentication,
> > and we want a p2p distributed service.
> >
> > Then, you want a centralized authorization service, where we think the
DHT
> > provides that also as a p2p distributed service.
> >
> > Are there other people on the list who think what Frank is advocating is
a
> > good idea?
> >
> > Brian
> >
> >
> > > -----Original Message-----
> > > From: Frank W. Miller [mailto:fwmiller at cornfed.com]
> > > Sent: Tuesday, November 28, 2006 11:33 AM
> > > To: Brian Rosen
> > > Cc: 'Eunsoo Shim'; p2p-sip at cs.columbia.edu
> > > Subject: RE: [p2p-sip] Fwd: what's wrong with DNS?
> > >
> > > On Tue, 2006-11-28 at 10:56 -0500, Brian Rosen wrote:
> > > > You understand that you are proposing a centralized service that
> > > > authenticates you every time you boot, and a centralized
authorization
> > > > service that authorizes you AT CALL TIME, at least the first time
per
> > > boot?
> > > >
> > >
> > > Not exactly.  Authorization happens at boot, i.e when your client
starts
> > > up.  Authentication happens at the first interaction between two UAs.
> > >
> > >
> > > > The direction we are going is that you use the centralized
credentialing
> > > > service to get credentials effectively once in a lifetime of the
> > > credential,
> > > > which might last at least months, and maybe years.  Thereafter, the
> > > > distributed hash table takes care of authentication and
authorization
> > > peer
> > > > to peer without any central service.
> > > >
> > >
> > > Understood.  I question the practicality of this for at least a couple
> > > of reasons.
> > >
> > > First, my view is that credentials should be changed
periodically.  The
> > > longer a credential is available, the more likely it is that it can be
> > > hacked.
> > >
> > > Second, the only real difference is how often the credentials are
issued
> > > and checked.  The DHT approach would do it "once" for all time, my
> > > proposal does it "once" at client boot time and at the beginning of a
> > > UA->UA interaction.  The need or lack of need for a centralized
> > > authority does not go away because of the use of a DHT to lengthen the
> > > time that credentials are valid.
> > >
> > >
> > > > I would definitely not call what you are proposing a peer to peer
> > > service;
> > > > it relies on centralized authentication and centralized
authorization
> > > > services in real time.
> > > >
> > >
> > > Well, its P2P except for initial authorization and subsequent initial
> > > authentications.
> > >
> > >
> > > FM
> > >
> >
> >
> > _______________________________________________
> > p2p-sip mailing list
> > p2p-sip at cs.columbia.edu
> > https://lists.cs.columbia.edu/cucslists/listinfo/p2p-sip
> >
> _______________________________________________
> p2p-sip mailing list
> p2p-sip at cs.columbia.edu
> https://lists.cs.columbia.edu/cucslists/listinfo/p2p-sip
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: https://lists.cs.columbia.edu/pipermail/p2p-sip/attachments/20061128/7b73de0f/attachment.html

From dbarrett at quinthar.com  Tue Nov 28 15:32:45 2006
From: dbarrett at quinthar.com (David Barrett)
Date: Tue, 28 Nov 2006 12:32:45 -0800
Subject: [p2p-sip] Fwd: what's wrong with DNS?
In-Reply-To: <08fe01c7131b$45f46dc0$63e06841@china.huawei.com>
Message-ID: <024b01c7132c$5cf7c920$0200a8c0@Quinthar>

I agree with Spencer, detail is useful for fleshing out these issues.

Regarding the DHT/certificate proposal, am I correct in understanding it
would work something like this?

1) User installs client from any number of vendors
2) User chooses from any number of certificate authorities (eg, "@blah.com")
3) That CA grants the user a unique name (eg, "david at blah.com")
4) The client generates a public/private keypair
5) The client uploads the public key to the CA
6) The CA digitally signs a certificate with its public key saying:
	"The client with public key X is 'david at blah.com'"
7) The client publishes its current IP address in a DHT
8) Some other user types "david at blah.com" into his P2P-SIP client
8) That client contacts "blah.com" to download the CA's public key
10) That client contacts the DHT to find the IP for "david at blah.com" 
11) That client connects to "david at blah.com" directly
12) That client asks "david at blah.com" for its certificate
13) That client validates the certificate's digital signature
14) That client trusts the remote peer is actually "david at blah.com"

Is this about right?  I think this is all great, except for one key usecase:

	What happens if my laptop gets stolen?

There are only two possible answers:

a) Your identity is stolen until your certificate expires (ie, months)
b) The CA publishes a "certificate revocation list" that is checked on call

The first case is just simply unacceptable.  Imagine if you lost your cell
phone and said "Hi, I'd like to prevent whoever stole my phone from using
it" and they said "Sure thing, let's just wait for the old one to expire in
6 months."  Users would reject this in a heartbeat.  I expect they'd
continue to reject it even if it were measured in days; possibly even hours.

The second case reduces this super complex really fancy distributed solution
back down to a simple centralized directory lookup conducted every time you
call someone.  And if we do that, why not just skip all the useless PKI and
just do a simple directory lookup on "@blah.com" (as Frank suggests)?

And if you do *that*, why are we bothering with the DHT -- why not just have
the "@blah.com" server also keep track of my IP address?

Or am I misunderstanding the PKI proposal?

-david

> -----Original Message-----
> From: p2p-sip-bounces at cs.columbia.edu [mailto:p2p-sip-
> bounces at cs.columbia.edu] On Behalf Of Spencer Dawkins
> Sent: Tuesday, November 28, 2006 10:30 AM
> To: p2p-sip at cs.columbia.edu
> Subject: Re: [p2p-sip] Fwd: what's wrong with DNS?
> 
> Without reference to the specific proposal, this is a good example of the
> level of detail we need in order to have discussions that go someplace and
> terminate in finite time - we'll be chartered to produce specifications,
> not
> to fill up working group mailing list archives!
> 
> Frank, thanks for being concrete. The followups show that this posting
> helped people understand what was being proposed.
> 
> Spencer
> 
> > I would propose something like this:
> >
> > There are two elements, User Agents and the Identity Service. Identity
> > refers to messages that are sent between an Internet host and the
> > Identity Service to authorize and authenticate UAIs.  The Identity
> > Service is assumed to be authoritative and trusted, and very likely a
> > centralized service.  Identity operations include the following
> > functions:
> >
> >   1.  Authorization
> >
> >   2.  Authentication
> >
> > Authorization
> >
> > I call the user portion of a SIP URI the User Agent Instance (UAI).
> > The UAI is allocated to the UA by some means outside of the scope of
> this
> > protocol, i.e. by visiting a website or something.  Authorization refers
> > to the registration of a UA instance for a UAI with an Identity Service.
> > UAs are generally authorized when a user wants to begin interacting with
> > other UAs, e.g. when their UA client program is started.  (Note that
> these
> > messages are notional.  They don't have to even be SIP.  The idea of
> > these templates is to list the information moving between the elements)
> >
> >
> >   Frank            Identity Service
> >     |      AUTHORIZE      |
> >     |-------------------->|
> >     |       200 OK        |
> >     |<--------------------|
> >     |                     |
> >
> >
> >   AUTHORIZE
> >   UAI: Frank
> >   Password: ... Frank's password ...
> >   CSeq: 2 AUTHORIZE
> >
> > Frank's UA requests authorization of the UAI.  The request includes a
> > shared secret that was setup when the UAI was issued.  The shared secret
> > must
> > be sent encrypted.
> >
> >
> >   200 OK
> >   UAI: Frank
> >   Authorization: ... Cookie sent to anyone authenticating Frank ...
> >   CSeq: 2 AUTHORIZE
> >
> > Frank's UA is returned a Cookie.  This cookie will be sent to other UAs
> > for subsequent interactions.  Other UAs can use the cookie for
> > authenticating
> > Frank's UA.  Note that this assumes the Authorization cookie is
> encrypted.
> >
> >
> > Authentication
> >
> > Authentication refers to the process of querying the Identity Service
> > regarding a UAI.  The procedure described herein is usually attempted
> > when a UA receives its first request for interaction from another UA.
> >
> >
> >    Bob            Identity Service
> >     |     AUTHENTICATE    |
> >     |-------------------->|
> >     |       200 OK        |
> >     |<--------------------|
> >     |                     |
> >
> >
> >
> >   AUTHENTICATE
> >   UAI: Frank
> >   Password: ... Bob's password ...
> >   CSeq: 2 AUTHORIZE
> >
> > Bob send's an authentication request to the Identity Service.  Bob is
> > only allowed to authenticate Frank if his UA is authorized using the
> > shared secret for Bob's UAI.  The shared secret must be encrypted.
> >
> >
> >   MSP 200 OK
> >   UAI: Frank
> >   Authentication: ... Cookie sent to Frank during authorization ...
> >   CSeq: 2 AUTHORIZE
> >
> > The identity service returns the cookie that was sent by Frank with
> > the initial interaction.  Bob's UAI can compare the two and determine
> > if Frank is really Frank.  Again, the cookie must be encrypted.
> >
> > This proposal requires some level of encryption in all these
> > messages.  I am actually in favor of SSL/TLS being required for
> > transport of these messages, whether that is for this separate
> > protocol blurb or for SIP in general.
> >
> > The management of UAIs for each p2p cloud becomes the authoritative
> > information for that cloud.  I can imagine services like FWD providing
> > it for free and I can also imagine commercial services charging for
> > the service.
> >
> > FM
> >
> >
> >
> > On Tue, 2006-11-28 at 09:30 -0500, Eunsoo Shim wrote:
> >> I am not sure what your proposal is.
> >> Could you provide more specifics on your proposal?
> >> For example, are you suggesting that we use existing DNS servers to
> >> store whatever we want to store at the P2P(SIP) network?
> >> Or are you suggesting that we use the DNS PROTOCOL as (part of) the
> >> protocol for P2P SIP and build a separate (global or ad-hoc) DNS-like
> >> infrastructure?
> >> Thanks.
> >>
> >> Eunsoo
> >>
> >> On 11/28/06, David Barrett <dbarrett at quinthar.com> wrote:
> >>         I think Matthew highlights the Achilles heel of so-called
> >>         "pure" P2P: it
> >>         never is.  In the real world, there's always a central service
> >>         behind the
> >>         curtain:
> >>
> >>         - In Skype's case, it's a central directory at a hard-coded IP
> >>         address.
> >>
> >>         - In P2P-SIP's case (if Frank and Brian are right) is some
> >>         kind of
> >>         centralized authorization/authentication/bootstrapping
> >>         service.
> >>
> >>         So granted, there's the question of "how do you find this
> >>         service?"  The
> >>         obvious choice favored by every other application on the
> >>         internet (and most
> >>         private networks) is "DNS".
> >>
> >>         But ignoring that, there's an even more important question:
> >>
> >>                 "Who controls this service?"
> >>
> >>         If we build some fancy DHT solution we imply "some unnamed
> >>         benevolent
> >>         dictator yet to be appointed" will rule us all.  We've sorta
> >>         assumed this
> >>         (or similarly vague explanations) for the entire course of the
> >>         P2P-SIP
> >>         discussion.  For a long while I was not only on board, I was
> >>         promoting this
> >>         "let's solve security later" approach.
> >>
> >>         But then Adam proposed a different owner, the IANA.  And the
> >>         way to do that
> >>         is through the existing, tried-and-true interface of DNS
> >>         (specifically,
> >>         dynamic DNS).  I think this is obvious and brilliant.
> >>
> >>         Building a nonpartial, internationally represented, neutral
> >>         global third
> >>         party isn't easy, and while the IANA hasn't wildly succeeded
> >>         in this,
> >>         they're closer than anybody else (and decades ahead of
> >>         anything we'd invent
> >>         while searching for our benevolent dictator).
> >>
> >>
> >>         Now I'll grant DNS is limited to only, say, 99.999% of
> >>         users.  And I
> >>         understand that there's a lot of interest in how the remaining
> >>         0.001% of
> >>         users will work without it.  But I'm having a hard time
> >>         understanding why
> >>         we're even discussing them.  Indeed, Matthew's repeated calls
> >>         for a
> >>         real-world use case of a network that doesn't already have,
> >>         won't quickly
> >>         get, or couldn't easily be configured to have DNS has remained
> >>         unanswered.
> >>         (And I'd add to that filter "outside of loud talking range.")
> >>
> >>         Furthermore, I'll grant you could build a DHT layer that
> >>         serves as a
> >>         read-through cache of the central security mechanism that
> >>         provides a faster
> >>         response and propagation time than DNS.  But Akamai, UltraDNS,
> >>         and other
> >>         massive, global organizations have proven that DNS can
> >>         propagate pretty dang
> >>         quick.
> >>
> >>         And while I think it's an excellent, open question on how to
> >>         improve DNS to
> >>         make it even faster (perhaps with some kind of P2P layer), or
> >>         how to make it
> >>         work better in partially or totally disconnected scenarios, or
> >>         how to make
> >>         it more resistant to spamming/phishing and so forth, that
> >>         seems like a much
> >>         bigger problem than SIP alone and thus better suited to its
> >>         own group.
> >>
> >>         And finally, I agree there are open questions as to how my
> >>         car-of-the-future
> >>         will talk to my garage door and sprinkler system.  But why is
> >>         this relevant?
> >>         I share Adam's concern that we're dreaming up a panacea to
> >>         every conceivable
> >>         problem rather than standardizing a solution to problems that
> >>         actually exist
> >>         with technologies that are already at hand.
> >>
> >>
> >>         I apologize for being obtuse, but dynamic DNS (even with
> >>         today's
> >>         limitations, and perhaps with future improvements) seems like
> >>         such an
> >>         obvious way to solve the distributed name resolution problem
> >>         that I'm
> >>         astonished it's being discounted so easily in favor of complex
> >>         solutions to
> >>         nebulous problems experienced by nonexistent userbases.
> >>
> >>         I'll admit the conclusion is anticlimactic -- I'm a P2P
> >>         developer who likes
> >>         high tech as much as anybody.  But where has my reasoning
> >>         fallen off track?
> >>
> >>         -david
> >>
> >>         > -----Original Message-----
> >>         > From: p2p-sip-bounces at cs.columbia.edu [mailto:p2p-sip-
> >>         > bounces at cs.columbia.edu] On Behalf Of Matthew Kaufman
> >>         > Sent: Monday, November 27, 2006 10:18 AM
> >>         > To: hsinnrei at adobe.com; slavitch at gmail.com
> >>         > Cc: p2p-sip at cs.columbia.edu
> >>         > Subject: Re: [p2p-sip] Fwd: what's wrong with DNS?
> >>         >
> >>         > Michael Slavitch:
> >>         > > As a form of assessing identity DNS is particularly weak
> >>         and requires
> >>         > > considerable care and expert feeding as provisioning is
> >>         always
> >>         > > ultimately human-based requiring some kind of authority.
> >>         >
> >>         > Agreed. Though in small disconnected networks, zeroconf
> >>         using mDNS avoids
> >>         > much of this, and in large networks the IT guys or your ISP
> >>         manage the
> >>         > complexity for you (though they haven't solved the security
> >>         problem yet)
> >>         >
> >>         > > A self-organizing automated system needs no such
> >>         crutches.
> >>         >
> >>         > Except for bootstrapping to get the first two nodes talking.
> >>         Which, if on
> >>         > the Internet, means a way to find IP addresses of those
> >>         nodes. Which, on
> >>         > today's Internet, could use DNS. Or, as the Skype folks do,
> >>         hardcoded
> >>         > bootstrap
> >>         > addresses. But either way, bootstrap is such a crutch.
> >>         >
> >>         > > A key provision of P2PSIP is to eliminate the need for
> >>         such
> >>         > > authorities in the first place because they are
> >>         superfluous for the
> >>         > > application.  An extended social network needs no central
> >>         domain
> >>         > > administrator nor a central certificate authority.   It
> >>         needs at most
> >>         > > a human facilitator not an engineer.
> >>         >
> >>         > So P2PSIP includes a web-of-trust security model? If so, why
> >>         not leverage
> >>         > existing solutions for that as well (like PGP)?
> >>         >
> >>         > > Such social networks are not trite. For example social
> >>         networks are
> >>         > > the basis of all peer reviews for medical
> >>         journals.  Shouting doesn't
> >>         > > work when 500 doctors are to be consulted, on several
> >>         continents, with
> >>         > > varied access to the Internet, and over time.  The person
> >>         running the
> >>         > > peer review is a journal editor, a person who uses
> >>         end-user tools, not
> >>         > > a network engineer.  DNS isn't the selection criteria
> >>         here, reputation
> >>         > > is.
> >>         >
> >>         > But DNS is how the journal's mailing list manager finds the
> >>         mail hosts
> >>         > for each of the 500 doctors. Delivery of the article and
> >>         reply comments
> >>         > across several continents over time works today just fine
> >>         without P2P
> >>         > or DHTs or reputation-based web-of-trust security models.
> >>         See this mailing
> >>         > list as an existence proof.
> >>         >
> >>         > Matthew Kaufman
> >>         > matthew at matthew.at
> >>         >
> >>         > _______________________________________________
> >>         > p2p-sip mailing list
> >>         > p2p-sip at cs.columbia.edu
> >>         > https://lists.cs.columbia.edu/cucslists/listinfo/p2p-sip
> >>
> >>         _______________________________________________
> >>         p2p-sip mailing list
> >>         p2p-sip at cs.columbia.edu
> >>         https://lists.cs.columbia.edu/cucslists/listinfo/p2p-sip
> >>
> >> _______________________________________________
> >> p2p-sip mailing list
> >> p2p-sip at cs.columbia.edu
> >> https://lists.cs.columbia.edu/cucslists/listinfo/p2p-sip
> >
> > _______________________________________________
> > p2p-sip mailing list
> > p2p-sip at cs.columbia.edu
> > https://lists.cs.columbia.edu/cucslists/listinfo/p2p-sip
> >
> 
> 
> _______________________________________________
> p2p-sip mailing list
> p2p-sip at cs.columbia.edu
> https://lists.cs.columbia.edu/cucslists/listinfo/p2p-sip


From slavitch at gmail.com  Tue Nov 28 15:36:47 2006
From: slavitch at gmail.com (Michael Slavitch)
Date: Tue, 28 Nov 2006 15:36:47 -0500
Subject: [p2p-sip] Fwd: what's wrong with DNS?
In-Reply-To: <024b01c7132c$5cf7c920$0200a8c0@Quinthar>
References: <08fe01c7131b$45f46dc0$63e06841@china.huawei.com>
	<024b01c7132c$5cf7c920$0200a8c0@Quinthar>
Message-ID: <1fc32c8f0611281236wa0cf919pa98beb811b1a7601@mail.gmail.com>

Again using Windows as the use-case, you  encrypt the file system and have a
strong password.

Both file system and access control depend on a strong password.

In fact this is exactly how it done in their new mobile operating system,
the implications being that it would be a strong vessel for P2PSIP.

Regards

M

On 11/28/06, David Barrett <dbarrett at quinthar.com> wrote:
>
> I agree with Spencer, detail is useful for fleshing out these issues.
>
> Regarding the DHT/certificate proposal, am I correct in understanding it
> would work something like this?
>
> 1) User installs client from any number of vendors
> 2) User chooses from any number of certificate authorities (eg, "@blah.com
> ")
> 3) That CA grants the user a unique name (eg, "david at blah.com")
> 4) The client generates a public/private keypair
> 5) The client uploads the public key to the CA
> 6) The CA digitally signs a certificate with its public key saying:
>         "The client with public key X is 'david at blah.com'"
> 7) The client publishes its current IP address in a DHT
> 8) Some other user types "david at blah.com" into his P2P-SIP client
> 8) That client contacts "blah.com" to download the CA's public key
> 10) That client contacts the DHT to find the IP for "david at blah.com"
> 11) That client connects to "david at blah.com" directly
> 12) That client asks "david at blah.com" for its certificate
> 13) That client validates the certificate's digital signature
> 14) That client trusts the remote peer is actually "david at blah.com"
>
> Is this about right?  I think this is all great, except for one key
> usecase:
>
>         What happens if my laptop gets stolen?
>
> There are only two possible answers:
>
> a) Your identity is stolen until your certificate expires (ie, months)
> b) The CA publishes a "certificate revocation list" that is checked on
> call
>
> The first case is just simply unacceptable.  Imagine if you lost your cell
> phone and said "Hi, I'd like to prevent whoever stole my phone from using
> it" and they said "Sure thing, let's just wait for the old one to expire
> in
> 6 months."  Users would reject this in a heartbeat.  I expect they'd
> continue to reject it even if it were measured in days; possibly even
> hours.
>
> The second case reduces this super complex really fancy distributed
> solution
> back down to a simple centralized directory lookup conducted every time
> you
> call someone.  And if we do that, why not just skip all the useless PKI
> and
> just do a simple directory lookup on "@blah.com" (as Frank suggests)?
>
> And if you do *that*, why are we bothering with the DHT -- why not just
> have
> the "@blah.com" server also keep track of my IP address?
>
> Or am I misunderstanding the PKI proposal?
>
> -david
>
> > -----Original Message-----
> > From: p2p-sip-bounces at cs.columbia.edu [mailto:p2p-sip-
> > bounces at cs.columbia.edu] On Behalf Of Spencer Dawkins
> > Sent: Tuesday, November 28, 2006 10:30 AM
> > To: p2p-sip at cs.columbia.edu
> > Subject: Re: [p2p-sip] Fwd: what's wrong with DNS?
> >
> > Without reference to the specific proposal, this is a good example of
> the
> > level of detail we need in order to have discussions that go someplace
> and
> > terminate in finite time - we'll be chartered to produce specifications,
> > not
> > to fill up working group mailing list archives!
> >
> > Frank, thanks for being concrete. The followups show that this posting
> > helped people understand what was being proposed.
> >
> > Spencer
> >
> > > I would propose something like this:
> > >
> > > There are two elements, User Agents and the Identity Service. Identity
> > > refers to messages that are sent between an Internet host and the
> > > Identity Service to authorize and authenticate UAIs.  The Identity
> > > Service is assumed to be authoritative and trusted, and very likely a
> > > centralized service.  Identity operations include the following
> > > functions:
> > >
> > >   1.  Authorization
> > >
> > >   2.  Authentication
> > >
> > > Authorization
> > >
> > > I call the user portion of a SIP URI the User Agent Instance (UAI).
> > > The UAI is allocated to the UA by some means outside of the scope of
> > this
> > > protocol, i.e. by visiting a website or something.  Authorization
> refers
> > > to the registration of a UA instance for a UAI with an Identity
> Service.
> > > UAs are generally authorized when a user wants to begin interacting
> with
> > > other UAs, e.g. when their UA client program is started.  (Note that
> > these
> > > messages are notional.  They don't have to even be SIP.  The idea of
> > > these templates is to list the information moving between the
> elements)
> > >
> > >
> > >   Frank            Identity Service
> > >     |      AUTHORIZE      |
> > >     |-------------------->|
> > >     |       200 OK        |
> > >     |<--------------------|
> > >     |                     |
> > >
> > >
> > >   AUTHORIZE
> > >   UAI: Frank
> > >   Password: ... Frank's password ...
> > >   CSeq: 2 AUTHORIZE
> > >
> > > Frank's UA requests authorization of the UAI.  The request includes a
> > > shared secret that was setup when the UAI was issued.  The shared
> secret
> > > must
> > > be sent encrypted.
> > >
> > >
> > >   200 OK
> > >   UAI: Frank
> > >   Authorization: ... Cookie sent to anyone authenticating Frank ...
> > >   CSeq: 2 AUTHORIZE
> > >
> > > Frank's UA is returned a Cookie.  This cookie will be sent to other
> UAs
> > > for subsequent interactions.  Other UAs can use the cookie for
> > > authenticating
> > > Frank's UA.  Note that this assumes the Authorization cookie is
> > encrypted.
> > >
> > >
> > > Authentication
> > >
> > > Authentication refers to the process of querying the Identity Service
> > > regarding a UAI.  The procedure described herein is usually attempted
> > > when a UA receives its first request for interaction from another UA.
> > >
> > >
> > >    Bob            Identity Service
> > >     |     AUTHENTICATE    |
> > >     |-------------------->|
> > >     |       200 OK        |
> > >     |<--------------------|
> > >     |                     |
> > >
> > >
> > >
> > >   AUTHENTICATE
> > >   UAI: Frank
> > >   Password: ... Bob's password ...
> > >   CSeq: 2 AUTHORIZE
> > >
> > > Bob send's an authentication request to the Identity Service.  Bob is
> > > only allowed to authenticate Frank if his UA is authorized using the
> > > shared secret for Bob's UAI.  The shared secret must be encrypted.
> > >
> > >
> > >   MSP 200 OK
> > >   UAI: Frank
> > >   Authentication: ... Cookie sent to Frank during authorization ...
> > >   CSeq: 2 AUTHORIZE
> > >
> > > The identity service returns the cookie that was sent by Frank with
> > > the initial interaction.  Bob's UAI can compare the two and determine
> > > if Frank is really Frank.  Again, the cookie must be encrypted.
> > >
> > > This proposal requires some level of encryption in all these
> > > messages.  I am actually in favor of SSL/TLS being required for
> > > transport of these messages, whether that is for this separate
> > > protocol blurb or for SIP in general.
> > >
> > > The management of UAIs for each p2p cloud becomes the authoritative
> > > information for that cloud.  I can imagine services like FWD providing
> > > it for free and I can also imagine commercial services charging for
> > > the service.
> > >
> > > FM
> > >
> > >
> > >
> > > On Tue, 2006-11-28 at 09:30 -0500, Eunsoo Shim wrote:
> > >> I am not sure what your proposal is.
> > >> Could you provide more specifics on your proposal?
> > >> For example, are you suggesting that we use existing DNS servers to
> > >> store whatever we want to store at the P2P(SIP) network?
> > >> Or are you suggesting that we use the DNS PROTOCOL as (part of) the
> > >> protocol for P2P SIP and build a separate (global or ad-hoc) DNS-like
> > >> infrastructure?
> > >> Thanks.
> > >>
> > >> Eunsoo
> > >>
> > >> On 11/28/06, David Barrett <dbarrett at quinthar.com> wrote:
> > >>         I think Matthew highlights the Achilles heel of so-called
> > >>         "pure" P2P: it
> > >>         never is.  In the real world, there's always a central
> service
> > >>         behind the
> > >>         curtain:
> > >>
> > >>         - In Skype's case, it's a central directory at a hard-coded
> IP
> > >>         address.
> > >>
> > >>         - In P2P-SIP's case (if Frank and Brian are right) is some
> > >>         kind of
> > >>         centralized authorization/authentication/bootstrapping
> > >>         service.
> > >>
> > >>         So granted, there's the question of "how do you find this
> > >>         service?"  The
> > >>         obvious choice favored by every other application on the
> > >>         internet (and most
> > >>         private networks) is "DNS".
> > >>
> > >>         But ignoring that, there's an even more important question:
> > >>
> > >>                 "Who controls this service?"
> > >>
> > >>         If we build some fancy DHT solution we imply "some unnamed
> > >>         benevolent
> > >>         dictator yet to be appointed" will rule us all.  We've sorta
> > >>         assumed this
> > >>         (or similarly vague explanations) for the entire course of
> the
> > >>         P2P-SIP
> > >>         discussion.  For a long while I was not only on board, I was
> > >>         promoting this
> > >>         "let's solve security later" approach.
> > >>
> > >>         But then Adam proposed a different owner, the IANA.  And the
> > >>         way to do that
> > >>         is through the existing, tried-and-true interface of DNS
> > >>         (specifically,
> > >>         dynamic DNS).  I think this is obvious and brilliant.
> > >>
> > >>         Building a nonpartial, internationally represented, neutral
> > >>         global third
> > >>         party isn't easy, and while the IANA hasn't wildly succeeded
> > >>         in this,
> > >>         they're closer than anybody else (and decades ahead of
> > >>         anything we'd invent
> > >>         while searching for our benevolent dictator).
> > >>
> > >>
> > >>         Now I'll grant DNS is limited to only, say, 99.999% of
> > >>         users.  And I
> > >>         understand that there's a lot of interest in how the
> remaining
> > >>         0.001% of
> > >>         users will work without it.  But I'm having a hard time
> > >>         understanding why
> > >>         we're even discussing them.  Indeed, Matthew's repeated calls
> > >>         for a
> > >>         real-world use case of a network that doesn't already have,
> > >>         won't quickly
> > >>         get, or couldn't easily be configured to have DNS has
> remained
> > >>         unanswered.
> > >>         (And I'd add to that filter "outside of loud talking range.")
> > >>
> > >>         Furthermore, I'll grant you could build a DHT layer that
> > >>         serves as a
> > >>         read-through cache of the central security mechanism that
> > >>         provides a faster
> > >>         response and propagation time than DNS.  But Akamai,
> UltraDNS,
> > >>         and other
> > >>         massive, global organizations have proven that DNS can
> > >>         propagate pretty dang
> > >>         quick.
> > >>
> > >>         And while I think it's an excellent, open question on how to
> > >>         improve DNS to
> > >>         make it even faster (perhaps with some kind of P2P layer), or
> > >>         how to make it
> > >>         work better in partially or totally disconnected scenarios,
> or
> > >>         how to make
> > >>         it more resistant to spamming/phishing and so forth, that
> > >>         seems like a much
> > >>         bigger problem than SIP alone and thus better suited to its
> > >>         own group.
> > >>
> > >>         And finally, I agree there are open questions as to how my
> > >>         car-of-the-future
> > >>         will talk to my garage door and sprinkler system.  But why is
> > >>         this relevant?
> > >>         I share Adam's concern that we're dreaming up a panacea to
> > >>         every conceivable
> > >>         problem rather than standardizing a solution to problems that
> > >>         actually exist
> > >>         with technologies that are already at hand.
> > >>
> > >>
> > >>         I apologize for being obtuse, but dynamic DNS (even with
> > >>         today's
> > >>         limitations, and perhaps with future improvements) seems like
> > >>         such an
> > >>         obvious way to solve the distributed name resolution problem
> > >>         that I'm
> > >>         astonished it's being discounted so easily in favor of
> complex
> > >>         solutions to
> > >>         nebulous problems experienced by nonexistent userbases.
> > >>
> > >>         I'll admit the conclusion is anticlimactic -- I'm a P2P
> > >>         developer who likes
> > >>         high tech as much as anybody.  But where has my reasoning
> > >>         fallen off track?
> > >>
> > >>         -david
> > >>
> > >>         > -----Original Message-----
> > >>         > From: p2p-sip-bounces at cs.columbia.edu [mailto:p2p-sip-
> > >>         > bounces at cs.columbia.edu] On Behalf Of Matthew Kaufman
> > >>         > Sent: Monday, November 27, 2006 10:18 AM
> > >>         > To: hsinnrei at adobe.com; slavitch at gmail.com
> > >>         > Cc: p2p-sip at cs.columbia.edu
> > >>         > Subject: Re: [p2p-sip] Fwd: what's wrong with DNS?
> > >>         >
> > >>         > Michael Slavitch:
> > >>         > > As a form of assessing identity DNS is particularly weak
> > >>         and requires
> > >>         > > considerable care and expert feeding as provisioning is
> > >>         always
> > >>         > > ultimately human-based requiring some kind of authority.
> > >>         >
> > >>         > Agreed. Though in small disconnected networks, zeroconf
> > >>         using mDNS avoids
> > >>         > much of this, and in large networks the IT guys or your ISP
> > >>         manage the
> > >>         > complexity for you (though they haven't solved the security
> > >>         problem yet)
> > >>         >
> > >>         > > A self-organizing automated system needs no such
> > >>         crutches.
> > >>         >
> > >>         > Except for bootstrapping to get the first two nodes
> talking.
> > >>         Which, if on
> > >>         > the Internet, means a way to find IP addresses of those
> > >>         nodes. Which, on
> > >>         > today's Internet, could use DNS. Or, as the Skype folks do,
> > >>         hardcoded
> > >>         > bootstrap
> > >>         > addresses. But either way, bootstrap is such a crutch.
> > >>         >
> > >>         > > A key provision of P2PSIP is to eliminate the need for
> > >>         such
> > >>         > > authorities in the first place because they are
> > >>         superfluous for the
> > >>         > > application.  An extended social network needs no central
> > >>         domain
> > >>         > > administrator nor a central certificate authority.   It
> > >>         needs at most
> > >>         > > a human facilitator not an engineer.
> > >>         >
> > >>         > So P2PSIP includes a web-of-trust security model? If so,
> why
> > >>         not leverage
> > >>         > existing solutions for that as well (like PGP)?
> > >>         >
> > >>         > > Such social networks are not trite. For example social
> > >>         networks are
> > >>         > > the basis of all peer reviews for medical
> > >>         journals.  Shouting doesn't
> > >>         > > work when 500 doctors are to be consulted, on several
> > >>         continents, with
> > >>         > > varied access to the Internet, and over time.  The person
> > >>         running the
> > >>         > > peer review is a journal editor, a person who uses
> > >>         end-user tools, not
> > >>         > > a network engineer.  DNS isn't the selection criteria
> > >>         here, reputation
> > >>         > > is.
> > >>         >
> > >>         > But DNS is how the journal's mailing list manager finds the
> > >>         mail hosts
> > >>         > for each of the 500 doctors. Delivery of the article and
> > >>         reply comments
> > >>         > across several continents over time works today just fine
> > >>         without P2P
> > >>         > or DHTs or reputation-based web-of-trust security models.
> > >>         See this mailing
> > >>         > list as an existence proof.
> > >>         >
> > >>         > Matthew Kaufman
> > >>         > matthew at matthew.at
> > >>         >
> > >>         > _______________________________________________
> > >>         > p2p-sip mailing list
> > >>         > p2p-sip at cs.columbia.edu
> > >>         > https://lists.cs.columbia.edu/cucslists/listinfo/p2p-sip
> > >>
> > >>         _______________________________________________
> > >>         p2p-sip mailing list
> > >>         p2p-sip at cs.columbia.edu
> > >>         https://lists.cs.columbia.edu/cucslists/listinfo/p2p-sip
> > >>
> > >> _______________________________________________
> > >> p2p-sip mailing list
> > >> p2p-sip at cs.columbia.edu
> > >> https://lists.cs.columbia.edu/cucslists/listinfo/p2p-sip
> > >
> > > _______________________________________________
> > > p2p-sip mailing list
> > > p2p-sip at cs.columbia.edu
> > > https://lists.cs.columbia.edu/cucslists/listinfo/p2p-sip
> > >
> >
> >
> > _______________________________________________
> > p2p-sip mailing list
> > p2p-sip at cs.columbia.edu
> > https://lists.cs.columbia.edu/cucslists/listinfo/p2p-sip
>
> _______________________________________________
> p2p-sip mailing list
> p2p-sip at cs.columbia.edu
> https://lists.cs.columbia.edu/cucslists/listinfo/p2p-sip
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: https://lists.cs.columbia.edu/pipermail/p2p-sip/attachments/20061128/3e900e65/attachment-0001.html

From fwmiller at cornfed.com  Tue Nov 28 15:43:31 2006
From: fwmiller at cornfed.com (Frank W. Miller)
Date: Tue, 28 Nov 2006 13:43:31 -0700
Subject: [p2p-sip] Fwd: what's wrong with DNS?
In-Reply-To: <08b301c71321$9fd25a40$640fa8c0@cis.neustar.com>
References: <08b301c71321$9fd25a40$640fa8c0@cis.neustar.com>
Message-ID: <1164746611.2622.67.camel@localhost.localdomain>


> We'll all agree we need a trusted entity to issue credentials.  The question
> is do we need one to validate the credential (that is, provide
> authentication services)?  I'd say, no, we don't.  If you get a credential
> from a trusted source, we can use p2p mechanisms to validate that
> credential.  We don't have to trust another peer; each peer can do the
> authentication itself.  In the current environment, the idea of doing a full
> credential check with RSA-1024 seems quite reasonable, done at a peer, per
> call, as long as it only takes ONE hunk of crypto.  We don't need to
> optimize that.  Similarly, we can do that kind of validation before we start
> accepting the DHT data exchange messages that manage the distributed hash
> table from a peer.  There doesn't seem to be a need for a centralized
> mechanism for that either.  
> 

This is my lack of security background showing through but how does this
actually work?  I mean, when you go to validate, how can you do so
without consulting the trusted entity?

So, lets say some UA wants to set up a call with you.  They send you a
credential and you RSA it. This gets you some other value.  Are you
saying that you already know what the resulting value is supposed to be
so you don't need to consult the authority?  If so, how did you find out
that value?



> Yes, I think we have agreed we will use a DHT.  Our charter actually says we
> choose one mandatory-to-implement DHT.  I don't think the text actually has
> to limit the overlay mechanism to using a DHT, but we'll see how torturous
> the text gets if we make it neutral as to mechanism.
> 


This worries me a little bit (although not hugely).  As an implementor,
I'd like to be able to use some flooding algorithm like Gnutella or
something just cuz its a lot easier to implement.  Is it possible to
just define the interface to the P2P overlay, maybe with some indication
of the algorithm being used internally, and leave it at that?  It seems
to me that DHT algorithms and P2P overlay algorithms in general are
going to evolve going forward.  It would be a shame to tie the protocol
to one unnecessarily.


FM




From slavitch at gmail.com  Tue Nov 28 15:51:43 2006
From: slavitch at gmail.com (Michael Slavitch)
Date: Tue, 28 Nov 2006 15:51:43 -0500
Subject: [p2p-sip] Fwd: what's wrong with DNS?
In-Reply-To: <4C1596FBF66C67478BCFE7B3F81FC1E01751D4039D@DF-MASTIFF-MSG.exchange.corp.microsoft.com>
References: <1164731556.2622.31.camel@localhost.localdomain>
	<088101c71313$b5be1c30$640fa8c0@cis.neustar.com>
	<1fc32c8f0611280953i7d302628i2775ce0405febda0@mail.gmail.com>
	<4C1596FBF66C67478BCFE7B3F81FC1E01751D402DD@DF-MASTIFF-MSG.exchange.corp.microsoft.com>
	<1fc32c8f0611281205v4a2bec5fy72d7d50b5abad3d6@mail.gmail.com>
	<4C1596FBF66C67478BCFE7B3F81FC1E01751D4039D@DF-MASTIFF-MSG.exchange.corp.microsoft.com>
Message-ID: <1fc32c8f0611281251v570854bgbbabd498975514eb@mail.gmail.com>

That is a current architectural limitation of Active Directory, one that
clearly delineates clients and servers as separate entities. Clients do not
normally trust each other. Remove those two restrictions and the need for
servers is eliminated using existing authentication regimes.

Regards

M

On 11/28/06, Sean Olson <seanol at exchange.microsoft.com> wrote:
>
>  User auth caching is for local login.  Kerb ticket caching works when **
> you** cannot access the DC. It won't work for the remote peer in a P2P
> system.
>
>
>
> *From:* Michael Slavitch [mailto:slavitch at gmail.com]
> *Sent:* Tuesday, November 28, 2006 12:05 PM
> *To:* Sean Olson
> *Cc:* Brian Rosen; p2p-sip at cs.columbia.edu
> *Subject:* Re: [p2p-sip] Fwd: what's wrong with DNS?
>
>
>
> User auth is cached locally on a PC , and can be cached for a long time
> without access to the DC, the ticket duration can be tuned.  On connection
> to the DC via a VPN or local connection the ticket is refreshed, which
> allows a single user login session to be indefinitely long, which is why
> login scripts fail for many users: they log in once per reboot.
>
> From a user perspective they logged in once many months ago with sessions
> that are stored via hibernation.
>
> For service tickets, the valid range is from 0 to 99,999 minutes. For user
> tickets, the valid range is from 0 to 99,999 hours. A value of zero
> effectively turns off expiration. Any other value sets a specific ticket
> lifetime. Under the hood tickets get renewed, DC's get contacted, but only
> once in a while.  A roaming user can be away for months without a DC
> login.
>
> A ticket value of zero is the common ticket length for road warriors for
> just this purpose.
>
> The same can be true for P2PSIP of course,  the tickets can be renewed on
> contact with supernodes, which are identical to domain controllers at the
> data level, with the provisio that their databases are cached.
>
> A domain controller is nothing but a DHT with a hash ratio of 100%, a full
> table.
>
> In fact, in my analysis, ongoing moves and changes that are inherent in
> Domain Controllers reduce the hash value to 98-99%.  In effect Active
> Directory is already a DHT, and a good proof of viability.
>
> Regards
>
> M
>
> On 11/28/06, Sean Olson <seanol at exchange.microsoft.com> wrote:
> > A key difference with domain auth is that you do need access to the DC
> for most auth needs. Kerb tickets last hours not months. Cert based auth can
> give you those longer durations at a cost. We've also implemented other auth
> techniques with P2P in Vista.
> >
> > -----Original Message-----
> > From: p2p-sip-bounces at cs.columbia.edu [mailto:p2p-sip-bounces at cs.columbia.edu
> ] On Behalf Of Michael Slavitch
> > Sent: Tuesday, November 28, 2006 9:53 AM
> > To: Brian Rosen
> > Cc: p2p-sip at cs.columbia.edu
> > Subject: Re: [p2p-sip] Fwd: what's wrong with DNS?
> >
> > Anyone using a Windows notebook that authenticates to a domain is
> > using this technology right now.  This is not open research.
> >
> > On 11/28/06, Brian Rosen < br at brianrosen.net> wrote:
> > >
> > > I think you are confusing a credential validation from a credential
> > > issuance.
> > >
> > > I think what you have in mind is that you get a credential that
> changes
> > > every few months: it's the password you use to authenticate.  That
> > > credential is validated every time you boot.
> > >
> > > The plan I think most of us have in mind is that you get a credential,
>
> > > perhaps something like an x.509 cert, good for a couple of months or
> years,
> > > that is validated every time you boot (e.g. connect to the overlay).
> > >
> > > The difference is you want a centralized service to do the
> authentication,
> > > and we want a p2p distributed service.
> > >
> > > Then, you want a centralized authorization service, where we think the
> DHT
> > > provides that also as a p2p distributed service.
> > >
> > > Are there other people on the list who think what Frank is advocating
> is a
> > > good idea?
> > >
> > > Brian
> > >
> > >
> > > > -----Original Message-----
> > > > From: Frank W. Miller [mailto: fwmiller at cornfed.com]
> > > > Sent: Tuesday, November 28, 2006 11:33 AM
> > > > To: Brian Rosen
> > > > Cc: 'Eunsoo Shim'; p2p-sip at cs.columbia.edu
> > > > Subject: RE: [p2p-sip] Fwd: what's wrong with DNS?
> > > >
> > > > On Tue, 2006-11-28 at 10:56 -0500, Brian Rosen wrote:
> > > > > You understand that you are proposing a centralized service that
> > > > > authenticates you every time you boot, and a centralized
> authorization
> > > > > service that authorizes you AT CALL TIME, at least the first time
> per
> > > > boot?
> > > > >
> > > >
> > > > Not exactly.  Authorization happens at boot, i.e when your client
> starts
> > > > up.  Authentication happens at the first interaction between two
> UAs.
> > > >
> > > >
> > > > > The direction we are going is that you use the centralized
> credentialing
> > > > > service to get credentials effectively once in a lifetime of the
> > > > credential,
> > > > > which might last at least months, and maybe years.  Thereafter,
> the
> > > > > distributed hash table takes care of authentication and
> authorization
> > > > peer
> > > > > to peer without any central service.
> > > > >
> > > >
> > > > Understood.  I question the practicality of this for at least a
> couple
> > > > of reasons.
> > > >
> > > > First, my view is that credentials should be changed
> periodically.  The
> > > > longer a credential is available, the more likely it is that it can
> be
> > > > hacked.
> > > >
> > > > Second, the only real difference is how often the credentials are
> issued
> > > > and checked.  The DHT approach would do it "once" for all time, my
> > > > proposal does it "once" at client boot time and at the beginning of
> a
> > > > UA->UA interaction.  The need or lack of need for a centralized
> > > > authority does not go away because of the use of a DHT to lengthen
> the
> > > > time that credentials are valid.
> > > >
> > > >
> > > > > I would definitely not call what you are proposing a peer to peer
> > > > service;
> > > > > it relies on centralized authentication and centralized
> authorization
> > > > > services in real time.
> > > > >
> > > >
> > > > Well, its P2P except for initial authorization and subsequent
> initial
> > > > authentications.
> > > >
> > > >
> > > > FM
> > > >
> > >
> > >
> > > _______________________________________________
> > > p2p-sip mailing list
> > > p2p-sip at cs.columbia.edu
> > > https://lists.cs.columbia.edu/cucslists/listinfo/p2p-sip
> > >
> > _______________________________________________
> > p2p-sip mailing list
> > p2p-sip at cs.columbia.edu
> > https://lists.cs.columbia.edu/cucslists/listinfo/p2p-sip
> >
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: https://lists.cs.columbia.edu/pipermail/p2p-sip/attachments/20061128/b6b8dd33/attachment.html

From spencer at mcsr-labs.org  Tue Nov 28 15:51:31 2006
From: spencer at mcsr-labs.org (Spencer Dawkins)
Date: Tue, 28 Nov 2006 14:51:31 -0600
Subject: [p2p-sip] Fwd: what's wrong with DNS?
References: <08b301c71321$9fd25a40$640fa8c0@cis.neustar.com>
	<1164746611.2622.67.camel@localhost.localdomain>
Message-ID: <09c701c7132e$fc7974b0$63e06841@china.huawei.com>

Hi, Frank,

>> Yes, I think we have agreed we will use a DHT.  Our charter actually says 
>> we
>> choose one mandatory-to-implement DHT.  I don't think the text actually 
>> has
>> to limit the overlay mechanism to using a DHT, but we'll see how 
>> torturous
>> the text gets if we make it neutral as to mechanism.
>>
> This worries me a little bit (although not hugely).  As an implementor,
> I'd like to be able to use some flooding algorithm like Gnutella or
> something just cuz its a lot easier to implement.  Is it possible to
> just define the interface to the P2P overlay, maybe with some indication
> of the algorithm being used internally, and leave it at that?  It seems
> to me that DHT algorithms and P2P overlay algorithms in general are
> going to evolve going forward.  It would be a shame to tie the protocol
> to one unnecessarily.

This is roughly the same situation as crypto algorithms - where the state of 
the art will never be stable.

We want to be "algorithm-agile", so we're not tied to today's state of the 
art forever, but we also select one algorithm as "mandatory-to-implement", 
so there is some chance that you'll be able to communicate with a device 
from another source.

Hope this helps,

Spencer 



From dbarrett at quinthar.com  Tue Nov 28 15:56:05 2006
From: dbarrett at quinthar.com (David Barrett)
Date: Tue, 28 Nov 2006 12:56:05 -0800
Subject: [p2p-sip] Fwd: what's wrong with DNS?
In-Reply-To: <1fc32c8f0611281236wa0cf919pa98beb811b1a7601@mail.gmail.com>
Message-ID: <027a01c7132f$a043b150$0200a8c0@Quinthar>

I'm not sure I understand this answer.  Are you saying that someone stealing
my laptop doesn't matter because my file system is encrypted, and thus they
can't log in as me anyway?  This seems to overstate the prevalence of
encrypted hard drives; I'd wager that nearly 0% of users on a global scale
do this.

 

Are you suggesting that P2P-SIP's security model should assume a totally
encrypted hard drive?

 

And even assuming this, are you saying that if someone steals my laptop
*and* my password, there is no resolution to the problem other than to wait
it out?

 

I'm not sure users will like this.

 

-david

 

  _____  

From: Michael Slavitch [mailto:slavitch at gmail.com] 
Sent: Tuesday, November 28, 2006 12:37 PM
To: David Barrett
Cc: Spencer Dawkins; p2p-sip at cs.columbia.edu
Subject: Re: [p2p-sip] Fwd: what's wrong with DNS?

 

Again using Windows as the use-case, you  encrypt the file system and have a
strong password.

Both file system and access control depend on a strong password.

In fact this is exactly how it done in their new mobile operating system,
the implications being that it would be a strong vessel for P2PSIP. 

Regards

M

On 11/28/06, David Barrett <dbarrett at quinthar.com> wrote:

I agree with Spencer, detail is useful for fleshing out these issues.

Regarding the DHT/certificate proposal, am I correct in understanding it
would work something like this?

1) User installs client from any number of vendors 
2) User chooses from any number of certificate authorities (eg, "@blah.com")
3) That CA grants the user a unique name (eg, "david at blah.com
<mailto:david at blah.com> ")
4) The client generates a public/private keypair
5) The client uploads the public key to the CA
6) The CA digitally signs a certificate with its public key saying:
        "The client with public key X is ' david at blah.com'"
7) The client publishes its current IP address in a DHT
8) Some other user types "david at blah.com" into his P2P-SIP client 
8) That client contacts "blah.com" to download the CA's public key
10) That client contacts the DHT to find the IP for "david at blah.com "
11) That client connects to "david at blah.com" directly
12) That client asks "david at blah.com" for its certificate
13) That client validates the certificate's digital signature 
14) That client trusts the remote peer is actually "david at blah.com"

Is this about right?  I think this is all great, except for one key usecase:

        What happens if my laptop gets stolen? 

There are only two possible answers:

a) Your identity is stolen until your certificate expires (ie, months)
b) The CA publishes a "certificate revocation list" that is checked on call

The first case is just simply unacceptable.  Imagine if you lost your cell 
phone and said "Hi, I'd like to prevent whoever stole my phone from using
it" and they said "Sure thing, let's just wait for the old one to expire in
6 months."  Users would reject this in a heartbeat.  I expect they'd 
continue to reject it even if it were measured in days; possibly even hours.

The second case reduces this super complex really fancy distributed solution
back down to a simple centralized directory lookup conducted every time you 
call someone.  And if we do that, why not just skip all the useless PKI and
just do a simple directory lookup on "@blah.com" (as Frank suggests)?

And if you do *that*, why are we bothering with the DHT -- why not just have

the "@blah.com" server also keep track of my IP address?

Or am I misunderstanding the PKI proposal?

-david

> -----Original Message-----
> From: p2p-sip-bounces at cs.columbia.edu [mailto:p2p-sip-
> bounces at cs.columbia.edu] On Behalf Of Spencer Dawkins
> Sent: Tuesday, November 28, 2006 10:30 AM 
> To: p2p-sip at cs.columbia.edu
> Subject: Re: [p2p-sip] Fwd: what's wrong with DNS?
>
> Without reference to the specific proposal, this is a good example of the 
> level of detail we need in order to have discussions that go someplace and
> terminate in finite time - we'll be chartered to produce specifications,
> not
> to fill up working group mailing list archives! 
>
> Frank, thanks for being concrete. The followups show that this posting
> helped people understand what was being proposed.
>
> Spencer
>
> > I would propose something like this: 
> >
> > There are two elements, User Agents and the Identity Service. Identity
> > refers to messages that are sent between an Internet host and the
> > Identity Service to authorize and authenticate UAIs.  The Identity 
> > Service is assumed to be authoritative and trusted, and very likely a
> > centralized service.  Identity operations include the following
> > functions:
> >
> >   1.  Authorization 
> >
> >   2.  Authentication
> >
> > Authorization
> >
> > I call the user portion of a SIP URI the User Agent Instance (UAI).
> > The UAI is allocated to the UA by some means outside of the scope of 
> this
> > protocol, i.e. by visiting a website or something.  Authorization refers
> > to the registration of a UA instance for a UAI with an Identity Service.
> > UAs are generally authorized when a user wants to begin interacting with

> > other UAs, e.g. when their UA client program is started.  (Note that
> these
> > messages are notional.  They don't have to even be SIP.  The idea of
> > these templates is to list the information moving between the elements) 
> >
> >
> >   Frank            Identity Service
> >     |      AUTHORIZE      |
> >     |-------------------->|
> >     |       200 OK        |
> >     |<--------------------| 
> >     |                     |
> >
> >
> >   AUTHORIZE
> >   UAI: Frank
> >   Password: ... Frank's password ...
> >   CSeq: 2 AUTHORIZE
> >
> > Frank's UA requests authorization of the UAI.  The request includes a 
> > shared secret that was setup when the UAI was issued.  The shared secret
> > must
> > be sent encrypted.
> >
> >
> >   200 OK
> >   UAI: Frank
> >   Authorization: ... Cookie sent to anyone authenticating Frank ... 
> >   CSeq: 2 AUTHORIZE
> >
> > Frank's UA is returned a Cookie.  This cookie will be sent to other UAs
> > for subsequent interactions.  Other UAs can use the cookie for
> > authenticating 
> > Frank's UA.  Note that this assumes the Authorization cookie is
> encrypted.
> >
> >
> > Authentication
> >
> > Authentication refers to the process of querying the Identity Service 
> > regarding a UAI.  The procedure described herein is usually attempted
> > when a UA receives its first request for interaction from another UA.
> >
> >
> >    Bob            Identity Service 
> >     |     AUTHENTICATE    |
> >     |-------------------->|
> >     |       200 OK        |
> >     |<--------------------|
> >     |                     |
> > 
> >
> >
> >   AUTHENTICATE
> >   UAI: Frank
> >   Password: ... Bob's password ...
> >   CSeq: 2 AUTHORIZE
> >
> > Bob send's an authentication request to the Identity Service.  Bob is 
> > only allowed to authenticate Frank if his UA is authorized using the
> > shared secret for Bob's UAI.  The shared secret must be encrypted.
> >
> >
> >   MSP 200 OK
> >   UAI: Frank 
> >   Authentication: ... Cookie sent to Frank during authorization ...
> >   CSeq: 2 AUTHORIZE
> >
> > The identity service returns the cookie that was sent by Frank with
> > the initial interaction.  Bob's UAI can compare the two and determine 
> > if Frank is really Frank.  Again, the cookie must be encrypted.
> >
> > This proposal requires some level of encryption in all these
> > messages.  I am actually in favor of SSL/TLS being required for 
> > transport of these messages, whether that is for this separate
> > protocol blurb or for SIP in general.
> >
> > The management of UAIs for each p2p cloud becomes the authoritative 
> > information for that cloud.  I can imagine services like FWD providing
> > it for free and I can also imagine commercial services charging for
> > the service.
> >
> > FM
> >
> >
> >
> > On Tue, 2006-11-28 at 09:30 -0500, Eunsoo Shim wrote:
> >> I am not sure what your proposal is.
> >> Could you provide more specifics on your proposal? 
> >> For example, are you suggesting that we use existing DNS servers to
> >> store whatever we want to store at the P2P(SIP) network?
> >> Or are you suggesting that we use the DNS PROTOCOL as (part of) the 
> >> protocol for P2P SIP and build a separate (global or ad-hoc) DNS-like
> >> infrastructure?
> >> Thanks.
> >>
> >> Eunsoo
> >>
> >> On 11/28/06, David Barrett < dbarrett at quinthar.com> wrote:
> >>         I think Matthew highlights the Achilles heel of so-called
> >>         "pure" P2P: it
> >>         never is.  In the real world, there's always a central service 
> >>         behind the
> >>         curtain:
> >>
> >>         - In Skype's case, it's a central directory at a hard-coded IP
> >>         address.
> >> 
> >>         - In P2P-SIP's case (if Frank and Brian are right) is some
> >>         kind of
> >>         centralized authorization/authentication/bootstrapping
> >>         service. 
> >>
> >>         So granted, there's the question of "how do you find this
> >>         service?"  The
> >>         obvious choice favored by every other application on the 
> >>         internet (and most
> >>         private networks) is "DNS".
> >>
> >>         But ignoring that, there's an even more important question:
> >> 
> >>                 "Who controls this service?"
> >>
> >>         If we build some fancy DHT solution we imply "some unnamed
> >>         benevolent
> >>         dictator yet to be appointed" will rule us all.  We've sorta 
> >>         assumed this
> >>         (or similarly vague explanations) for the entire course of the
> >>         P2P-SIP
> >>         discussion.  For a long while I was not only on board, I was 
> >>         promoting this
> >>         "let's solve security later" approach.
> >>
> >>         But then Adam proposed a different owner, the IANA.  And the
> >>         way to do that
> >>         is through the existing, tried-and-true interface of DNS
> >>         (specifically,
> >>         dynamic DNS).  I think this is obvious and brilliant. 
> >>
> >>         Building a nonpartial, internationally represented, neutral
> >>         global third
> >>         party isn't easy, and while the IANA hasn't wildly succeeded 
> >>         in this,
> >>         they're closer than anybody else (and decades ahead of
> >>         anything we'd invent
> >>         while searching for our benevolent dictator). 
> >>
> >>
> >>         Now I'll grant DNS is limited to only, say, 99.999% of
> >>         users.  And I
> >>         understand that there's a lot of interest in how the remaining 
> >>         0.001% of
> >>         users will work without it.  But I'm having a hard time
> >>         understanding why
> >>         we're even discussing them.  Indeed, Matthew's repeated calls 
> >>         for a
> >>         real-world use case of a network that doesn't already have,
> >>         won't quickly
> >>         get, or couldn't easily be configured to have DNS has remained 
> >>         unanswered.
> >>         (And I'd add to that filter "outside of loud talking range.")
> >>
> >>         Furthermore, I'll grant you could build a DHT layer that 
> >>         serves as a
> >>         read-through cache of the central security mechanism that
> >>         provides a faster
> >>         response and propagation time than DNS.  But Akamai, UltraDNS, 
> >>         and other
> >>         massive, global organizations have proven that DNS can
> >>         propagate pretty dang
> >>         quick.
> >>
> >>         And while I think it's an excellent, open question on how to 
> >>         improve DNS to
> >>         make it even faster (perhaps with some kind of P2P layer), or
> >>         how to make it
> >>         work better in partially or totally disconnected scenarios, or 
> >>         how to make
> >>         it more resistant to spamming/phishing and so forth, that
> >>         seems like a much
> >>         bigger problem than SIP alone and thus better suited to its 
> >>         own group.
> >>
> >>         And finally, I agree there are open questions as to how my
> >>         car-of-the-future
> >>         will talk to my garage door and sprinkler system.  But why is 
> >>         this relevant?
> >>         I share Adam's concern that we're dreaming up a panacea to
> >>         every conceivable
> >>         problem rather than standardizing a solution to problems that 
> >>         actually exist
> >>         with technologies that are already at hand.
> >>
> >>
> >>         I apologize for being obtuse, but dynamic DNS (even with 
> >>         today's
> >>         limitations, and perhaps with future improvements) seems like
> >>         such an
> >>         obvious way to solve the distributed name resolution problem 
> >>         that I'm
> >>         astonished it's being discounted so easily in favor of complex
> >>         solutions to
> >>         nebulous problems experienced by nonexistent userbases. 
> >>
> >>         I'll admit the conclusion is anticlimactic -- I'm a P2P
> >>         developer who likes
> >>         high tech as much as anybody.  But where has my reasoning 
> >>         fallen off track?
> >>
> >>         -david
> >>
> >>         > -----Original Message-----
> >>         > From: p2p-sip-bounces at cs.columbia.edu [mailto:p2p-sip-
> >>         > bounces at cs.columbia.edu] On Behalf Of Matthew Kaufman
> >>         > Sent: Monday, November 27, 2006 10:18 AM 
> >>         > To: hsinnrei at adobe.com; slavitch at gmail.com
> >>         > Cc: p2p-sip at cs.columbia.edu
> >>         > Subject: Re: [p2p-sip] Fwd: what's wrong with DNS?
> >>         >
> >>         > Michael Slavitch:
> >>         > > As a form of assessing identity DNS is particularly weak 
> >>         and requires
> >>         > > considerable care and expert feeding as provisioning is
> >>         always
> >>         > > ultimately human-based requiring some kind of authority. 
> >>         >
> >>         > Agreed. Though in small disconnected networks, zeroconf
> >>         using mDNS avoids
> >>         > much of this, and in large networks the IT guys or your ISP 
> >>         manage the
> >>         > complexity for you (though they haven't solved the security
> >>         problem yet)
> >>         >
> >>         > > A self-organizing automated system needs no such 
> >>         crutches.
> >>         >
> >>         > Except for bootstrapping to get the first two nodes talking.
> >>         Which, if on
> >>         > the Internet, means a way to find IP addresses of those 
> >>         nodes. Which, on
> >>         > today's Internet, could use DNS. Or, as the Skype folks do,
> >>         hardcoded
> >>         > bootstrap
> >>         > addresses. But either way, bootstrap is such a crutch. 
> >>         >
> >>         > > A key provision of P2PSIP is to eliminate the need for
> >>         such
> >>         > > authorities in the first place because they are 
> >>         superfluous for the
> >>         > > application.  An extended social network needs no central
> >>         domain
> >>         > > administrator nor a central certificate authority.   It 
> >>         needs at most
> >>         > > a human facilitator not an engineer.
> >>         >
> >>         > So P2PSIP includes a web-of-trust security model? If so, why 
> >>         not leverage
> >>         > existing solutions for that as well (like PGP)?
> >>         >
> >>         > > Such social networks are not trite. For example social 
> >>         networks are
> >>         > > the basis of all peer reviews for medical
> >>         journals.  Shouting doesn't
> >>         > > work when 500 doctors are to be consulted, on several 
> >>         continents, with
> >>         > > varied access to the Internet, and over time.  The person
> >>         running the
> >>         > > peer review is a journal editor, a person who uses 
> >>         end-user tools, not
> >>         > > a network engineer.  DNS isn't the selection criteria
> >>         here, reputation
> >>         > > is.
> >>         > 
> >>         > But DNS is how the journal's mailing list manager finds the
> >>         mail hosts
> >>         > for each of the 500 doctors. Delivery of the article and
> >>         reply comments 
> >>         > across several continents over time works today just fine
> >>         without P2P
> >>         > or DHTs or reputation-based web-of-trust security models.
> >>         See this mailing 
> >>         > list as an existence proof.
> >>         >
> >>         > Matthew Kaufman
> >>         > matthew at matthew.at 
> >>         >
> >>         > _______________________________________________
> >>         > p2p-sip mailing list
> >>         > p2p-sip at cs.columbia.edu
> >>         > https://lists.cs.columbia.edu/cucslists/listinfo/p2p-sip
> >>
> >>         _______________________________________________ 
> >>         p2p-sip mailing list
> >>         p2p-sip at cs.columbia.edu
> >>         https://lists.cs.columbia.edu/cucslists/listinfo/p2p-sip
> >>
> >> _______________________________________________
> >> p2p-sip mailing list
> >> p2p-sip at cs.columbia.edu
> >> https://lists.cs.columbia.edu/cucslists/listinfo/p2p-sip
> >
> > _______________________________________________ 
> > p2p-sip mailing list
> > p2p-sip at cs.columbia.edu
> > https://lists.cs.columbia.edu/cucslists/listinfo/p2p-sip 
> >
>
>
> _______________________________________________
> p2p-sip mailing list
> p2p-sip at cs.columbia.edu
> https://lists.cs.columbia.edu/cucslists/listinfo/p2p-sip

_______________________________________________
p2p-sip mailing list
p2p-sip at cs.columbia.edu
https://lists.cs.columbia.edu/cucslists/listinfo/p2p-sip

 

-------------- next part --------------
An HTML attachment was scrubbed...
URL: https://lists.cs.columbia.edu/pipermail/p2p-sip/attachments/20061128/ae260669/attachment-0001.html

From ekr at networkresonance.com  Tue Nov 28 15:56:59 2006
From: ekr at networkresonance.com (Eric Rescorla)
Date: Tue, 28 Nov 2006 12:56:59 -0800
Subject: [p2p-sip] Fwd: what's wrong with DNS?
In-Reply-To: <1164746611.2622.67.camel@localhost.localdomain> (Frank W.
	Miller's message of "Tue, 28 Nov 2006 13:43:31 -0700")
References: <08b301c71321$9fd25a40$640fa8c0@cis.neustar.com>
	<1164746611.2622.67.camel@localhost.localdomain>
Message-ID: <867ixf1e90.fsf@raman.networkresonance.com>

"Frank W. Miller" <fwmiller at cornfed.com> writes:

>> We'll all agree we need a trusted entity to issue credentials.  The question
>> is do we need one to validate the credential (that is, provide
>> authentication services)?  I'd say, no, we don't.  If you get a credential
>> from a trusted source, we can use p2p mechanisms to validate that
>> credential.  We don't have to trust another peer; each peer can do the
>> authentication itself.  In the current environment, the idea of doing a full
>> credential check with RSA-1024 seems quite reasonable, done at a peer, per
>> call, as long as it only takes ONE hunk of crypto.  We don't need to
>> optimize that.  Similarly, we can do that kind of validation before we start
>> accepting the DHT data exchange messages that manage the distributed hash
>> table from a peer.  There doesn't seem to be a need for a centralized
>> mechanism for that either.  
>> 
>
> This is my lack of security background showing through but how does this
> actually work?  I mean, when you go to validate, how can you do so
> without consulting the trusted entity?
>
> So, lets say some UA wants to set up a call with you.  They send you a
> credential and you RSA it. This gets you some other value.  Are you
> saying that you already know what the resulting value is supposed to be
> so you don't need to consult the authority?  If so, how did you find out
> that value?

The credential contains:
   - Your information
   - The identity of the trusted third party
   - A signature from the trusted third party.

RSA lets you verify that that signature is in fact from the trusted
third party is in fact over that data.


>> Yes, I think we have agreed we will use a DHT.  Our charter actually says we
>> choose one mandatory-to-implement DHT.  I don't think the text actually has
>> to limit the overlay mechanism to using a DHT, but we'll see how torturous
>> the text gets if we make it neutral as to mechanism.
>> 
>
>
> This worries me a little bit (although not hugely).  As an implementor,
> I'd like to be able to use some flooding algorithm like Gnutella or
> something just cuz its a lot easier to implement.  Is it possible to
> just define the interface to the P2P overlay, maybe with some indication
> of the algorithm being used internally, and leave it at that?  It seems
> to me that DHT algorithms and P2P overlay algorithms in general are
> going to evolve going forward.  It would be a shame to tie the protocol
> to one unnecessarily.

The issue isn't about tying it to a single protocol but ensuring
that two independent implementations can talk to each other.
That requires that they implement a common protocol.

-Ekr

From dbarrett at quinthar.com  Tue Nov 28 16:29:37 2006
From: dbarrett at quinthar.com (David Barrett)
Date: Tue, 28 Nov 2006 13:29:37 -0800
Subject: [p2p-sip] Fwd: what's wrong with DNS?
In-Reply-To: <24CCCC428EFEA2469BF046DB3C7A8D22141377@namail5.corp.adobe.com>
Message-ID: <029b01c71334$4f05b630$0200a8c0@Quinthar>

> -----Original Message-----
> From: Henry Sinnreich [mailto:hsinnrei at adobe.com]
> Subject: RE: [p2p-sip] Fwd: what's wrong with DNS?
> 
> BTW: Existing P2P traffic is independent of the DNS and it dominates the
> Internet traffic. Numbers between 60% and 80% have been published.

I apologize, but I think that's just simply false and misleading.

First, virtually all of that traffic is file-sharing traffic designed to
preserve anonymity, which is the direct opposite of our problem.  Using that
as justification to keep our system "clean" of DNS is a red herring.

But even ignoring that, most of that traffic is Bittorrent, which *does* use
DNS not only for search (to find torrents on btjunkie.org and so forth), but
to find peers (to contact the tracker).

At the end of the day, DNS is a fundamental part of the internet, used by
virtually very user not only ever day, but almost every minute.  If DNS went
down globally, the vast majority of internet traffic would also slow to a
stop (web, email, bittorrent, etc).  There would be a few survivors -- Skype
with its hard-coded IPs would persevere, along with eDonkey, Kazaa and other
anonymous or totally closed/private services, but about everything else
would grind to a halt.  Planning on how to keep working without DNS is like
planning on how to work without power -- you just don't.  If DNS goes down,
we have bigger problems to worry about.

We certainly *can* reinvent the wheel and figure out how to keep working in
this apocalypse scenario, I just don't see why we should bother. 

-david


From dbarrett at quinthar.com  Tue Nov 28 16:42:33 2006
From: dbarrett at quinthar.com (David Barrett)
Date: Tue, 28 Nov 2006 13:42:33 -0800
Subject: [p2p-sip] Fwd: what's wrong with DNS?
In-Reply-To: <4C1596FBF66C67478BCFE7B3F81FC1E01751D403EB@DF-MASTIFF-MSG.exchange.corp.microsoft.com>
Message-ID: <02a601c71336$1d12edd0$0200a8c0@Quinthar>

If they're in the same room, why do they need VoIP?  The range of wifi isn't
much wider than the range of the spoken voice.

Even if they did somehow have wifi but not voice range, mDNS (as Matthew has
repeatedly pointed out) is exactly suited for this problem.  If they want
VoIP, they'll probably also want file transfers, IM, and all the other
accoutrements of internet life.  Why would we have one solution for VoIP,
and use something totally different for everything else?

Why not just make mDNS really easy to use, and then use that for these
fringe VoIP cases?

-david

> -----Original Message-----
> From: Sean Olson [mailto:seanol at exchange.microsoft.com]
> Sent: Tuesday, November 28, 2006 1:36 PM
> To: David Barrett; 'Henry Sinnreich'; 'Brian Rosen'; 'Matthew Kaufman';
> slavitch at gmail.com
> Cc: p2p-sip at cs.columbia.edu; 'Kundan Singh'
> Subject: RE: [p2p-sip] Fwd: what's wrong with DNS?
> 
> It would be nice to support an ad-hoc WiFi network among peers and use
> P2PSIP for voice, chat, etc. This is a scenario where DNS does not exist
> necessarily and it is actually fairly common.  Take for example, a set of
> consultants working on site with a customer. They probably don't have
> direct access to the customer's network and maybe not to the broader
> Internet either. They will however likely have WiFi and be able to setup a
> small ad-hoc network amongst themselves.
> 
> Is this scenario out-of-scope for this WG?
> 
> 
> -----Original Message-----
> From: p2p-sip-bounces at cs.columbia.edu [mailto:p2p-sip-
> bounces at cs.columbia.edu] On Behalf Of David Barrett
> Sent: Tuesday, November 28, 2006 1:30 PM
> To: 'Henry Sinnreich'; 'Brian Rosen'; 'Matthew Kaufman';
> slavitch at gmail.com
> Cc: p2p-sip at cs.columbia.edu; 'Kundan Singh'
> Subject: Re: [p2p-sip] Fwd: what's wrong with DNS?
> 
> > -----Original Message-----
> > From: Henry Sinnreich [mailto:hsinnrei at adobe.com]
> > Subject: RE: [p2p-sip] Fwd: what's wrong with DNS?
> >
> > BTW: Existing P2P traffic is independent of the DNS and it dominates the
> > Internet traffic. Numbers between 60% and 80% have been published.
> 
> I apologize, but I think that's just simply false and misleading.
> 
> First, virtually all of that traffic is file-sharing traffic designed to
> preserve anonymity, which is the direct opposite of our problem.  Using
> that
> as justification to keep our system "clean" of DNS is a red herring.
> 
> But even ignoring that, most of that traffic is Bittorrent, which *does*
> use
> DNS not only for search (to find torrents on btjunkie.org and so forth),
> but
> to find peers (to contact the tracker).
> 
> At the end of the day, DNS is a fundamental part of the internet, used by
> virtually very user not only ever day, but almost every minute.  If DNS
> went
> down globally, the vast majority of internet traffic would also slow to a
> stop (web, email, bittorrent, etc).  There would be a few survivors --
> Skype
> with its hard-coded IPs would persevere, along with eDonkey, Kazaa and
> other
> anonymous or totally closed/private services, but about everything else
> would grind to a halt.  Planning on how to keep working without DNS is
> like
> planning on how to work without power -- you just don't.  If DNS goes
> down,
> we have bigger problems to worry about.
> 
> We certainly *can* reinvent the wheel and figure out how to keep working
> in
> this apocalypse scenario, I just don't see why we should bother.
> 
> -david
> 
> _______________________________________________
> p2p-sip mailing list
> p2p-sip at cs.columbia.edu
> https://lists.cs.columbia.edu/cucslists/listinfo/p2p-sip


From seanol at exchange.microsoft.com  Tue Nov 28 16:44:49 2006
From: seanol at exchange.microsoft.com (Sean Olson)
Date: Tue, 28 Nov 2006 13:44:49 -0800
Subject: [p2p-sip] Fwd: what's wrong with DNS?
In-Reply-To: <02a601c71336$1d12edd0$0200a8c0@Quinthar>
References: <4C1596FBF66C67478BCFE7B3F81FC1E01751D403EB@DF-MASTIFF-MSG.exchange.corp.microsoft.com>
	<02a601c71336$1d12edd0$0200a8c0@Quinthar>
Message-ID: <4C1596FBF66C67478BCFE7B3F81FC1E01751D403F7@DF-MASTIFF-MSG.exchange.corp.microsoft.com>

I'm not advocating limiting this to voice. mDNS is great except when multicast is not available. But my real question is not so much about technology but the scenario.

Is this a scenario that is being considered?


-----Original Message-----
From: David Barrett [mailto:dbarrett at quinthar.com]
Sent: Tuesday, November 28, 2006 1:43 PM
To: Sean Olson; 'Henry Sinnreich'; 'Brian Rosen'; 'Matthew Kaufman'; slavitch at gmail.com
Cc: p2p-sip at cs.columbia.edu; 'Kundan Singh'
Subject: RE: [p2p-sip] Fwd: what's wrong with DNS?

If they're in the same room, why do they need VoIP?  The range of wifi isn't
much wider than the range of the spoken voice.

Even if they did somehow have wifi but not voice range, mDNS (as Matthew has
repeatedly pointed out) is exactly suited for this problem.  If they want
VoIP, they'll probably also want file transfers, IM, and all the other
accoutrements of internet life.  Why would we have one solution for VoIP,
and use something totally different for everything else?

Why not just make mDNS really easy to use, and then use that for these
fringe VoIP cases?

-david

> -----Original Message-----
> From: Sean Olson [mailto:seanol at exchange.microsoft.com]
> Sent: Tuesday, November 28, 2006 1:36 PM
> To: David Barrett; 'Henry Sinnreich'; 'Brian Rosen'; 'Matthew Kaufman';
> slavitch at gmail.com
> Cc: p2p-sip at cs.columbia.edu; 'Kundan Singh'
> Subject: RE: [p2p-sip] Fwd: what's wrong with DNS?
>
> It would be nice to support an ad-hoc WiFi network among peers and use
> P2PSIP for voice, chat, etc. This is a scenario where DNS does not exist
> necessarily and it is actually fairly common.  Take for example, a set of
> consultants working on site with a customer. They probably don't have
> direct access to the customer's network and maybe not to the broader
> Internet either. They will however likely have WiFi and be able to setup a
> small ad-hoc network amongst themselves.
>
> Is this scenario out-of-scope for this WG?
>
>
> -----Original Message-----
> From: p2p-sip-bounces at cs.columbia.edu [mailto:p2p-sip-
> bounces at cs.columbia.edu] On Behalf Of David Barrett
> Sent: Tuesday, November 28, 2006 1:30 PM
> To: 'Henry Sinnreich'; 'Brian Rosen'; 'Matthew Kaufman';
> slavitch at gmail.com
> Cc: p2p-sip at cs.columbia.edu; 'Kundan Singh'
> Subject: Re: [p2p-sip] Fwd: what's wrong with DNS?
>
> > -----Original Message-----
> > From: Henry Sinnreich [mailto:hsinnrei at adobe.com]
> > Subject: RE: [p2p-sip] Fwd: what's wrong with DNS?
> >
> > BTW: Existing P2P traffic is independent of the DNS and it dominates the
> > Internet traffic. Numbers between 60% and 80% have been published.
>
> I apologize, but I think that's just simply false and misleading.
>
> First, virtually all of that traffic is file-sharing traffic designed to
> preserve anonymity, which is the direct opposite of our problem.  Using
> that
> as justification to keep our system "clean" of DNS is a red herring.
>
> But even ignoring that, most of that traffic is Bittorrent, which *does*
> use
> DNS not only for search (to find torrents on btjunkie.org and so forth),
> but
> to find peers (to contact the tracker).
>
> At the end of the day, DNS is a fundamental part of the internet, used by
> virtually very user not only ever day, but almost every minute.  If DNS
> went
> down globally, the vast majority of internet traffic would also slow to a
> stop (web, email, bittorrent, etc).  There would be a few survivors --
> Skype
> with its hard-coded IPs would persevere, along with eDonkey, Kazaa and
> other
> anonymous or totally closed/private services, but about everything else
> would grind to a halt.  Planning on how to keep working without DNS is
> like
> planning on how to work without power -- you just don't.  If DNS goes
> down,
> we have bigger problems to worry about.
>
> We certainly *can* reinvent the wheel and figure out how to keep working
> in
> this apocalypse scenario, I just don't see why we should bother.
>
> -david
>
> _______________________________________________
> p2p-sip mailing list
> p2p-sip at cs.columbia.edu
> https://lists.cs.columbia.edu/cucslists/listinfo/p2p-sip



From dbarrett at quinthar.com  Tue Nov 28 16:52:56 2006
From: dbarrett at quinthar.com (David Barrett)
Date: Tue, 28 Nov 2006 13:52:56 -0800
Subject: [p2p-sip] Fwd: what's wrong with DNS?
In-Reply-To: <4C1596FBF66C67478BCFE7B3F81FC1E01751D403F7@DF-MASTIFF-MSG.exchange.corp.microsoft.com>
Message-ID: <02a701c71337$90b99940$0200a8c0@Quinthar>

Incidentally, I believe ad hoc networks *do* support multicast.  Indeed,
multicast on the LAN is actually quite common (it's multicast on the WAN
that's generally not available).

Furthermore, it's my impression that the scope of our charter *is* limited
to voice.  If we don't draw the line there, then where *do* we draw the
line?

-david
 
> -----Original Message-----
> From: Sean Olson [mailto:seanol at exchange.microsoft.com]
> Sent: Tuesday, November 28, 2006 1:45 PM
> To: David Barrett; 'Henry Sinnreich'; 'Brian Rosen'; 'Matthew Kaufman';
> slavitch at gmail.com
> Cc: p2p-sip at cs.columbia.edu; 'Kundan Singh'
> Subject: RE: [p2p-sip] Fwd: what's wrong with DNS?
> 
> I'm not advocating limiting this to voice. mDNS is great except when
> multicast is not available. But my real question is not so much about
> technology but the scenario.
> 
> Is this a scenario that is being considered?
> 
> 
> -----Original Message-----
> From: David Barrett [mailto:dbarrett at quinthar.com]
> Sent: Tuesday, November 28, 2006 1:43 PM
> To: Sean Olson; 'Henry Sinnreich'; 'Brian Rosen'; 'Matthew Kaufman';
> slavitch at gmail.com
> Cc: p2p-sip at cs.columbia.edu; 'Kundan Singh'
> Subject: RE: [p2p-sip] Fwd: what's wrong with DNS?
> 
> If they're in the same room, why do they need VoIP?  The range of wifi
> isn't
> much wider than the range of the spoken voice.
> 
> Even if they did somehow have wifi but not voice range, mDNS (as Matthew
> has
> repeatedly pointed out) is exactly suited for this problem.  If they want
> VoIP, they'll probably also want file transfers, IM, and all the other
> accoutrements of internet life.  Why would we have one solution for VoIP,
> and use something totally different for everything else?
> 
> Why not just make mDNS really easy to use, and then use that for these
> fringe VoIP cases?
> 
> -david
> 
> > -----Original Message-----
> > From: Sean Olson [mailto:seanol at exchange.microsoft.com]
> > Sent: Tuesday, November 28, 2006 1:36 PM
> > To: David Barrett; 'Henry Sinnreich'; 'Brian Rosen'; 'Matthew Kaufman';
> > slavitch at gmail.com
> > Cc: p2p-sip at cs.columbia.edu; 'Kundan Singh'
> > Subject: RE: [p2p-sip] Fwd: what's wrong with DNS?
> >
> > It would be nice to support an ad-hoc WiFi network among peers and use
> > P2PSIP for voice, chat, etc. This is a scenario where DNS does not exist
> > necessarily and it is actually fairly common.  Take for example, a set
> of
> > consultants working on site with a customer. They probably don't have
> > direct access to the customer's network and maybe not to the broader
> > Internet either. They will however likely have WiFi and be able to setup
> a
> > small ad-hoc network amongst themselves.
> >
> > Is this scenario out-of-scope for this WG?
> >
> >
> > -----Original Message-----
> > From: p2p-sip-bounces at cs.columbia.edu [mailto:p2p-sip-
> > bounces at cs.columbia.edu] On Behalf Of David Barrett
> > Sent: Tuesday, November 28, 2006 1:30 PM
> > To: 'Henry Sinnreich'; 'Brian Rosen'; 'Matthew Kaufman';
> > slavitch at gmail.com
> > Cc: p2p-sip at cs.columbia.edu; 'Kundan Singh'
> > Subject: Re: [p2p-sip] Fwd: what's wrong with DNS?
> >
> > > -----Original Message-----
> > > From: Henry Sinnreich [mailto:hsinnrei at adobe.com]
> > > Subject: RE: [p2p-sip] Fwd: what's wrong with DNS?
> > >
> > > BTW: Existing P2P traffic is independent of the DNS and it dominates
> the
> > > Internet traffic. Numbers between 60% and 80% have been published.
> >
> > I apologize, but I think that's just simply false and misleading.
> >
> > First, virtually all of that traffic is file-sharing traffic designed to
> > preserve anonymity, which is the direct opposite of our problem.  Using
> > that
> > as justification to keep our system "clean" of DNS is a red herring.
> >
> > But even ignoring that, most of that traffic is Bittorrent, which *does*
> > use
> > DNS not only for search (to find torrents on btjunkie.org and so forth),
> > but
> > to find peers (to contact the tracker).
> >
> > At the end of the day, DNS is a fundamental part of the internet, used
> by
> > virtually very user not only ever day, but almost every minute.  If DNS
> > went
> > down globally, the vast majority of internet traffic would also slow to
> a
> > stop (web, email, bittorrent, etc).  There would be a few survivors --
> > Skype
> > with its hard-coded IPs would persevere, along with eDonkey, Kazaa and
> > other
> > anonymous or totally closed/private services, but about everything else
> > would grind to a halt.  Planning on how to keep working without DNS is
> > like
> > planning on how to work without power -- you just don't.  If DNS goes
> > down,
> > we have bigger problems to worry about.
> >
> > We certainly *can* reinvent the wheel and figure out how to keep working
> > in
> > this apocalypse scenario, I just don't see why we should bother.
> >
> > -david
> >
> > _______________________________________________
> > p2p-sip mailing list
> > p2p-sip at cs.columbia.edu
> > https://lists.cs.columbia.edu/cucslists/listinfo/p2p-sip


From seanol at exchange.microsoft.com  Tue Nov 28 16:56:36 2006
From: seanol at exchange.microsoft.com (Sean Olson)
Date: Tue, 28 Nov 2006 13:56:36 -0800
Subject: [p2p-sip] Fwd: what's wrong with DNS?
In-Reply-To: <02a701c71337$90b99940$0200a8c0@Quinthar>
References: <4C1596FBF66C67478BCFE7B3F81FC1E01751D403F7@DF-MASTIFF-MSG.exchange.corp.microsoft.com>
	<02a701c71337$90b99940$0200a8c0@Quinthar>
Message-ID: <4C1596FBF66C67478BCFE7B3F81FC1E01751D40406@DF-MASTIFF-MSG.exchange.corp.microsoft.com>

I don't understand -- we didn't limit the charter of the SIP WG to just voice. Why would it be different for P2PSIP?  Why would you get less service from the P2P network than a traditional trapezoid? IM, file transfer, video, etc. are all reasonable modalities for P2P. Am I missing something that would make these harder than voice?

For multicast, the issue is not technical support but admin enablement. Some do; some don't. I've got nothing against using multicast as one option. But if it is the only option, that would be a deployment blocker in many real-world customers.



-----Original Message-----
From: David Barrett [mailto:dbarrett at quinthar.com]
Sent: Tuesday, November 28, 2006 1:53 PM
To: Sean Olson; 'Henry Sinnreich'; 'Brian Rosen'; 'Matthew Kaufman'; slavitch at gmail.com
Cc: p2p-sip at cs.columbia.edu; 'Kundan Singh'
Subject: RE: [p2p-sip] Fwd: what's wrong with DNS?

Incidentally, I believe ad hoc networks *do* support multicast.  Indeed,
multicast on the LAN is actually quite common (it's multicast on the WAN
that's generally not available).

Furthermore, it's my impression that the scope of our charter *is* limited
to voice.  If we don't draw the line there, then where *do* we draw the
line?

-david

> -----Original Message-----
> From: Sean Olson [mailto:seanol at exchange.microsoft.com]
> Sent: Tuesday, November 28, 2006 1:45 PM
> To: David Barrett; 'Henry Sinnreich'; 'Brian Rosen'; 'Matthew Kaufman';
> slavitch at gmail.com
> Cc: p2p-sip at cs.columbia.edu; 'Kundan Singh'
> Subject: RE: [p2p-sip] Fwd: what's wrong with DNS?
>
> I'm not advocating limiting this to voice. mDNS is great except when
> multicast is not available. But my real question is not so much about
> technology but the scenario.
>
> Is this a scenario that is being considered?
>
>
> -----Original Message-----
> From: David Barrett [mailto:dbarrett at quinthar.com]
> Sent: Tuesday, November 28, 2006 1:43 PM
> To: Sean Olson; 'Henry Sinnreich'; 'Brian Rosen'; 'Matthew Kaufman';
> slavitch at gmail.com
> Cc: p2p-sip at cs.columbia.edu; 'Kundan Singh'
> Subject: RE: [p2p-sip] Fwd: what's wrong with DNS?
>
> If they're in the same room, why do they need VoIP?  The range of wifi
> isn't
> much wider than the range of the spoken voice.
>
> Even if they did somehow have wifi but not voice range, mDNS (as Matthew
> has
> repeatedly pointed out) is exactly suited for this problem.  If they want
> VoIP, they'll probably also want file transfers, IM, and all the other
> accoutrements of internet life.  Why would we have one solution for VoIP,
> and use something totally different for everything else?
>
> Why not just make mDNS really easy to use, and then use that for these
> fringe VoIP cases?
>
> -david
>
> > -----Original Message-----
> > From: Sean Olson [mailto:seanol at exchange.microsoft.com]
> > Sent: Tuesday, November 28, 2006 1:36 PM
> > To: David Barrett; 'Henry Sinnreich'; 'Brian Rosen'; 'Matthew Kaufman';
> > slavitch at gmail.com
> > Cc: p2p-sip at cs.columbia.edu; 'Kundan Singh'
> > Subject: RE: [p2p-sip] Fwd: what's wrong with DNS?
> >
> > It would be nice to support an ad-hoc WiFi network among peers and use
> > P2PSIP for voice, chat, etc. This is a scenario where DNS does not exist
> > necessarily and it is actually fairly common.  Take for example, a set
> of
> > consultants working on site with a customer. They probably don't have
> > direct access to the customer's network and maybe not to the broader
> > Internet either. They will however likely have WiFi and be able to setup
> a
> > small ad-hoc network amongst themselves.
> >
> > Is this scenario out-of-scope for this WG?
> >
> >
> > -----Original Message-----
> > From: p2p-sip-bounces at cs.columbia.edu [mailto:p2p-sip-
> > bounces at cs.columbia.edu] On Behalf Of David Barrett
> > Sent: Tuesday, November 28, 2006 1:30 PM
> > To: 'Henry Sinnreich'; 'Brian Rosen'; 'Matthew Kaufman';
> > slavitch at gmail.com
> > Cc: p2p-sip at cs.columbia.edu; 'Kundan Singh'
> > Subject: Re: [p2p-sip] Fwd: what's wrong with DNS?
> >
> > > -----Original Message-----
> > > From: Henry Sinnreich [mailto:hsinnrei at adobe.com]
> > > Subject: RE: [p2p-sip] Fwd: what's wrong with DNS?
> > >
> > > BTW: Existing P2P traffic is independent of the DNS and it dominates
> the
> > > Internet traffic. Numbers between 60% and 80% have been published.
> >
> > I apologize, but I think that's just simply false and misleading.
> >
> > First, virtually all of that traffic is file-sharing traffic designed to
> > preserve anonymity, which is the direct opposite of our problem.  Using
> > that
> > as justification to keep our system "clean" of DNS is a red herring.
> >
> > But even ignoring that, most of that traffic is Bittorrent, which *does*
> > use
> > DNS not only for search (to find torrents on btjunkie.org and so forth),
> > but
> > to find peers (to contact the tracker).
> >
> > At the end of the day, DNS is a fundamental part of the internet, used
> by
> > virtually very user not only ever day, but almost every minute.  If DNS
> > went
> > down globally, the vast majority of internet traffic would also slow to
> a
> > stop (web, email, bittorrent, etc).  There would be a few survivors --
> > Skype
> > with its hard-coded IPs would persevere, along with eDonkey, Kazaa and
> > other
> > anonymous or totally closed/private services, but about everything else
> > would grind to a halt.  Planning on how to keep working without DNS is
> > like
> > planning on how to work without power -- you just don't.  If DNS goes
> > down,
> > we have bigger problems to worry about.
> >
> > We certainly *can* reinvent the wheel and figure out how to keep working
> > in
> > this apocalypse scenario, I just don't see why we should bother.
> >
> > -david
> >
> > _______________________________________________
> > p2p-sip mailing list
> > p2p-sip at cs.columbia.edu
> > https://lists.cs.columbia.edu/cucslists/listinfo/p2p-sip



From dbarrett at quinthar.com  Tue Nov 28 17:14:37 2006
From: dbarrett at quinthar.com (David Barrett)
Date: Tue, 28 Nov 2006 14:14:37 -0800
Subject: [p2p-sip] Fwd: what's wrong with DNS?
In-Reply-To: <4C1596FBF66C67478BCFE7B3F81FC1E01751D40406@DF-MASTIFF-MSG.exchange.corp.microsoft.com>
Message-ID: <02ba01c7133a$980b5960$0200a8c0@Quinthar>

Ah, perhaps this is my mistake.  Isn't it limited to the general class of
audio/video/realtime communications?  For example, is background swarming
file-transfers out of scope?  I apologize; I clearly don't know what I'm
talking about here regarding scope, I'm sorry for the confusion.

To take it from another angle, what is out of scope?  Are we competing head
to head with UPnP for generalized resource advertisement and discovery?

-david

> -----Original Message-----
> From: Sean Olson [mailto:seanol at exchange.microsoft.com]
> Sent: Tuesday, November 28, 2006 1:57 PM
> To: David Barrett; 'Henry Sinnreich'; 'Brian Rosen'; 'Matthew Kaufman';
> slavitch at gmail.com
> Cc: p2p-sip at cs.columbia.edu; 'Kundan Singh'
> Subject: RE: [p2p-sip] Fwd: what's wrong with DNS?
> 
> I don't understand -- we didn't limit the charter of the SIP WG to just
> voice. Why would it be different for P2PSIP?  Why would you get less
> service from the P2P network than a traditional trapezoid? IM, file
> transfer, video, etc. are all reasonable modalities for P2P. Am I missing
> something that would make these harder than voice?
> 
> For multicast, the issue is not technical support but admin enablement.
> Some do; some don't. I've got nothing against using multicast as one
> option. But if it is the only option, that would be a deployment blocker
> in many real-world customers.
> 
> 
> 
> -----Original Message-----
> From: David Barrett [mailto:dbarrett at quinthar.com]
> Sent: Tuesday, November 28, 2006 1:53 PM
> To: Sean Olson; 'Henry Sinnreich'; 'Brian Rosen'; 'Matthew Kaufman';
> slavitch at gmail.com
> Cc: p2p-sip at cs.columbia.edu; 'Kundan Singh'
> Subject: RE: [p2p-sip] Fwd: what's wrong with DNS?
> 
> Incidentally, I believe ad hoc networks *do* support multicast.  Indeed,
> multicast on the LAN is actually quite common (it's multicast on the WAN
> that's generally not available).
> 
> Furthermore, it's my impression that the scope of our charter *is* limited
> to voice.  If we don't draw the line there, then where *do* we draw the
> line?
> 
> -david
> 
> > -----Original Message-----
> > From: Sean Olson [mailto:seanol at exchange.microsoft.com]
> > Sent: Tuesday, November 28, 2006 1:45 PM
> > To: David Barrett; 'Henry Sinnreich'; 'Brian Rosen'; 'Matthew Kaufman';
> > slavitch at gmail.com
> > Cc: p2p-sip at cs.columbia.edu; 'Kundan Singh'
> > Subject: RE: [p2p-sip] Fwd: what's wrong with DNS?
> >
> > I'm not advocating limiting this to voice. mDNS is great except when
> > multicast is not available. But my real question is not so much about
> > technology but the scenario.
> >
> > Is this a scenario that is being considered?
> >
> >
> > -----Original Message-----
> > From: David Barrett [mailto:dbarrett at quinthar.com]
> > Sent: Tuesday, November 28, 2006 1:43 PM
> > To: Sean Olson; 'Henry Sinnreich'; 'Brian Rosen'; 'Matthew Kaufman';
> > slavitch at gmail.com
> > Cc: p2p-sip at cs.columbia.edu; 'Kundan Singh'
> > Subject: RE: [p2p-sip] Fwd: what's wrong with DNS?
> >
> > If they're in the same room, why do they need VoIP?  The range of wifi
> > isn't
> > much wider than the range of the spoken voice.
> >
> > Even if they did somehow have wifi but not voice range, mDNS (as Matthew
> > has
> > repeatedly pointed out) is exactly suited for this problem.  If they
> want
> > VoIP, they'll probably also want file transfers, IM, and all the other
> > accoutrements of internet life.  Why would we have one solution for
> VoIP,
> > and use something totally different for everything else?
> >
> > Why not just make mDNS really easy to use, and then use that for these
> > fringe VoIP cases?
> >
> > -david
> >
> > > -----Original Message-----
> > > From: Sean Olson [mailto:seanol at exchange.microsoft.com]
> > > Sent: Tuesday, November 28, 2006 1:36 PM
> > > To: David Barrett; 'Henry Sinnreich'; 'Brian Rosen'; 'Matthew
> Kaufman';
> > > slavitch at gmail.com
> > > Cc: p2p-sip at cs.columbia.edu; 'Kundan Singh'
> > > Subject: RE: [p2p-sip] Fwd: what's wrong with DNS?
> > >
> > > It would be nice to support an ad-hoc WiFi network among peers and use
> > > P2PSIP for voice, chat, etc. This is a scenario where DNS does not
> exist
> > > necessarily and it is actually fairly common.  Take for example, a set
> > of
> > > consultants working on site with a customer. They probably don't have
> > > direct access to the customer's network and maybe not to the broader
> > > Internet either. They will however likely have WiFi and be able to
> setup
> > a
> > > small ad-hoc network amongst themselves.
> > >
> > > Is this scenario out-of-scope for this WG?
> > >
> > >
> > > -----Original Message-----
> > > From: p2p-sip-bounces at cs.columbia.edu [mailto:p2p-sip-
> > > bounces at cs.columbia.edu] On Behalf Of David Barrett
> > > Sent: Tuesday, November 28, 2006 1:30 PM
> > > To: 'Henry Sinnreich'; 'Brian Rosen'; 'Matthew Kaufman';
> > > slavitch at gmail.com
> > > Cc: p2p-sip at cs.columbia.edu; 'Kundan Singh'
> > > Subject: Re: [p2p-sip] Fwd: what's wrong with DNS?
> > >
> > > > -----Original Message-----
> > > > From: Henry Sinnreich [mailto:hsinnrei at adobe.com]
> > > > Subject: RE: [p2p-sip] Fwd: what's wrong with DNS?
> > > >
> > > > BTW: Existing P2P traffic is independent of the DNS and it dominates
> > the
> > > > Internet traffic. Numbers between 60% and 80% have been published.
> > >
> > > I apologize, but I think that's just simply false and misleading.
> > >
> > > First, virtually all of that traffic is file-sharing traffic designed
> to
> > > preserve anonymity, which is the direct opposite of our problem.
> Using
> > > that
> > > as justification to keep our system "clean" of DNS is a red herring.
> > >
> > > But even ignoring that, most of that traffic is Bittorrent, which
> *does*
> > > use
> > > DNS not only for search (to find torrents on btjunkie.org and so
> forth),
> > > but
> > > to find peers (to contact the tracker).
> > >
> > > At the end of the day, DNS is a fundamental part of the internet, used
> > by
> > > virtually very user not only ever day, but almost every minute.  If
> DNS
> > > went
> > > down globally, the vast majority of internet traffic would also slow
> to
> > a
> > > stop (web, email, bittorrent, etc).  There would be a few survivors --
> > > Skype
> > > with its hard-coded IPs would persevere, along with eDonkey, Kazaa and
> > > other
> > > anonymous or totally closed/private services, but about everything
> else
> > > would grind to a halt.  Planning on how to keep working without DNS is
> > > like
> > > planning on how to work without power -- you just don't.  If DNS goes
> > > down,
> > > we have bigger problems to worry about.
> > >
> > > We certainly *can* reinvent the wheel and figure out how to keep
> working
> > > in
> > > this apocalypse scenario, I just don't see why we should bother.
> > >
> > > -david
> > >
> > > _______________________________________________
> > > p2p-sip mailing list
> > > p2p-sip at cs.columbia.edu
> > > https://lists.cs.columbia.edu/cucslists/listinfo/p2p-sip


From matthew at matthew.at  Tue Nov 28 17:21:13 2006
From: matthew at matthew.at (Matthew Kaufman)
Date: Tue, 28 Nov 2006 14:21:13 -0800
Subject: [p2p-sip] Fwd: what's wrong with DNS?
In-Reply-To: <4C1596FBF66C67478BCFE7B3F81FC1E01751D403F7@DF-MASTIFF-MSG.exchange.corp.microsoft.com>
Message-ID: <001b01c7133b$83ca7430$029b0a0a@matthewdeskxps>

The "ad-hoc wifi network" will support zeroconf mDNS, as will almost all
other ad-hoc LAN setups. Anything bigger will need a DHCP server and other
infrastructure (like a local DNS server), even if not connected to the
global Internet.

That means that a DNS + mDNS API should be sufficient to both locate node IP
addresses (important) and look up stored credentials (possibly required,
depending on security model). Now we just need to standardize on NAT
traversal (STUN/TURN/ICE/etc.) and whatever mandatory extensions to SIP
itself that might be required to do infrastructure-less P2P SIP
connections... and that last part is where I think a "P2P SIP" group might
produce the greatest value.

If there's consensus that a Chord-like DHT (demonstrably both old DHT
technology and not resilient in the face of likely churn levels or any
number of denial-of-service attack scenarios) is absolutely necessary in
order to call ourselves "P2P", then I guess we should find or invent
something that it is actually required for. Or we could remove it from the
list of requirements... which would be my vote, if I had one.

Matthew Kaufman
matthew at matthew.at




From adamfisk at gmail.com  Tue Nov 28 18:20:34 2006
From: adamfisk at gmail.com (Adam Fisk)
Date: Tue, 28 Nov 2006 18:20:34 -0500
Subject: [p2p-sip] Fwd: what's wrong with DNS?
In-Reply-To: <015301c712b9$5daeb950$0200a8c0@Quinthar>
References: <200611271818.kARII0191955@where.matthew.at>
	<015301c712b9$5daeb950$0200a8c0@Quinthar>
Message-ID: <f40963db0611281520p6aa81677r62132b4a86e7cc1a@mail.gmail.com>

I couldn't have said it better, David (I probably would have said it much
worse!).  I have yet to hear any arguments against this approach compelling
enough to warrant the complexity of the proposed solutions.  I continue to
think the solution is not appealing to people because it's too simple and
not sexy enough, regardless of real-world applicability -- it's
anticlimactic, as you said.

I unfortunately don't have the incentive or the time to keep pushing this
perspective, though.  If the decisions of this list actually affected me in
some way, I'd argue this thing to the ground, because I think it's clearly
the winning idea.  It's tragic for decisions at this high a level to be
decided by attrition rather than technical merit, but that appears to be
what's happening barring some groundswell of support from the group's core
participants.  Maybe there's some compelling argument out there against the
DNS solution for the connected-internet case, but I have yet to hear it.

There's a lot of talk about what happens in "real p2p systems."  Both David
and I have a great deal of experience with these systems, folks (core
developers of LimeWire/Last Bamboo and RedSwoosh/iGlance).  We don't need
convincing about the merits of DHTs, the various bootstrapping solutions out
there, etc etc.  I think it's interesting, though, that the p2p guys are the
ones arguing for the more "centralized" approach.  It's in the interest of
seeing the work here deployed on a meaningful scale, which to me means
implementing the simplest solution that will work.

-Adam
-------------- next part --------------
An HTML attachment was scrubbed...
URL: https://lists.cs.columbia.edu/pipermail/p2p-sip/attachments/20061128/fe6fc350/attachment.html

From huang-ming.pan at comcast.net  Tue Nov 28 18:25:01 2006
From: huang-ming.pan at comcast.net (Peter Pan)
Date: Tue, 28 Nov 2006 15:25:01 -0800
Subject: [p2p-sip] Fwd: what's wrong with DNS?
References: <027a01c7132f$a043b150$0200a8c0@Quinthar>
Message-ID: <002f01c71344$6ded7190$030aa8c0@comcast.net>

Signed credentials are encrypted in local filesystem as files or Windows
registry entries. The filesystem needs not be encrypted entirely.

If you ever act as multiple Skype users with your laptop, you should have
found that you need different passwords to retrieve individual credentials
as you switch among the userids you've applied from Skype server. This is
what I meant by "sophistication" (of Skype) long time ago, which is an issue
of software design instead of protocol standardization :(

Peter
----- Original Message -----
From: David Barrett
To: 'Michael Slavitch'
Cc: p2p-sip at cs.columbia.edu
Sent: Tuesday, November 28, 2006 12:56 PM
Subject: Re: [p2p-sip] Fwd: what's wrong with DNS?


I'm not sure I understand this answer.  Are you saying that someone stealing
my laptop doesn't matter because my file system is encrypted, and thus they
can't log in as me anyway?  This seems to overstate the prevalence of
encrypted hard drives; I'd wager that nearly 0% of users on a global scale
do this.

Are you suggesting that P2P-SIP's security model should assume a totally
encrypted hard drive?

And even assuming this, are you saying that if someone steals my laptop
*and* my password, there is no resolution to the problem other than to wait
it out?

I'm not sure users will like this.

-david







From adamfisk at gmail.com  Tue Nov 28 18:35:26 2006
From: adamfisk at gmail.com (Adam Fisk)
Date: Tue, 28 Nov 2006 18:35:26 -0500
Subject: [p2p-sip] Fwd: what's wrong with DNS?
In-Reply-To: <7b683f3f0611280630t369ea8aesc2ceb1a1efea110c@mail.gmail.com>
References: <200611271818.kARII0191955@where.matthew.at>
	<015301c712b9$5daeb950$0200a8c0@Quinthar>
	<7b683f3f0611280630t369ea8aesc2ceb1a1efea110c@mail.gmail.com>
Message-ID: <f40963db0611281535i4bcaddfem1d885944a397c012@mail.gmail.com>

David's breakdown described in best, earlier in this thread.  Here it is
again:

1)       Every user chooses a domain name.  Mine might be
'dbarrett.p2psip.org'

 2)       When I start my client up, I choose any non-firewalled peer (how
do I find it?) and REGISTER myself with it as something like 'dbarrett@
dbarrett.p2psip.org'

 3)       After registering with the peer/proxy, I contact some
highly-redundant set of dynamic DNS servers and say "this is the latest IP
address of my proxy, point to it whenever anyone does a DNS lookup on
dbarrett.p2psip.org"

 4)       Your off the shelf, totally unaltered SIP client asks DNS for the
IP of 'dbarrett.p2psip.org', gets the latest IP of my proxy, and looks up
the user 'dbarrett at dbarrett.p2psip.org', and then contacts me as normal.

 5)       If the peer I chose as my proxy goes down, I detect it immediately
and choose a new non-firewwalled peer (again, how do I find it?), REGISTER
with that as in step (2), and update my dynamic DNS binding as in step (3).

 6)       Repeat.


On 11/28/06, Eunsoo Shim <eunsooshim at gmail.com> wrote:
>
> I am not sure what your proposal is.
> Could you provide more specifics on your proposal?
> For example, are you suggesting that we use existing DNS servers to store
> whatever we want to store at the P2P(SIP) network?
> Or are you suggesting that we use the DNS PROTOCOL as (part of) the
> protocol for P2P SIP and build a separate (global or ad-hoc) DNS-like
> infrastructure?
> Thanks.
>
> Eunsoo
>
> On 11/28/06, David Barrett <dbarrett at quinthar.com> wrote:
> >
> > I think Matthew highlights the Achilles heel of so-called "pure" P2P: it
> > never is.  In the real world, there's always a central service behind
> > the
> > curtain:
> >
> > - In Skype's case, it's a central directory at a hard-coded IP address.
> >
> > - In P2P-SIP's case (if Frank and Brian are right) is some kind of
> > centralized authorization/authentication/bootstrapping service.
> >
> > So granted, there's the question of "how do you find this service?"  The
> >
> > obvious choice favored by every other application on the internet (and
> > most
> > private networks) is "DNS".
> >
> > But ignoring that, there's an even more important question:
> >
> >         "Who controls this service?"
> >
> > If we build some fancy DHT solution we imply "some unnamed benevolent
> > dictator yet to be appointed" will rule us all.  We've sorta assumed
> > this
> > (or similarly vague explanations) for the entire course of the P2P-SIP
> > discussion.  For a long while I was not only on board, I was promoting
> > this
> > "let's solve security later" approach.
> >
> > But then Adam proposed a different owner, the IANA.  And the way to do
> > that
> > is through the existing, tried-and-true interface of DNS (specifically,
> > dynamic DNS).  I think this is obvious and brilliant.
> >
> > Building a nonpartial, internationally represented, neutral global third
> > party isn't easy, and while the IANA hasn't wildly succeeded in this,
> > they're closer than anybody else (and decades ahead of anything we'd
> > invent
> > while searching for our benevolent dictator).
> >
> >
> > Now I'll grant DNS is limited to only, say, 99.999% of users.  And I
> > understand that there's a lot of interest in how the remaining 0.001% of
> > users will work without it.  But I'm having a hard time understanding
> > why
> > we're even discussing them.  Indeed, Matthew's repeated calls for a
> > real-world use case of a network that doesn't already have, won't
> > quickly
> > get, or couldn't easily be configured to have DNS has remained
> > unanswered.
> > (And I'd add to that filter "outside of loud talking range.")
> >
> > Furthermore, I'll grant you could build a DHT layer that serves as a
> > read-through cache of the central security mechanism that provides a
> > faster
> > response and propagation time than DNS.  But Akamai, UltraDNS, and other
> > massive, global organizations have proven that DNS can propagate pretty
> > dang
> > quick.
> >
> > And while I think it's an excellent, open question on how to improve DNS
> > to
> > make it even faster (perhaps with some kind of P2P layer), or how to
> > make it
> > work better in partially or totally disconnected scenarios, or how to
> > make
> > it more resistant to spamming/phishing and so forth, that seems like a
> > much
> > bigger problem than SIP alone and thus better suited to its own group.
> >
> > And finally, I agree there are open questions as to how my
> > car-of-the-future
> > will talk to my garage door and sprinkler system.  But why is this
> > relevant?
> > I share Adam's concern that we're dreaming up a panacea to every
> > conceivable
> > problem rather than standardizing a solution to problems that actually
> > exist
> > with technologies that are already at hand.
> >
> >
> > I apologize for being obtuse, but dynamic DNS (even with today's
> > limitations, and perhaps with future improvements) seems like such an
> > obvious way to solve the distributed name resolution problem that I'm
> > astonished it's being discounted so easily in favor of complex solutions
> > to
> > nebulous problems experienced by nonexistent userbases.
> >
> > I'll admit the conclusion is anticlimactic -- I'm a P2P developer who
> > likes
> > high tech as much as anybody.  But where has my reasoning fallen off
> > track?
> >
> > -david
> >
> > > -----Original Message-----
> > > From: p2p-sip-bounces at cs.columbia.edu [mailto:p2p-sip-
> > > bounces at cs.columbia.edu] On Behalf Of Matthew Kaufman
> > > Sent: Monday, November 27, 2006 10:18 AM
> > > To: hsinnrei at adobe.com; slavitch at gmail.com
> > > Cc: p2p-sip at cs.columbia.edu
> > > Subject: Re: [p2p-sip] Fwd: what's wrong with DNS?
> > >
> > > Michael Slavitch:
> > > > As a form of assessing identity DNS is particularly weak and
> > requires
> > > > considerable care and expert feeding as provisioning is always
> > > > ultimately human-based requiring some kind of authority.
> > >
> > > Agreed. Though in small disconnected networks, zeroconf using mDNS
> > avoids
> > > much of this, and in large networks the IT guys or your ISP manage the
> > > complexity for you (though they haven't solved the security problem
> > yet)
> > >
> > > > A self-organizing automated system needs no such crutches.
> > >
> > > Except for bootstrapping to get the first two nodes talking. Which, if
> > on
> > > the Internet, means a way to find IP addresses of those nodes. Which,
> > on
> > > today's Internet, could use DNS. Or, as the Skype folks do, hardcoded
> > > bootstrap
> > > addresses. But either way, bootstrap is such a crutch.
> > >
> > > > A key provision of P2PSIP is to eliminate the need for such
> > > > authorities in the first place because they are superfluous for the
> > > > application.  An extended social network needs no central domain
> > > > administrator nor a central certificate authority.   It needs at
> > most
> > > > a human facilitator not an engineer.
> > >
> > > So P2PSIP includes a web-of-trust security model? If so, why not
> > leverage
> > > existing solutions for that as well (like PGP)?
> > >
> > > > Such social networks are not trite. For example social networks are
> > > > the basis of all peer reviews for medical journals.  Shouting
> > doesn't
> > > > work when 500 doctors are to be consulted, on several continents,
> > with
> > > > varied access to the Internet, and over time.  The person running
> > the
> > > > peer review is a journal editor, a person who uses end-user tools,
> > not
> > > > a network engineer.  DNS isn't the selection criteria here,
> > reputation
> > > > is.
> > >
> > > But DNS is how the journal's mailing list manager finds the mail hosts
> >
> > > for each of the 500 doctors. Delivery of the article and reply
> > comments
> > > across several continents over time works today just fine without P2P
> > > or DHTs or reputation-based web-of-trust security models. See this
> > mailing
> > > list as an existence proof.
> > >
> > > Matthew Kaufman
> > > matthew at matthew.at
> > >
> > > _______________________________________________
> > > p2p-sip mailing list
> > > p2p-sip at cs.columbia.edu
> > > https://lists.cs.columbia.edu/cucslists/listinfo/p2p-sip
> >
> > _______________________________________________
> > p2p-sip mailing list
> > p2p-sip at cs.columbia.edu
> > https://lists.cs.columbia.edu/cucslists/listinfo/p2p-sip
> >
>
>
> _______________________________________________
> p2p-sip mailing list
> p2p-sip at cs.columbia.edu
> https://lists.cs.columbia.edu/cucslists/listinfo/p2p-sip
>
>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: https://lists.cs.columbia.edu/pipermail/p2p-sip/attachments/20061128/6f8c0b51/attachment-0001.html

From matthew at matthew.at  Tue Nov 28 19:29:45 2006
From: matthew at matthew.at (Matthew Kaufman)
Date: Tue, 28 Nov 2006 16:29:45 -0800
Subject: [p2p-sip] Fwd: what's wrong with DNS?
In-Reply-To: <f40963db0611281535i4bcaddfem1d885944a397c012@mail.gmail.com>
Message-ID: <002501c7134d$78ab4bd0$029b0a0a@matthewdeskxps>

Adam Fisk:

David's breakdown described in best, earlier in this thread.  Here it is
again:

1)       Every user chooses a domain name.  Mine might be
'dbarrett.p2psip.org' 

2)       When I start my client up, I choose any non-firewalled peer (how do
I find it?) and REGISTER myself with it as something like
'dbarrett at dbarrett.p2psip.org'

3)       After registering with the peer/proxy, I contact some
highly-redundant set of dynamic DNS servers and say "this is the latest IP
address of my proxy, point to it whenever anyone does a DNS lookup on
dbarrett.p2psip.org"

4)       Your off the shelf, totally unaltered SIP client asks DNS for the
IP of 'dbarrett.p2psip.org', gets the latest IP of my proxy, and looks up
the user 'dbarrett at dbarrett.p2psip.org', and then contacts me as normal.

5)       If the peer I chose as my proxy goes down, I detect it immediately
and choose a new non-firewwalled peer (again, how do I find it?), REGISTER
with that as in step (2), and update my dynamic DNS binding as in step (3).

6)       Repeat.

 

I would add to this that if you are yourself not firewalled, you should be
able to REGISTER with "yourself" and your machine should be able to directly
process inbound calls that arrive.

 

If you are firewalled, clearly a node is needed that can relay call setup
messages to you, and pretty clearly those should just be normal SIP
messages. There's a good argument for allowing one to register with more
than one at a time and point the dynamic DNS appropriately.

 

Additionally, you should also answer mDNS queries for you.p2psip.org with
your local self-proxy, so that machines on your LAN can call you directly
whether or not the DNS is reachable, if their SIP clients are mDNS-aware.

 

As for your open issue of finding a non-firewalled peer to register with?
Use the DNS to get an address of one inside of p2psip.org (if that's the
domain within which you're operating) Note that this covers the case of an
organization that wants the clients registered with a stable machine or
machines that it operates instead of with client machines acting as
registrars. The organization can also set policy around what addresses can
be registered, if credentials are required, etc.

 

The real work that needs doing is coming up with a minimal set of
recommended existing standards plus the minimal extensions necessary such
that a working system can be built. Clearly STUN or better is needed for NAT
traversal. Clearly SIP itself is the call signaling, though extensions might
enable neat P2P capabilities. There's some work around using mDNS for the
detached-from-real-Internet cases. There's some work around what the
necessary security infrastructure, if any, might need to be. There's some
work around the canonical way of finding non-firewalled peers and
distributing load among them. There might need to be work around how the DNS
entries get updated. And there might need to be some work around building a
big distributed dynamic DNS system for the case where a domain needs many
thousands of peer entries but can't afford or doesn't want to operate a
centralized cluster of those servers. I would point out, however, that the
last item is already an active research topic, and not really a P2P SIP
issue at all.

 

Using DNS as the protocol for the node-finding mechanism (and possibly as
the credential-lookup mechanism) means that inbound calls from existing SIP
clients should "just work", and that each domain can set its own policies
around whether the node-lookup (DNS) function is centralized or distributed
and whether the registrar function itself is centralized or distributed.
Then enabling zeroconf mDNS on top of that covers the case where you can't
talk to anyone but other clients on your ad-hoc network. And I think that's
100% of the enumerated use cases.

 

With no time spent on new signaling messages for outdated DHT architectures
as mandatory client components.

 

 

Matthew Kaufman

matthew at matthew.at

 

ps. To reflect on what Adam said about people who've built "real p2p
systems", we built a secure P2P VOIP softphone at amicima, and we spent a
lot of time thinking about per-domain policy for architecture (distributed
vs. centralized name to address lookup and NAT traversal functionality) and
security (centralized vs. distributed trust models, certificate signing,
etc.), which our MFPNet protocol supported. We used our secure media flow
protocol to transport the signaling messages, the audio, and the
simultaneous file transfers rather than pure SIP for signaling and RTP for
the audio (because we think that media prioritization is a big deal and that
even RTP is going to wish it had TCP-compatible congestion control some
day), but what we see here is essentially the same problem, and I too am
looking to make sure that what gets specified as the "P2P extensions to SIP"
is the simplest solution that will serve real user needs.

 

-------------- next part --------------
An HTML attachment was scrubbed...
URL: https://lists.cs.columbia.edu/pipermail/p2p-sip/attachments/20061128/04421453/attachment.html

From dbarrett at quinthar.com  Tue Nov 28 20:14:36 2006
From: dbarrett at quinthar.com (David Barrett)
Date: Tue, 28 Nov 2006 17:14:36 -0800
Subject: [p2p-sip] Fwd: what's wrong with DNS?
In-Reply-To: <f40963db0611281535i4bcaddfem1d885944a397c012@mail.gmail.com>
Message-ID: <031701c71353$bc853bd0$0200a8c0@Quinthar>

The outstanding question is in (2) and (5): how do I find a willing,
non-NAT'd volunteer SIP proxy?  (And the same question can be asked of STUN
and TURN servers.)  And for this task, a DHT might be well suited.
(Non-NAT'd nodes that are willing to volunteer their resources run a
standard SIP proxy and STUN/TURN server, and then register themselves in a
global DHT.  One is as good as any other, so a DHT is well-suited to this
task.)

 

But the user identification itself is done and enforced through DNS.  The
only way to steal dbarrett.quinthar.com <mailto:dbarrett at phone.quinthar.com>
from me is to steal my domain name, which already has legal and trademark
protections, international conflict resolutions, and so on.  

 

So wherever "I" am on the internet, I update "my" IP address for
dbarrett.quinthar.com using dynamic DNS.  This way I can run a webserver,
Jabber server, FTP server, SIP phone, or whatever else I want on my laptop,
and it follows me wherever I go.

 

There is of course the problem that most of the time my laptop is behind a
NAT, so the IP address stored in the DNS record is probably 192.168.xxx.yyy,
but at least if we're behind the same NAT you'll just be able to type in my
DNS name and connect to my computer.  (Though I could publish multiple IP
addresses in the DNS record - one for the LAN, one for my NAT.)

 

And this neatly extends to an mDNS world, thus enabling you to access me in
exactly the same way, whether we're on opposite sides of the planet via
internet, or sitting next to each other in the middle of a desert connected
by a 12" crossover cable.

 

We could even use other DNS records to publish my latest public key hash, my
human name, my real telephone number, and so on.

 

And all this could work *today*.

 

-david

 

  _____  

From: Adam Fisk [mailto:adamfisk at gmail.com] 
Sent: Tuesday, November 28, 2006 3:35 PM
To: Eunsoo Shim
Cc: David Barrett; p2p-sip at cs.columbia.edu
Subject: Re: [p2p-sip] Fwd: what's wrong with DNS?

 

David's breakdown described in best, earlier in this thread.  Here it is
again:

1)       Every user chooses a domain name.  Mine might be
'dbarrett.p2psip.org' 

2)       When I start my client up, I choose any non-firewalled peer (how do
I find it?) and REGISTER myself with it as something like
'dbarrett at dbarrett.p2psip.org'

3)       After registering with the peer/proxy, I contact some
highly-redundant set of dynamic DNS servers and say "this is the latest IP
address of my proxy, point to it whenever anyone does a DNS lookup on
dbarrett.p2psip.org"

4)       Your off the shelf, totally unaltered SIP client asks DNS for the
IP of 'dbarrett.p2psip.org', gets the latest IP of my proxy, and looks up
the user 'dbarrett at dbarrett.p2psip.org', and then contacts me as normal.

5)       If the peer I chose as my proxy goes down, I detect it immediately
and choose a new non-firewwalled peer (again, how do I find it?), REGISTER
with that as in step (2), and update my dynamic DNS binding as in step (3).

6)       Repeat.

 

On 11/28/06, Eunsoo Shim <eunsooshim at gmail.com> wrote:

I am not sure what your proposal is.
Could you provide more specifics on your proposal?
For example, are you suggesting that we use existing DNS servers to store
whatever we want to store at the P2P(SIP) network?
Or are you suggesting that we use the DNS PROTOCOL as (part of) the protocol
for P2P SIP and build a separate (global or ad-hoc) DNS-like infrastructure?
Thanks.

Eunsoo

 

On 11/28/06, David Barrett <dbarrett at quinthar.com> wrote:

I think Matthew highlights the Achilles heel of so-called "pure" P2P: it
never is.  In the real world, there's always a central service behind the
curtain:

- In Skype's case, it's a central directory at a hard-coded IP address. 

- In P2P-SIP's case (if Frank and Brian are right) is some kind of
centralized authorization/authentication/bootstrapping service.

So granted, there's the question of "how do you find this service?"  The 
obvious choice favored by every other application on the internet (and most
private networks) is "DNS".

But ignoring that, there's an even more important question:

        "Who controls this service?" 

If we build some fancy DHT solution we imply "some unnamed benevolent
dictator yet to be appointed" will rule us all.  We've sorta assumed this
(or similarly vague explanations) for the entire course of the P2P-SIP 
discussion.  For a long while I was not only on board, I was promoting this
"let's solve security later" approach.

But then Adam proposed a different owner, the IANA.  And the way to do that
is through the existing, tried-and-true interface of DNS (specifically, 
dynamic DNS).  I think this is obvious and brilliant.

Building a nonpartial, internationally represented, neutral global third
party isn't easy, and while the IANA hasn't wildly succeeded in this,
they're closer than anybody else (and decades ahead of anything we'd invent 
while searching for our benevolent dictator).


Now I'll grant DNS is limited to only, say, 99.999% of users.  And I
understand that there's a lot of interest in how the remaining 0.001% of
users will work without it.  But I'm having a hard time understanding why 
we're even discussing them.  Indeed, Matthew's repeated calls for a
real-world use case of a network that doesn't already have, won't quickly
get, or couldn't easily be configured to have DNS has remained unanswered. 
(And I'd add to that filter "outside of loud talking range.")

Furthermore, I'll grant you could build a DHT layer that serves as a
read-through cache of the central security mechanism that provides a faster 
response and propagation time than DNS.  But Akamai, UltraDNS, and other
massive, global organizations have proven that DNS can propagate pretty dang
quick.

And while I think it's an excellent, open question on how to improve DNS to 
make it even faster (perhaps with some kind of P2P layer), or how to make it
work better in partially or totally disconnected scenarios, or how to make
it more resistant to spamming/phishing and so forth, that seems like a much 
bigger problem than SIP alone and thus better suited to its own group.

And finally, I agree there are open questions as to how my car-of-the-future
will talk to my garage door and sprinkler system.  But why is this relevant?

I share Adam's concern that we're dreaming up a panacea to every conceivable
problem rather than standardizing a solution to problems that actually exist
with technologies that are already at hand.


I apologize for being obtuse, but dynamic DNS (even with today's 
limitations, and perhaps with future improvements) seems like such an
obvious way to solve the distributed name resolution problem that I'm
astonished it's being discounted so easily in favor of complex solutions to 
nebulous problems experienced by nonexistent userbases.

I'll admit the conclusion is anticlimactic -- I'm a P2P developer who likes
high tech as much as anybody.  But where has my reasoning fallen off track? 

-david

> -----Original Message-----
> From: p2p-sip-bounces at cs.columbia.edu [mailto:p2p-sip-
> bounces at cs.columbia.edu] On Behalf Of Matthew Kaufman
> Sent: Monday, November 27, 2006 10:18 AM
> To: hsinnrei at adobe.com; slavitch at gmail.com  <mailto:slavitch at gmail.com> 
> Cc: p2p-sip at cs.columbia.edu
> Subject: Re: [p2p-sip] Fwd: what's wrong with DNS?
> 
> Michael Slavitch:
> > As a form of assessing identity DNS is particularly weak and requires 
> > considerable care and expert feeding as provisioning is always
> > ultimately human-based requiring some kind of authority.
>
> Agreed. Though in small disconnected networks, zeroconf using mDNS avoids 
> much of this, and in large networks the IT guys or your ISP manage the
> complexity for you (though they haven't solved the security problem yet)
>
> > A self-organizing automated system needs no such crutches. 
>
> Except for bootstrapping to get the first two nodes talking. Which, if on
> the Internet, means a way to find IP addresses of those nodes. Which, on
> today's Internet, could use DNS. Or, as the Skype folks do, hardcoded 
> bootstrap
> addresses. But either way, bootstrap is such a crutch.
>
> > A key provision of P2PSIP is to eliminate the need for such
> > authorities in the first place because they are superfluous for the 
> > application.  An extended social network needs no central domain
> > administrator nor a central certificate authority.   It needs at most
> > a human facilitator not an engineer.
>
> So P2PSIP includes a web-of-trust security model? If so, why not leverage
> existing solutions for that as well (like PGP)?
>
> > Such social networks are not trite. For example social networks are 
> > the basis of all peer reviews for medical journals.  Shouting doesn't
> > work when 500 doctors are to be consulted, on several continents, with
> > varied access to the Internet, and over time.  The person running the 
> > peer review is a journal editor, a person who uses end-user tools, not
> > a network engineer.  DNS isn't the selection criteria here, reputation
> > is.
>
> But DNS is how the journal's mailing list manager finds the mail hosts 
> for each of the 500 doctors. Delivery of the article and reply comments
> across several continents over time works today just fine without P2P
> or DHTs or reputation-based web-of-trust security models. See this mailing

> list as an existence proof.
>
> Matthew Kaufman
> matthew at matthew.at
>
> _______________________________________________ 
> p2p-sip mailing list 
> p2p-sip at cs.columbia.edu
> https://lists.cs.columbia.edu/cucslists/listinfo/p2p-sip

_______________________________________________
p2p-sip mailing list
p2p-sip at cs.columbia.edu
https://lists.cs.columbia.edu/cucslists/listinfo/p2p-sip
<https://lists.cs.columbia.edu/cucslists/listinfo/p2p-sip> 






_______________________________________________
p2p-sip mailing list
p2p-sip at cs.columbia.edu
https://lists.cs.columbia.edu/cucslists/listinfo/p2p-sip



 

-------------- next part --------------
An HTML attachment was scrubbed...
URL: https://lists.cs.columbia.edu/pipermail/p2p-sip/attachments/20061128/b72c4b80/attachment-0001.html

From hgs at cs.columbia.edu  Tue Nov 28 21:39:55 2006
From: hgs at cs.columbia.edu (Henning Schulzrinne)
Date: Tue, 28 Nov 2006 21:39:55 -0500
Subject: [p2p-sip] Fwd: what's wrong with DNS?
In-Reply-To: <f40963db0611281520p6aa81677r62132b4a86e7cc1a@mail.gmail.com>
References: <200611271818.kARII0191955@where.matthew.at>
	<015301c712b9$5daeb950$0200a8c0@Quinthar>
	<f40963db0611281520p6aa81677r62132b4a86e7cc1a@mail.gmail.com>
Message-ID: <76E67E10-17AA-4616-AE3F-83687EBDC924@cs.columbia.edu>

For what it's worth, there was some discussion on the list and in  
meetings that mDNS and similar approaches could well be valuable, but  
that existing venues, such as SIPPING, would be sufficient to discuss  
the likely minor details that need to be specified. Indeed, we're  
working on such a specification. (One of the tactical problems is  
that mDNS is not an IETF standard and unlikely to become one in its  
current form, if I'm judging the zeroconf state correctly. Somebody  
with more familiarity with the group's fate can chime in here, as  
zeroconf is no longer an active WG.)

The main problem I see with DNS is more practical: setting up an  
authoritative DNS server correctly seems to require a fair amount of  
deep networking knowledge. Words like "lame delegation" come to mind.  
(This gets worse since you are supposed to set up at least two such  
servers that synchronize with each other, in independent networks.)  
Whether that's a failure of current implementations or something more  
fundamental, I don't know. One of the attractions of the DHT model is  
that you can set up a local system by just installing software on a  
laptop, plus maybe a credentialing system for larger systems with  
potentially malicious users.

While every enterprise runs a DNS server, I'd wager that the large  
majority are not set up to allow individual users (as opposed to,  
say, the DHCP server) to update individual DNS entries. This isn't an  
issue for a carrier wanting to build a public service, but then I'm  
confused as to why such a carrier doesn't just use the current SIP  
server model. Just saving a few registrar boxes doesn't seem worth  
the hassle.

In general, nobody keeps people from doing work. I personally think  
that having a team of interested individuals work out what needs to  
happen for dynamic DNS to work on a realistic scale would be quite  
interesting. Depending on the details, this work could well find a  
home in SIPPING or a DNS-related WG, or maybe neither, if this is all  
just combining existing protocols.

The ability to find willing registrars and STUN servers is likely to  
be shared by any DHT or DNS solution, although a simpler centralized  
registry (e.g., via SLP, see our work on domain-based SLP) is also  
feasible.

Henning

On Nov 28, 2006, at 6:20 PM, Adam Fisk wrote:

> I couldn't have said it better, David (I probably would have said  
> it much worse!).  I have yet to hear any arguments against this  
> approach compelling enough to warrant the complexity of the  
> proposed solutions.  I continue to think the solution is not  
> appealing to people because it's too simple and not sexy enough,  
> regardless of real-world applicability -- it's anticlimactic, as  
> you said.
>
> I unfortunately don't have the incentive or the time to keep  
> pushing this perspective, though.  If the decisions of this list  
> actually affected me in some way, I'd argue this thing to the  
> ground, because I think it's clearly the winning idea.  It's tragic  
> for decisions at this high a level to be decided by attrition  
> rather than technical merit, but that appears to be what's  
> happening barring some groundswell of support from the group's core  
> participants.  Maybe there's some compelling argument out there  
> against the DNS solution for the connected-internet case, but I  
> have yet to hear it.
>
> There's a lot of talk about what happens in "real p2p systems."   
> Both David and I have a great deal of experience with these  
> systems, folks (core developers of LimeWire/Last Bamboo and  
> RedSwoosh/iGlance).  We don't need convincing about the merits of  
> DHTs, the various bootstrapping solutions out there, etc etc.  I  
> think it's interesting, though, that the p2p guys are the ones  
> arguing for the more "centralized" approach.  It's in the interest  
> of seeing the work here deployed on a meaningful scale, which to me  
> means implementing the simplest solution that will work.
>
> -Adam
>
>
>
> _______________________________________________
> p2p-sip mailing list
> p2p-sip at cs.columbia.edu
> https://lists.cs.columbia.edu/cucslists/listinfo/p2p-sip


From dbarrett at quinthar.com  Wed Nov 29 02:29:35 2006
From: dbarrett at quinthar.com (David Barrett)
Date: Tue, 28 Nov 2006 23:29:35 -0800
Subject: [p2p-sip] Fwd: what's wrong with DNS?
In-Reply-To: <76E67E10-17AA-4616-AE3F-83687EBDC924@cs.columbia.edu>
Message-ID: <002401c71388$27ab3480$0202a8c0@Quinthar>

> -----Original Message-----
> From: Henning Schulzrinne [mailto:hgs at cs.columbia.edu]
> Subject: Re: [p2p-sip] Fwd: what's wrong with DNS?
> 
> (One of the tactical problems is
> that mDNS is not an IETF standard and unlikely to become one in its
> current form, if I'm judging the zeroconf state correctly. Somebody
> with more familiarity with the group's fate can chime in here, as
> zeroconf is no longer an active WG.)

mDNS appears to be on the standards track:

http://files.multicastdns.org/draft-cheshire-dnsext-multicastdns.txt

Submitted 2006/8/10, expires 2007/2/10.


> The main problem I see with DNS is more practical: setting up an
> authoritative DNS server correctly seems to require a fair amount of
> deep networking knowledge. ...
> Whether that's a failure of current implementations or something more
> fundamental, I don't know.

Good question.  It might be hard merely because there's been no reason to
make it easy, until now.  There seems no fundamental reason why it need be
complex.

-david




From dbarrett at quinthar.com  Wed Nov 29 03:49:05 2006
From: dbarrett at quinthar.com (David Barrett)
Date: Wed, 29 Nov 2006 00:49:05 -0800
Subject: [p2p-sip] Fwd: what's wrong with DNS?
In-Reply-To: <002501c7134d$78ab4bd0$029b0a0a@matthewdeskxps>
Message-ID: <004a01c71393$3a5d1ed0$0202a8c0@Quinthar>

I completely agree with Matthew here.  Basing this on DNS gives the domain
owner extreme flexibility over how he wants to manage the users on that
domain.  For example:

 

-          A large provider (eg, Yahoo) might offer a standard public SIP
proxy for millions of general users at *@yahoo.com



-          A large company (eg, Intel) might manage its own private SIP
proxy for thousands of employees *@intel.com



-          A small company on a DSL connection might have a simple desktop
computer running a standard SIP proxy a dozen employees at *@smallco.com,
updating the IP of the proxy using a standard dynamic DNS client.



-          An individual, fixed user might have a static IP on his
cablemodem and thus run both a SIP proxy and SIP client on the same computer
for dbarrett at dbarrett.quinthar.com



-          An individual, mobile user might have a laptop running an
integrated SIP-proxy/SIP-client/dynamic-DNS/mDNS client that updates the
dbarrett.quinthar.com <mailto:dbarrett at dbarrett.quinthar.com>  record to
wherever the laptop is right now (perhaps using a DHT to find volunteer
anonymous SIP proxies and STUN/TURN servers if behind a NAT), and that
advertises itself on dbarrett at dbarrett.quinthar.com.local for totally
infrastructureless, ad-hoc environments

 

This spans the entire range of the largest to the smallest, the most fixed
to the most mobile environment.  With the exception of some kind of DHT to
find a cloud of public resources (ie, SIP proxies that will let anybody
register with them from any domain, as well as STUN and TURN servers),
everything is done with existing technologies.

 

Furthermore, only the "individual, fixed" and "individual, mobile" use cases
require anything that can't be deployed off the shelf right now, and even
that is just a repackaging of existing technologies (with the exception of
the public-resource DHT, if we choose to go that route).  Indeed, it's not
hard to imagine a well-packaged SIP client that has an install wizard
saying:

 

What is your SIP address?

a)       I already have an address assigned to me (eg, dbarrett at yahoo.com)

b)      I have no address, please help me create one

 

(Choose B) Every address consist of a username and a domain.  What domain do
you want to use?

a)       I want to use an existing domain (choose from a list of public
providers, eg @yahoo.com)

b)      I want to buy a new domain (choose from a list of domain providers)

c)       I want to get a free domain (choose from a list of free providers,
such as DynDNS)

d)      I want to use a domain I already own (enter the domain)

 

(Choose D) Testing. dbarrett.quinthar.com is registered with 'DynDNS, Inc',
which is a recognized dynamic DNS provider.  Enter your DynDNS name and
password and we will automatically update dbarrett.quinthar.com to be
associated with this computer, even if it changes networks.

 

(Enter name/pw) Testing. we've confirmed our ability to update
dbarrett.quinthar.com.  Your new SIP address is
dbarrett at dbarrett.quinthar.com. 

 

Basically, make the client aware of dynamic DNS, use the standard dynamic
DNS update protocols, and it can be made really easy for the user.
Similarly, this SIP client could act as an mDNS server (on OS's that don't
already support it) and thus autodetect other P2P-SIP clients on an ad-hoc
network.  

 

It can be totally self-configuring and seamlessly scalable from huge to tiny
deployments, without requiring a benevolent dictator or inventing a
replacement for the IANA.

 

-david

 

  _____  

From: p2p-sip-bounces at cs.columbia.edu
[mailto:p2p-sip-bounces at cs.columbia.edu] On Behalf Of Matthew Kaufman
Sent: Tuesday, November 28, 2006 4:30 PM
To: 'Adam Fisk'; 'Eunsoo Shim'
Cc: p2p-sip at cs.columbia.edu
Subject: Re: [p2p-sip] Fwd: what's wrong with DNS?

 

Adam Fisk:

David's breakdown described in best, earlier in this thread.  Here it is
again:

1)       Every user chooses a domain name.  Mine might be
'dbarrett.p2psip.org' 

2)       When I start my client up, I choose any non-firewalled peer (how do
I find it?) and REGISTER myself with it as something like
'dbarrett at dbarrett.p2psip.org'

3)       After registering with the peer/proxy, I contact some
highly-redundant set of dynamic DNS servers and say "this is the latest IP
address of my proxy, point to it whenever anyone does a DNS lookup on
dbarrett.p2psip.org"

4)       Your off the shelf, totally unaltered SIP client asks DNS for the
IP of 'dbarrett.p2psip.org', gets the latest IP of my proxy, and looks up
the user 'dbarrett at dbarrett.p2psip.org', and then contacts me as normal.

5)       If the peer I chose as my proxy goes down, I detect it immediately
and choose a new non-firewwalled peer (again, how do I find it?), REGISTER
with that as in step (2), and update my dynamic DNS binding as in step (3).

6)       Repeat.

 

I would add to this that if you are yourself not firewalled, you should be
able to REGISTER with "yourself" and your machine should be able to directly
process inbound calls that arrive.

 

If you are firewalled, clearly a node is needed that can relay call setup
messages to you, and pretty clearly those should just be normal SIP
messages. There's a good argument for allowing one to register with more
than one at a time and point the dynamic DNS appropriately.

 

Additionally, you should also answer mDNS queries for you.p2psip.org with
your local self-proxy, so that machines on your LAN can call you directly
whether or not the DNS is reachable, if their SIP clients are mDNS-aware.

 

As for your open issue of finding a non-firewalled peer to register with?
Use the DNS to get an address of one inside of p2psip.org (if that's the
domain within which you're operating) Note that this covers the case of an
organization that wants the clients registered with a stable machine or
machines that it operates instead of with client machines acting as
registrars. The organization can also set policy around what addresses can
be registered, if credentials are required, etc.

 

The real work that needs doing is coming up with a minimal set of
recommended existing standards plus the minimal extensions necessary such
that a working system can be built. Clearly STUN or better is needed for NAT
traversal. Clearly SIP itself is the call signaling, though extensions might
enable neat P2P capabilities. There's some work around using mDNS for the
detached-from-real-Internet cases. There's some work around what the
necessary security infrastructure, if any, might need to be. There's some
work around the canonical way of finding non-firewalled peers and
distributing load among them. There might need to be work around how the DNS
entries get updated. And there might need to be some work around building a
big distributed dynamic DNS system for the case where a domain needs many
thousands of peer entries but can't afford or doesn't want to operate a
centralized cluster of those servers. I would point out, however, that the
last item is already an active research topic, and not really a P2P SIP
issue at all.

 

Using DNS as the protocol for the node-finding mechanism (and possibly as
the credential-lookup mechanism) means that inbound calls from existing SIP
clients should "just work", and that each domain can set its own policies
around whether the node-lookup (DNS) function is centralized or distributed
and whether the registrar function itself is centralized or distributed.
Then enabling zeroconf mDNS on top of that covers the case where you can't
talk to anyone but other clients on your ad-hoc network. And I think that's
100% of the enumerated use cases.

 

With no time spent on new signaling messages for outdated DHT architectures
as mandatory client components.

 

 

Matthew Kaufman

matthew at matthew.at

 

ps. To reflect on what Adam said about people who've built "real p2p
systems", we built a secure P2P VOIP softphone at amicima, and we spent a
lot of time thinking about per-domain policy for architecture (distributed
vs. centralized name to address lookup and NAT traversal functionality) and
security (centralized vs. distributed trust models, certificate signing,
etc.), which our MFPNet protocol supported. We used our secure media flow
protocol to transport the signaling messages, the audio, and the
simultaneous file transfers rather than pure SIP for signaling and RTP for
the audio (because we think that media prioritization is a big deal and that
even RTP is going to wish it had TCP-compatible congestion control some
day), but what we see here is essentially the same problem, and I too am
looking to make sure that what gets specified as the "P2P extensions to SIP"
is the simplest solution that will serve real user needs.

 

-------------- next part --------------
An HTML attachment was scrubbed...
URL: https://lists.cs.columbia.edu/pipermail/p2p-sip/attachments/20061129/8dc0b6cc/attachment-0001.html

From edpimentl at gmail.com  Wed Nov 29 04:33:05 2006
From: edpimentl at gmail.com (Ed Pimentel)
Date: Wed, 29 Nov 2006 04:33:05 -0500
Subject: [p2p-sip] what's wrong with DNS?
In-Reply-To: <1fc32c8f0611201914pcba511bka99ecbdaa65136b@mail.gmail.com>
References: <24CCCC428EFEA2469BF046DB3C7A8D22141340@namail5.corp.adobe.com>
	<1164071250.14261.4.camel@sioux.systems.cs.cornell.edu>
	<1fc32c8f0611201914pcba511bka99ecbdaa65136b@mail.gmail.com>
Message-ID: <E2849D71-44B9-451F-BFE2-0AB44D484DBB@gmail.com>

In vehement agreement on thinking of new more efficient methods to  
look up, besides DNS.
We do not have to view every problem like a nail and use the hammer(DNS)
However if a  kindler, gentler and 1000 points of light  - better/ 
faster/cheaper DNS is to be implemented
please look into steroid performance of CoDnS
http://www.cs.cornell.edu/people/egs/beehive/codons-sigcomm04/ 
codons.html


On Nov 20, 2006, at 10:14 PM, Michael Slavitch wrote:

> DNS was designed in the days when long-term store was scarce and so
> was bandwidth. Technology was difficult and obtuse, with the number of
> human implementors measured in the hundreds, and the number of users
> in the thousands. I remember that because I was there. There was no
> other way to pass this data. Something had to be first and that was
> DNS, which reflected the times and the organizations that invented it
> in the first place.
>

Sincerely,

Ed Pimentel
AgileCO.net

Mail:   edpimentl at gmail.com
sip:    sip:2150112 at truphone.com
IM:     edpimentl [AOL | Jabber | Yahoo | MSN ]
Voip:   edpimentl [SKype | GoogleTalk ]

Mobile Content Marketing/Management/Digital Delivery
http://mobilecentral.ws

Mobile Payment/P2P
http://agilepay.ws/ws

Mobile Social Network (Private Label)
http://GooGaYa.com



-------------- next part --------------
An HTML attachment was scrubbed...
URL: https://lists.cs.columbia.edu/pipermail/p2p-sip/attachments/20061129/333024e3/attachment.html

From br at brianrosen.net  Wed Nov 29 08:01:48 2006
From: br at brianrosen.net (Brian Rosen)
Date: Wed, 29 Nov 2006 08:01:48 -0500
Subject: [p2p-sip] Fwd: what's wrong with DNS?
In-Reply-To: <004a01c71393$3a5d1ed0$0202a8c0@Quinthar>
Message-ID: <013801c713b6$9011f4e0$640fa8c0@cis.neustar.com>

<not as chair, although at present, I?m not chair of anything>
I suggest that you take Henning?s advice, write up an Internet Draft, and
submit it to sipping.  
It may be an interesting enough idea to enough folks that you can get an RFC
out of it.

It does not meet the charter of p2psip as determined by list and BOF
consensus.  It really is a different set of ideas.  Those ideas may have
merit, they just don?t fit here.

As a suggestion if you do decide to proceed:
Consider defining the idea as a DDDS application
Consider that dbarrett at dbarrett.quinthar.com is not really what you want,
you want dbarrett at quinthar.com.  You want that sip address to resolve to a
host who is named dbarret.quinthar.com for this purpose.  A DDDS application
could start with that string, convert it to dbarrett.quinthar.com, and do a
DNS query for a NAPTR that got you what you wanted.
You might want something in the quinthar.com domain that indicates that it
participates in the scheme you describe, since a normal sip client getting a
request for dbarrett at quinthar.com would try to find the sip proxy at
quinthar.com.  If quinthar.com doesn't operate a sip proxy, but instead has
a resolution scheme like you imagine, you might want it to say so
explicitly.

Please do consider the security implications of this proposal.  I think
those are the real Achilles heel of the idea.  Proxy servers provide a
useful capability that controls incoming calls.  Our current ideas work
pretty much like a "Closed User Group", using a credential to authorize
placement of a call within the CUG.  We may need more controls that that.

Brian


________________________________________
From: p2p-sip-bounces at cs.columbia.edu
[mailto:p2p-sip-bounces at cs.columbia.edu] On Behalf Of David Barrett
Sent: Wednesday, November 29, 2006 3:49 AM
To: matthew at matthew.at; 'Adam Fisk'; 'Eunsoo Shim'
Cc: p2p-sip at cs.columbia.edu
Subject: Re: [p2p-sip] Fwd: what's wrong with DNS?

I completely agree with Matthew here.? Basing this on DNS gives the domain
owner extreme flexibility over how he wants to manage the users on that
domain.? For example:

- A large provider (eg, Yahoo) might offer a standard public SIP proxy for
millions of general users at *@yahoo.com
- A large company (eg, Intel) might manage its own private SIP proxy for
thousands of employees *@intel.com
- A small company on a DSL connection might have a simple desktop computer
running a standard SIP proxy a dozen employees at *@smallco.com, updating
the IP of the proxy using a standard dynamic DNS client.
- An individual, fixed user might have a static IP on his cablemodem and
thus run both a SIP proxy and SIP client on the same computer for
dbarrett at dbarrett.quinthar.com
- An individual, mobile user might have a laptop running an integrated
SIP-proxy/SIP-client/dynamic-DNS/mDNS client that updates the
dbarrett.quinthar.com record to wherever the laptop is right now (perhaps
using a DHT to find volunteer anonymous SIP proxies and STUN/TURN servers if
behind a NAT), and that advertises itself on
dbarrett at dbarrett.quinthar.com.local for totally infrastructureless, ad-hoc
environments

This spans the entire range of the largest to the smallest, the most fixed
to the most mobile environment.? With the exception of some kind of DHT to
find a cloud of public resources (ie, SIP proxies that will let anybody
register with them from any domain, as well as STUN and TURN servers),
everything is done with existing technologies.

Furthermore, only the ?individual, fixed? and ?individual, mobile? use cases
require anything that can?t be deployed off the shelf right now, and even
that is just a repackaging of existing technologies (with the exception of
the public-resource DHT, if we choose to go that route).? Indeed, it?s not
hard to imagine a well-packaged SIP client that has an install wizard
saying:

What is your SIP address?
a) I already have an address assigned to me (eg, dbarrett at yahoo.com)
b) I have no address, please help me create one

(Choose B) Every address consist of a username and a domain.? What domain do
you want to use?
a) I want to use an existing domain (choose from a list of public providers,
eg @yahoo.com)
b) I want to buy a new domain (choose from a list of domain providers)
c) I want to get a free domain (choose from a list of free providers, such
as DynDNS)
d) I want to use a domain I already own (enter the domain)

(Choose D) Testing
 dbarrett.quinthar.com is registered with ?DynDNS, Inc?,
which is a recognized dynamic DNS provider.? Enter your DynDNS name and
password and we will automatically update dbarrett.quinthar.com to be
associated with this computer, even if it changes networks.

(Enter name/pw) Testing
 we?ve confirmed our ability to update
dbarrett.quinthar.com.? Your new SIP address is
dbarrett at dbarrett.quinthar.com. 

Basically, make the client aware of dynamic DNS, use the standard dynamic
DNS update protocols, and it can be made really easy for the user.?
Similarly, this SIP client could act as an mDNS server (on OS?s that don?t
already support it) and thus autodetect other P2P-SIP clients on an ad-hoc
network.? 

It can be totally self-configuring and seamlessly scalable from huge to tiny
deployments, without requiring a benevolent dictator or inventing a
replacement for the IANA.

-david

________________________________________
From: p2p-sip-bounces at cs.columbia.edu
[mailto:p2p-sip-bounces at cs.columbia.edu] On Behalf Of Matthew Kaufman
Sent: Tuesday, November 28, 2006 4:30 PM
To: 'Adam Fisk'; 'Eunsoo Shim'
Cc: p2p-sip at cs.columbia.edu
Subject: Re: [p2p-sip] Fwd: what's wrong with DNS?

Adam Fisk:
David's breakdown described in best, earlier in this thread.? Here it is
again:
1)?????? Every user chooses a domain name.? Mine might be
'dbarrett.p2psip.org' 
2)?????? When I start my client up, I choose any non-firewalled peer (how do
I find it?) and REGISTER myself with it as something like
'dbarrett at dbarrett.p2psip.org'
3)?????? After registering with the peer/proxy, I contact some
highly-redundant set of dynamic DNS servers and say "this is the latest IP
address of my proxy, point to it whenever anyone does a DNS lookup on
dbarrett.p2psip.org"
4)?????? Your off the shelf, totally unaltered SIP client asks DNS for the
IP of 'dbarrett.p2psip.org', gets the latest IP of my proxy, and looks up
the user 'dbarrett at dbarrett.p2psip.org', and then contacts me as normal.
5)?????? If the peer I chose as my proxy goes down, I detect it immediately
and choose a new non-firewwalled peer (again, how do I find it?), REGISTER
with that as in step (2), and update my dynamic DNS binding as in step (3).
6)?????? Repeat.

I would add to this that if you are yourself not firewalled, you should be
able to REGISTER with ?yourself? and your machine should be able to directly
process inbound calls that arrive.

If you are firewalled, clearly a node is needed that can relay call setup
messages to you, and pretty clearly those should just be normal SIP
messages. There?s a good argument for allowing one to register with more
than one at a time and point the dynamic DNS appropriately.

Additionally, you should also answer mDNS queries for you.p2psip.org with
your local self-proxy, so that machines on your LAN can call you directly
whether or not the DNS is reachable, if their SIP clients are mDNS-aware.

As for your open issue of finding a non-firewalled peer to register with?
Use the DNS to get an address of one inside of p2psip.org (if that?s the
domain within which you?re operating) Note that this covers the case of an
organization that wants the clients registered with a stable machine or
machines that it operates instead of with client machines acting as
registrars. The organization can also set policy around what addresses can
be registered, if credentials are required, etc.

The real work that needs doing is coming up with a minimal set of
recommended existing standards plus the minimal extensions necessary such
that a working system can be built. Clearly STUN or better is needed for NAT
traversal. Clearly SIP itself is the call signaling, though extensions might
enable neat P2P capabilities. There?s some work around using mDNS for the
detached-from-real-Internet cases. There?s some work around what the
necessary security infrastructure, if any, might need to be. There?s some
work around the canonical way of finding non-firewalled peers and
distributing load among them. There might need to be work around how the DNS
entries get updated. And there might need to be some work around building a
big distributed dynamic DNS system for the case where a domain needs many
thousands of peer entries but can?t afford or doesn?t want to operate a
centralized cluster of those servers. I would point out, however, that the
last item is already an active research topic, and not really a P2P SIP
issue at all.

Using DNS as the protocol for the node-finding mechanism (and possibly as
the credential-lookup mechanism) means that inbound calls from existing SIP
clients should ?just work?, and that each domain can set its own policies
around whether the node-lookup (DNS) function is centralized or distributed
and whether the registrar function itself is centralized or distributed.
Then enabling zeroconf mDNS on top of that covers the case where you can?t
talk to anyone but other clients on your ad-hoc network. And I think that?s
100% of the enumerated use cases.

With no time spent on new signaling messages for outdated DHT architectures
as mandatory client components.


Matthew Kaufman
matthew at matthew.at

ps. To reflect on what Adam said about people who?ve built ?real p2p
systems?, we built a secure P2P VOIP softphone at amicima, and we spent a
lot of time thinking about per-domain policy for architecture (distributed
vs. centralized name to address lookup and NAT traversal functionality) and
security (centralized vs. distributed trust models, certificate signing,
etc.), which our MFPNet protocol supported. We used our secure media flow
protocol to transport the signaling messages, the audio, and the
simultaneous file transfers rather than pure SIP for signaling and RTP for
the audio (because we think that media prioritization is a big deal and that
even RTP is going to wish it had TCP-compatible congestion control some
day), but what we see here is essentially the same problem, and I too am
looking to make sure that what gets specified as the ?P2P extensions to SIP?
is the simplest solution that will serve real user needs.




From hsinnrei at adobe.com  Wed Nov 29 09:08:24 2006
From: hsinnrei at adobe.com (Henry Sinnreich)
Date: Wed, 29 Nov 2006 06:08:24 -0800
Subject: [p2p-sip] Fwd: what's wrong with DNS?
In-Reply-To: <013801c713b6$9011f4e0$640fa8c0@cis.neustar.com>
Message-ID: <24CCCC428EFEA2469BF046DB3C7A8D22141384@namail5.corp.adobe.com>

Full agreement.

This list was created for P2P SIP as hammered out in the existing charter.

Other good ideas and approaches must start with an I-D that people can discuss.

I am personally using dyndns for my home SIP service and am sure new DNS approaches deserve a better status as IETF WGs. Henning's proposed approaches for such work is right. There is no need to tie up this WG with another topic.

Thanks, Henry

-----Original Message-----
From: p2p-sip-bounces at cs.columbia.edu [mailto:p2p-sip-bounces at cs.columbia.edu] On Behalf Of Brian Rosen
Sent: Wednesday, November 29, 2006 7:02 AM
To: 'David Barrett'; matthew at matthew.at; 'Adam Fisk'; 'Eunsoo Shim'
Cc: p2p-sip at cs.columbia.edu
Subject: Re: [p2p-sip] Fwd: what's wrong with DNS?

<not as chair, although at present, I'm not chair of anything>
I suggest that you take Henning's advice, write up an Internet Draft, and
submit it to sipping.  
It may be an interesting enough idea to enough folks that you can get an RFC
out of it.

It does not meet the charter of p2psip as determined by list and BOF
consensus.  It really is a different set of ideas.  Those ideas may have
merit, they just don't fit here.

As a suggestion if you do decide to proceed:
Consider defining the idea as a DDDS application
Consider that dbarrett at dbarrett.quinthar.com is not really what you want,
you want dbarrett at quinthar.com.  You want that sip address to resolve to a
host who is named dbarret.quinthar.com for this purpose.  A DDDS application
could start with that string, convert it to dbarrett.quinthar.com, and do a
DNS query for a NAPTR that got you what you wanted.
You might want something in the quinthar.com domain that indicates that it
participates in the scheme you describe, since a normal sip client getting a
request for dbarrett at quinthar.com would try to find the sip proxy at
quinthar.com.  If quinthar.com doesn't operate a sip proxy, but instead has
a resolution scheme like you imagine, you might want it to say so
explicitly.

Please do consider the security implications of this proposal.  I think
those are the real Achilles heel of the idea.  Proxy servers provide a
useful capability that controls incoming calls.  Our current ideas work
pretty much like a "Closed User Group", using a credential to authorize
placement of a call within the CUG.  We may need more controls that that.

Brian


________________________________________
From: p2p-sip-bounces at cs.columbia.edu
[mailto:p2p-sip-bounces at cs.columbia.edu] On Behalf Of David Barrett
Sent: Wednesday, November 29, 2006 3:49 AM
To: matthew at matthew.at; 'Adam Fisk'; 'Eunsoo Shim'
Cc: p2p-sip at cs.columbia.edu
Subject: Re: [p2p-sip] Fwd: what's wrong with DNS?

I completely agree with Matthew here.? Basing this on DNS gives the domain
owner extreme flexibility over how he wants to manage the users on that
domain.? For example:

- A large provider (eg, Yahoo) might offer a standard public SIP proxy for
millions of general users at *@yahoo.com
- A large company (eg, Intel) might manage its own private SIP proxy for
thousands of employees *@intel.com
- A small company on a DSL connection might have a simple desktop computer
running a standard SIP proxy a dozen employees at *@smallco.com, updating
the IP of the proxy using a standard dynamic DNS client.
- An individual, fixed user might have a static IP on his cablemodem and
thus run both a SIP proxy and SIP client on the same computer for
dbarrett at dbarrett.quinthar.com
- An individual, mobile user might have a laptop running an integrated
SIP-proxy/SIP-client/dynamic-DNS/mDNS client that updates the
dbarrett.quinthar.com record to wherever the laptop is right now (perhaps
using a DHT to find volunteer anonymous SIP proxies and STUN/TURN servers if
behind a NAT), and that advertises itself on
dbarrett at dbarrett.quinthar.com.local for totally infrastructureless, ad-hoc
environments

This spans the entire range of the largest to the smallest, the most fixed
to the most mobile environment.? With the exception of some kind of DHT to
find a cloud of public resources (ie, SIP proxies that will let anybody
register with them from any domain, as well as STUN and TURN servers),
everything is done with existing technologies.

Furthermore, only the "individual, fixed" and "individual, mobile" use cases
require anything that can't be deployed off the shelf right now, and even
that is just a repackaging of existing technologies (with the exception of
the public-resource DHT, if we choose to go that route).? Indeed, it's not
hard to imagine a well-packaged SIP client that has an install wizard
saying:

What is your SIP address?
a) I already have an address assigned to me (eg, dbarrett at yahoo.com)
b) I have no address, please help me create one

(Choose B) Every address consist of a username and a domain.? What domain do
you want to use?
a) I want to use an existing domain (choose from a list of public providers,
eg @yahoo.com)
b) I want to buy a new domain (choose from a list of domain providers)
c) I want to get a free domain (choose from a list of free providers, such
as DynDNS)
d) I want to use a domain I already own (enter the domain)

(Choose D) Testing... dbarrett.quinthar.com is registered with 'DynDNS, Inc',
which is a recognized dynamic DNS provider.? Enter your DynDNS name and
password and we will automatically update dbarrett.quinthar.com to be
associated with this computer, even if it changes networks.

(Enter name/pw) Testing... we've confirmed our ability to update
dbarrett.quinthar.com.? Your new SIP address is
dbarrett at dbarrett.quinthar.com. 

Basically, make the client aware of dynamic DNS, use the standard dynamic
DNS update protocols, and it can be made really easy for the user.?
Similarly, this SIP client could act as an mDNS server (on OS's that don't
already support it) and thus autodetect other P2P-SIP clients on an ad-hoc
network.? 

It can be totally self-configuring and seamlessly scalable from huge to tiny
deployments, without requiring a benevolent dictator or inventing a
replacement for the IANA.

-david

________________________________________
From: p2p-sip-bounces at cs.columbia.edu
[mailto:p2p-sip-bounces at cs.columbia.edu] On Behalf Of Matthew Kaufman
Sent: Tuesday, November 28, 2006 4:30 PM
To: 'Adam Fisk'; 'Eunsoo Shim'
Cc: p2p-sip at cs.columbia.edu
Subject: Re: [p2p-sip] Fwd: what's wrong with DNS?

Adam Fisk:
David's breakdown described in best, earlier in this thread.? Here it is
again:
1)?????? Every user chooses a domain name.? Mine might be
'dbarrett.p2psip.org' 
2)?????? When I start my client up, I choose any non-firewalled peer (how do
I find it?) and REGISTER myself with it as something like
'dbarrett at dbarrett.p2psip.org'
3)?????? After registering with the peer/proxy, I contact some
highly-redundant set of dynamic DNS servers and say "this is the latest IP
address of my proxy, point to it whenever anyone does a DNS lookup on
dbarrett.p2psip.org"
4)?????? Your off the shelf, totally unaltered SIP client asks DNS for the
IP of 'dbarrett.p2psip.org', gets the latest IP of my proxy, and looks up
the user 'dbarrett at dbarrett.p2psip.org', and then contacts me as normal.
5)?????? If the peer I chose as my proxy goes down, I detect it immediately
and choose a new non-firewwalled peer (again, how do I find it?), REGISTER
with that as in step (2), and update my dynamic DNS binding as in step (3).
6)?????? Repeat.

I would add to this that if you are yourself not firewalled, you should be
able to REGISTER with "yourself" and your machine should be able to directly
process inbound calls that arrive.

If you are firewalled, clearly a node is needed that can relay call setup
messages to you, and pretty clearly those should just be normal SIP
messages. There's a good argument for allowing one to register with more
than one at a time and point the dynamic DNS appropriately.

Additionally, you should also answer mDNS queries for you.p2psip.org with
your local self-proxy, so that machines on your LAN can call you directly
whether or not the DNS is reachable, if their SIP clients are mDNS-aware.

As for your open issue of finding a non-firewalled peer to register with?
Use the DNS to get an address of one inside of p2psip.org (if that's the
domain within which you're operating) Note that this covers the case of an
organization that wants the clients registered with a stable machine or
machines that it operates instead of with client machines acting as
registrars. The organization can also set policy around what addresses can
be registered, if credentials are required, etc.

The real work that needs doing is coming up with a minimal set of
recommended existing standards plus the minimal extensions necessary such
that a working system can be built. Clearly STUN or better is needed for NAT
traversal. Clearly SIP itself is the call signaling, though extensions might
enable neat P2P capabilities. There's some work around using mDNS for the
detached-from-real-Internet cases. There's some work around what the
necessary security infrastructure, if any, might need to be. There's some
work around the canonical way of finding non-firewalled peers and
distributing load among them. There might need to be work around how the DNS
entries get updated. And there might need to be some work around building a
big distributed dynamic DNS system for the case where a domain needs many
thousands of peer entries but can't afford or doesn't want to operate a
centralized cluster of those servers. I would point out, however, that the
last item is already an active research topic, and not really a P2P SIP
issue at all.

Using DNS as the protocol for the node-finding mechanism (and possibly as
the credential-lookup mechanism) means that inbound calls from existing SIP
clients should "just work", and that each domain can set its own policies
around whether the node-lookup (DNS) function is centralized or distributed
and whether the registrar function itself is centralized or distributed.
Then enabling zeroconf mDNS on top of that covers the case where you can't
talk to anyone but other clients on your ad-hoc network. And I think that's
100% of the enumerated use cases.

With no time spent on new signaling messages for outdated DHT architectures
as mandatory client components.


Matthew Kaufman
matthew at matthew.at

ps. To reflect on what Adam said about people who've built "real p2p
systems", we built a secure P2P VOIP softphone at amicima, and we spent a
lot of time thinking about per-domain policy for architecture (distributed
vs. centralized name to address lookup and NAT traversal functionality) and
security (centralized vs. distributed trust models, certificate signing,
etc.), which our MFPNet protocol supported. We used our secure media flow
protocol to transport the signaling messages, the audio, and the
simultaneous file transfers rather than pure SIP for signaling and RTP for
the audio (because we think that media prioritization is a big deal and that
even RTP is going to wish it had TCP-compatible congestion control some
day), but what we see here is essentially the same problem, and I too am
looking to make sure that what gets specified as the "P2P extensions to SIP"
is the simplest solution that will serve real user needs.



_______________________________________________
p2p-sip mailing list
p2p-sip at cs.columbia.edu
https://lists.cs.columbia.edu/cucslists/listinfo/p2p-sip


From tme at multicasttech.com  Wed Nov 29 14:04:45 2006
From: tme at multicasttech.com (Marshall Eubanks)
Date: Wed, 29 Nov 2006 14:04:45 -0500
Subject: [p2p-sip] Fwd: what's wrong with DNS?
In-Reply-To: <002401c71388$27ab3480$0202a8c0@Quinthar>
References: <002401c71388$27ab3480$0202a8c0@Quinthar>
Message-ID: <F30D5A23-2AC9-404D-9537-5AAB8F454026@multicasttech.com>

Hello;

On Nov 29, 2006, at 2:29 AM, David Barrett wrote:

>> -----Original Message-----
>> From: Henning Schulzrinne [mailto:hgs at cs.columbia.edu]
>> Subject: Re: [p2p-sip] Fwd: what's wrong with DNS?
>>
>> (One of the tactical problems is
>> that mDNS is not an IETF standard and unlikely to become one in its
>> current form, if I'm judging the zeroconf state correctly. Somebody
>> with more familiarity with the group's fate can chime in here, as
>> zeroconf is no longer an active WG.)
>
> mDNS appears to be on the standards track:
>
> http://files.multicastdns.org/draft-cheshire-dnsext-multicastdns.txt

There are two variants of mDNS, as I understand it.

There is the Stuart Cheshire version that is not a WG product, but
is standards track, and is used :

http://tools.ietf.org/html/draft-cheshire-dnsext-multicastdns-06

Then there is the official WG product,
http://www.ietf.org/internet-drafts/draft-ietf-dnsext-mdns-47.txt

which may or may not be deployed. (There was a long email discussion  
of this forking of the
mDNS effort, I believe on the ietf list in 2005. Suffice it to say  
that Stuart was not very
happy about it.)

Regards
Marshall


>
> Submitted 2006/8/10, expires 2007/2/10.
>
>
>> The main problem I see with DNS is more practical: setting up an
>> authoritative DNS server correctly seems to require a fair amount of
>> deep networking knowledge. ...
>> Whether that's a failure of current implementations or something more
>> fundamental, I don't know.
>
> Good question.  It might be hard merely because there's been no  
> reason to
> make it easy, until now.  There seems no fundamental reason why it  
> need be
> complex.
>
> -david
>
>
>
> _______________________________________________
> p2p-sip mailing list
> p2p-sip at cs.columbia.edu
> https://lists.cs.columbia.edu/cucslists/listinfo/p2p-sip


From dbarrett at quinthar.com  Wed Nov 29 14:23:33 2006
From: dbarrett at quinthar.com (David Barrett)
Date: Wed, 29 Nov 2006 11:23:33 -0800
Subject: [p2p-sip] Fwd: what's wrong with DNS?
In-Reply-To: <24CCCC428EFEA2469BF046DB3C7A8D22141384@namail5.corp.adobe.com>
Message-ID: <00ce01c713eb$dc718d20$0202a8c0@Quinthar>

That's totally fair; I agree what we're discussing is far off the normal
P2P-SIP track.  I've just been trying to nail down precisely what we're
doing and why, as it seems a little unfocused.  For example, each of the
following questions has been individually answered different ways by
different people, and I'm struggling to get a coherent picture in my head of
what the true answers are all together:

- Do we have a centralized service in the sky that does realtime
authentication/authorization before every call is set up?  Or do we have
absolutely no servers and anybody can install a client and choose a name and
it magically just works and is totally secure?  I've heard everything across
the board on this, even though they're wildly divergent in terms of
architecture.

- Do we support the "change password" use case (I lost my laptop and my
competitor is taking my calls)?  I've heard absolutely nothing on this
front, suggesting the answer is "no" despite it being absolutely crucial.
Furthermore, I've heard no explanation for how we can support that use case
without a centralized realtime authentication/authorization service.

- I've heard absolute certainty that we'll have a DHT *somewhere*, but no
clear explanation for what it'll hold.  Credential information?  Does it
look up SIP proxies?  *Is* it a SIP proxy somehow?  It seems like we're
scrambling to justify its use.

- I've heard absolute certainty that DNS/mDNS can't be trusted for some
critical set of users, though nobody has been able to give a real-world
situation where that's true, and logically it seems like it will nearly
always be available.

- I heard gasps when I suggested that VoIP isn't useful when people are
sitting in easy voice and eyeball range of each other and thus that use case
isn't valuable, but no acknowledgement or rebuttal.

Basically, I've heard a wide ranging desire to support every protocol, every
networking environment, every conceivable use case, across every conceivable
timeline and deployment size.  But I've heard no consensus on anything that
we *won't* do. 

Except, of course, use DNS.


Regardless, like Adam, I've got no real skin in the game and thus only have
a sideline interest in helping P2P-SIP succeed, so I'll step down and let
you guys continue your work.  Thanks for the vigorous debate!

-david

> -----Original Message-----
> From: Henry Sinnreich [mailto:hsinnrei at adobe.com]
> Sent: Wednesday, November 29, 2006 6:08 AM
> To: Brian Rosen; David Barrett; matthew at matthew.at; Adam Fisk; Eunsoo Shim
> Cc: p2p-sip at cs.columbia.edu
> Subject: RE: [p2p-sip] Fwd: what's wrong with DNS?
> 
> Full agreement.
> 
> This list was created for P2P SIP as hammered out in the existing charter.
> 
> Other good ideas and approaches must start with an I-D that people can
> discuss.
> 
> I am personally using dyndns for my home SIP service and am sure new DNS
> approaches deserve a better status as IETF WGs. Henning's proposed
> approaches for such work is right. There is no need to tie up this WG with
> another topic.
> 
> Thanks, Henry
> 
> -----Original Message-----
> From: p2p-sip-bounces at cs.columbia.edu [mailto:p2p-sip-
> bounces at cs.columbia.edu] On Behalf Of Brian Rosen
> Sent: Wednesday, November 29, 2006 7:02 AM
> To: 'David Barrett'; matthew at matthew.at; 'Adam Fisk'; 'Eunsoo Shim'
> Cc: p2p-sip at cs.columbia.edu
> Subject: Re: [p2p-sip] Fwd: what's wrong with DNS?
> 
> <not as chair, although at present, I'm not chair of anything>
> I suggest that you take Henning's advice, write up an Internet Draft, and
> submit it to sipping.
> It may be an interesting enough idea to enough folks that you can get an
> RFC
> out of it.
> 
> It does not meet the charter of p2psip as determined by list and BOF
> consensus.  It really is a different set of ideas.  Those ideas may have
> merit, they just don't fit here.
> 
> As a suggestion if you do decide to proceed:
> Consider defining the idea as a DDDS application
> Consider that dbarrett at dbarrett.quinthar.com is not really what you want,
> you want dbarrett at quinthar.com.  You want that sip address to resolve to a
> host who is named dbarret.quinthar.com for this purpose.  A DDDS
> application
> could start with that string, convert it to dbarrett.quinthar.com, and do
> a
> DNS query for a NAPTR that got you what you wanted.
> You might want something in the quinthar.com domain that indicates that it
> participates in the scheme you describe, since a normal sip client getting
> a
> request for dbarrett at quinthar.com would try to find the sip proxy at
> quinthar.com.  If quinthar.com doesn't operate a sip proxy, but instead
> has
> a resolution scheme like you imagine, you might want it to say so
> explicitly.
> 
> Please do consider the security implications of this proposal.  I think
> those are the real Achilles heel of the idea.  Proxy servers provide a
> useful capability that controls incoming calls.  Our current ideas work
> pretty much like a "Closed User Group", using a credential to authorize
> placement of a call within the CUG.  We may need more controls that that.
> 
> Brian
> 
> 
> ________________________________________
> From: p2p-sip-bounces at cs.columbia.edu
> [mailto:p2p-sip-bounces at cs.columbia.edu] On Behalf Of David Barrett
> Sent: Wednesday, November 29, 2006 3:49 AM
> To: matthew at matthew.at; 'Adam Fisk'; 'Eunsoo Shim'
> Cc: p2p-sip at cs.columbia.edu
> Subject: Re: [p2p-sip] Fwd: what's wrong with DNS?
> 
> I completely agree with Matthew here.? Basing this on DNS gives the domain
> owner extreme flexibility over how he wants to manage the users on that
> domain.? For example:
> 
> - A large provider (eg, Yahoo) might offer a standard public SIP proxy for
> millions of general users at *@yahoo.com
> - A large company (eg, Intel) might manage its own private SIP proxy for
> thousands of employees *@intel.com
> - A small company on a DSL connection might have a simple desktop computer
> running a standard SIP proxy a dozen employees at *@smallco.com, updating
> the IP of the proxy using a standard dynamic DNS client.
> - An individual, fixed user might have a static IP on his cablemodem and
> thus run both a SIP proxy and SIP client on the same computer for
> dbarrett at dbarrett.quinthar.com
> - An individual, mobile user might have a laptop running an integrated
> SIP-proxy/SIP-client/dynamic-DNS/mDNS client that updates the
> dbarrett.quinthar.com record to wherever the laptop is right now (perhaps
> using a DHT to find volunteer anonymous SIP proxies and STUN/TURN servers
> if
> behind a NAT), and that advertises itself on
> dbarrett at dbarrett.quinthar.com.local for totally infrastructureless, ad-
> hoc
> environments
> 
> This spans the entire range of the largest to the smallest, the most fixed
> to the most mobile environment.? With the exception of some kind of DHT to
> find a cloud of public resources (ie, SIP proxies that will let anybody
> register with them from any domain, as well as STUN and TURN servers),
> everything is done with existing technologies.
> 
> Furthermore, only the "individual, fixed" and "individual, mobile" use
> cases
> require anything that can't be deployed off the shelf right now, and even
> that is just a repackaging of existing technologies (with the exception of
> the public-resource DHT, if we choose to go that route).? Indeed, it's not
> hard to imagine a well-packaged SIP client that has an install wizard
> saying:
> 
> What is your SIP address?
> a) I already have an address assigned to me (eg, dbarrett at yahoo.com)
> b) I have no address, please help me create one
> 
> (Choose B) Every address consist of a username and a domain.? What domain
> do
> you want to use?
> a) I want to use an existing domain (choose from a list of public
> providers,
> eg @yahoo.com)
> b) I want to buy a new domain (choose from a list of domain providers)
> c) I want to get a free domain (choose from a list of free providers, such
> as DynDNS)
> d) I want to use a domain I already own (enter the domain)
> 
> (Choose D) Testing... dbarrett.quinthar.com is registered with 'DynDNS,
> Inc',
> which is a recognized dynamic DNS provider.? Enter your DynDNS name and
> password and we will automatically update dbarrett.quinthar.com to be
> associated with this computer, even if it changes networks.
> 
> (Enter name/pw) Testing... we've confirmed our ability to update
> dbarrett.quinthar.com.? Your new SIP address is
> dbarrett at dbarrett.quinthar.com.
> 
> Basically, make the client aware of dynamic DNS, use the standard dynamic
> DNS update protocols, and it can be made really easy for the user.
> Similarly, this SIP client could act as an mDNS server (on OS's that don't
> already support it) and thus autodetect other P2P-SIP clients on an ad-hoc
> network.
> 
> It can be totally self-configuring and seamlessly scalable from huge to
> tiny
> deployments, without requiring a benevolent dictator or inventing a
> replacement for the IANA.
> 
> -david
> 
> ________________________________________
> From: p2p-sip-bounces at cs.columbia.edu
> [mailto:p2p-sip-bounces at cs.columbia.edu] On Behalf Of Matthew Kaufman
> Sent: Tuesday, November 28, 2006 4:30 PM
> To: 'Adam Fisk'; 'Eunsoo Shim'
> Cc: p2p-sip at cs.columbia.edu
> Subject: Re: [p2p-sip] Fwd: what's wrong with DNS?
> 
> Adam Fisk:
> David's breakdown described in best, earlier in this thread.? Here it is
> again:
> 1)?????? Every user chooses a domain name.? Mine might be
> 'dbarrett.p2psip.org'
> 2)?????? When I start my client up, I choose any non-firewalled peer (how
> do
> I find it?) and REGISTER myself with it as something like
> 'dbarrett at dbarrett.p2psip.org'
> 3)?????? After registering with the peer/proxy, I contact some
> highly-redundant set of dynamic DNS servers and say "this is the latest IP
> address of my proxy, point to it whenever anyone does a DNS lookup on
> dbarrett.p2psip.org"
> 4)?????? Your off the shelf, totally unaltered SIP client asks DNS for the
> IP of 'dbarrett.p2psip.org', gets the latest IP of my proxy, and looks up
> the user 'dbarrett at dbarrett.p2psip.org', and then contacts me as normal.
> 5)?????? If the peer I chose as my proxy goes down, I detect it
> immediately
> and choose a new non-firewwalled peer (again, how do I find it?), REGISTER
> with that as in step (2), and update my dynamic DNS binding as in step
> (3).
> 6)?????? Repeat.
> 
> I would add to this that if you are yourself not firewalled, you should be
> able to REGISTER with "yourself" and your machine should be able to
> directly
> process inbound calls that arrive.
> 
> If you are firewalled, clearly a node is needed that can relay call setup
> messages to you, and pretty clearly those should just be normal SIP
> messages. There's a good argument for allowing one to register with more
> than one at a time and point the dynamic DNS appropriately.
> 
> Additionally, you should also answer mDNS queries for you.p2psip.org with
> your local self-proxy, so that machines on your LAN can call you directly
> whether or not the DNS is reachable, if their SIP clients are mDNS-aware.
> 
> As for your open issue of finding a non-firewalled peer to register with?
> Use the DNS to get an address of one inside of p2psip.org (if that's the
> domain within which you're operating) Note that this covers the case of an
> organization that wants the clients registered with a stable machine or
> machines that it operates instead of with client machines acting as
> registrars. The organization can also set policy around what addresses can
> be registered, if credentials are required, etc.
> 
> The real work that needs doing is coming up with a minimal set of
> recommended existing standards plus the minimal extensions necessary such
> that a working system can be built. Clearly STUN or better is needed for
> NAT
> traversal. Clearly SIP itself is the call signaling, though extensions
> might
> enable neat P2P capabilities. There's some work around using mDNS for the
> detached-from-real-Internet cases. There's some work around what the
> necessary security infrastructure, if any, might need to be. There's some
> work around the canonical way of finding non-firewalled peers and
> distributing load among them. There might need to be work around how the
> DNS
> entries get updated. And there might need to be some work around building
> a
> big distributed dynamic DNS system for the case where a domain needs many
> thousands of peer entries but can't afford or doesn't want to operate a
> centralized cluster of those servers. I would point out, however, that the
> last item is already an active research topic, and not really a P2P SIP
> issue at all.
> 
> Using DNS as the protocol for the node-finding mechanism (and possibly as
> the credential-lookup mechanism) means that inbound calls from existing
> SIP
> clients should "just work", and that each domain can set its own policies
> around whether the node-lookup (DNS) function is centralized or
> distributed
> and whether the registrar function itself is centralized or distributed.
> Then enabling zeroconf mDNS on top of that covers the case where you can't
> talk to anyone but other clients on your ad-hoc network. And I think
> that's
> 100% of the enumerated use cases.
> 
> With no time spent on new signaling messages for outdated DHT
> architectures
> as mandatory client components.
> 
> 
> Matthew Kaufman
> matthew at matthew.at
> 
> ps. To reflect on what Adam said about people who've built "real p2p
> systems", we built a secure P2P VOIP softphone at amicima, and we spent a
> lot of time thinking about per-domain policy for architecture (distributed
> vs. centralized name to address lookup and NAT traversal functionality)
> and
> security (centralized vs. distributed trust models, certificate signing,
> etc.), which our MFPNet protocol supported. We used our secure media flow
> protocol to transport the signaling messages, the audio, and the
> simultaneous file transfers rather than pure SIP for signaling and RTP for
> the audio (because we think that media prioritization is a big deal and
> that
> even RTP is going to wish it had TCP-compatible congestion control some
> day), but what we see here is essentially the same problem, and I too am
> looking to make sure that what gets specified as the "P2P extensions to
> SIP"
> is the simplest solution that will serve real user needs.
> 
> 
> 
> _______________________________________________
> p2p-sip mailing list
> p2p-sip at cs.columbia.edu
> https://lists.cs.columbia.edu/cucslists/listinfo/p2p-sip



From edpimentl at gmail.com  Wed Nov 29 14:40:17 2006
From: edpimentl at gmail.com (EdPimentl)
Date: Wed, 29 Nov 2006 14:40:17 -0500
Subject: [p2p-sip] Fwd: what's wrong with DNS?
In-Reply-To: <00ce01c713eb$dc718d20$0202a8c0@Quinthar>
References: <24CCCC428EFEA2469BF046DB3C7A8D22141384@namail5.corp.adobe.com>
	<00ce01c713eb$dc718d20$0202a8c0@Quinthar>
Message-ID: <9dc4a1670611291140k40e5a762k8298badf0304b56e@mail.gmail.com>

The condundrum is that P2P_SIP as the infrastructure/technology and protocol
is its embryonic phase. A number of the questions which are being ask is
what will tackle during the operationalization of the service, along with
the business process, billing supports system and operational support
systems.

Having said that the WG wants to make sure that there is inherent
flexibility in P2P_SIP for JohnQPublic to call/connect/collaborate with
SusieAnyBody with out the fear  they will be encarcerated in an
unwilling-partner ims four-wall garden.
So if they ever completely builded with their enumerous ever changing
interfaces, their customer will be safely outside.


On 11/29/06, David Barrett <dbarrett at quinthar.com> wrote:
>
> That's totally fair; I agree what we're discussing is far off the normal
> P2P-SIP track.  I've just been trying to nail down precisely what we're
> doing and why, as it seems a little unfocused.  For example, each of the
> following questions has been individually answered different ways by
> different people, and I'm struggling to get a coherent picture in my head
> of
> what the true answers are all together:
>
> - Do we have a centralized service in the sky that does realtime
> authentication/authorization before every call is set up?  Or do we have
> absolutely no servers and anybody can install a client and choose a name
> and
> it magically just works and is totally secure?  I've heard everything
> across
> the board on this, even though they're wildly divergent in terms of
> architecture.
>
> - Do we support the "change password" use case (I lost my laptop and my
> competitor is taking my calls)?  I've heard absolutely nothing on this
> front, suggesting the answer is "no" despite it being absolutely crucial.
> Furthermore, I've heard no explanation for how we can support that use
> case
> without a centralized realtime authentication/authorization service.
>
> - I've heard absolute certainty that we'll have a DHT *somewhere*, but no
> clear explanation for what it'll hold.  Credential information?  Does it
> look up SIP proxies?  *Is* it a SIP proxy somehow?  It seems like we're
> scrambling to justify its use.
>
> - I've heard absolute certainty that DNS/mDNS can't be trusted for some
> critical set of users, though nobody has been able to give a real-world
> situation where that's true, and logically it seems like it will nearly
> always be available.
>
> - I heard gasps when I suggested that VoIP isn't useful when people are
> sitting in easy voice and eyeball range of each other and thus that use
> case
> isn't valuable, but no acknowledgement or rebuttal.
>
> Basically, I've heard a wide ranging desire to support every protocol,
> every
> networking environment, every conceivable use case, across every
> conceivable
> timeline and deployment size.  But I've heard no consensus on anything
> that
> we *won't* do.
>
> Except, of course, use DNS.
>
>
> Regardless, like Adam, I've got no real skin in the game and thus only
> have
> a sideline interest in helping P2P-SIP succeed, so I'll step down and let
> you guys continue your work.  Thanks for the vigorous debate!
>
> -david
>
> > -----Original Message-----
> > From: Henry Sinnreich [mailto:hsinnrei at adobe.com]
> > Sent: Wednesday, November 29, 2006 6:08 AM
> > To: Brian Rosen; David Barrett; matthew at matthew.at; Adam Fisk; Eunsoo
> Shim
> > Cc: p2p-sip at cs.columbia.edu
> > Subject: RE: [p2p-sip] Fwd: what's wrong with DNS?
> >
> > Full agreement.
> >
> > This list was created for P2P SIP as hammered out in the existing
> charter.
> >
> > Other good ideas and approaches must start with an I-D that people can
> > discuss.
> >
> > I am personally using dyndns for my home SIP service and am sure new DNS
> > approaches deserve a better status as IETF WGs. Henning's proposed
> > approaches for such work is right. There is no need to tie up this WG
> with
> > another topic.
> >
> > Thanks, Henry
> >
> > -----Original Message-----
> > From: p2p-sip-bounces at cs.columbia.edu [mailto:p2p-sip-
> > bounces at cs.columbia.edu] On Behalf Of Brian Rosen
> > Sent: Wednesday, November 29, 2006 7:02 AM
> > To: 'David Barrett'; matthew at matthew.at; 'Adam Fisk'; 'Eunsoo Shim'
> > Cc: p2p-sip at cs.columbia.edu
> > Subject: Re: [p2p-sip] Fwd: what's wrong with DNS?
> >
> > <not as chair, although at present, I'm not chair of anything>
> > I suggest that you take Henning's advice, write up an Internet Draft,
> and
> > submit it to sipping.
> > It may be an interesting enough idea to enough folks that you can get an
> > RFC
> > out of it.
> >
> > It does not meet the charter of p2psip as determined by list and BOF
> > consensus.  It really is a different set of ideas.  Those ideas may have
> > merit, they just don't fit here.
> >
> > As a suggestion if you do decide to proceed:
> > Consider defining the idea as a DDDS application
> > Consider that dbarrett at dbarrett.quinthar.com is not really what you
> want,
> > you want dbarrett at quinthar.com.  You want that sip address to resolve to
> a
> > host who is named dbarret.quinthar.com for this purpose.  A DDDS
> > application
> > could start with that string, convert it to dbarrett.quinthar.com, and
> do
> > a
> > DNS query for a NAPTR that got you what you wanted.
> > You might want something in the quinthar.com domain that indicates that
> it
> > participates in the scheme you describe, since a normal sip client
> getting
> > a
> > request for dbarrett at quinthar.com would try to find the sip proxy at
> > quinthar.com.  If quinthar.com doesn't operate a sip proxy, but instead
> > has
> > a resolution scheme like you imagine, you might want it to say so
> > explicitly.
> >
> > Please do consider the security implications of this proposal.  I think
> > those are the real Achilles heel of the idea.  Proxy servers provide a
> > useful capability that controls incoming calls.  Our current ideas work
> > pretty much like a "Closed User Group", using a credential to authorize
> > placement of a call within the CUG.  We may need more controls that
> that.
> >
> > Brian
> >
> >
> > ________________________________________
> > From: p2p-sip-bounces at cs.columbia.edu
> > [mailto:p2p-sip-bounces at cs.columbia.edu] On Behalf Of David Barrett
> > Sent: Wednesday, November 29, 2006 3:49 AM
> > To: matthew at matthew.at; 'Adam Fisk'; 'Eunsoo Shim'
> > Cc: p2p-sip at cs.columbia.edu
> > Subject: Re: [p2p-sip] Fwd: what's wrong with DNS?
> >
> > I completely agree with Matthew here. Basing this on DNS gives the
> domain
> > owner extreme flexibility over how he wants to manage the users on that
> > domain. For example:
> >
> > - A large provider (eg, Yahoo) might offer a standard public SIP proxy
> for
> > millions of general users at *@yahoo.com
> > - A large company (eg, Intel) might manage its own private SIP proxy for
> > thousands of employees *@intel.com
> > - A small company on a DSL connection might have a simple desktop
> computer
> > running a standard SIP proxy a dozen employees at *@smallco.com,
> updating
> > the IP of the proxy using a standard dynamic DNS client.
> > - An individual, fixed user might have a static IP on his cablemodem and
> > thus run both a SIP proxy and SIP client on the same computer for
> > dbarrett at dbarrett.quinthar.com
> > - An individual, mobile user might have a laptop running an integrated
> > SIP-proxy/SIP-client/dynamic-DNS/mDNS client that updates the
> > dbarrett.quinthar.com record to wherever the laptop is right now
> (perhaps
> > using a DHT to find volunteer anonymous SIP proxies and STUN/TURN
> servers
> > if
> > behind a NAT), and that advertises itself on
> > dbarrett at dbarrett.quinthar.com.local for totally infrastructureless, ad-
> > hoc
> > environments
> >
> > This spans the entire range of the largest to the smallest, the most
> fixed
> > to the most mobile environment. With the exception of some kind of DHT
> to
> > find a cloud of public resources (ie, SIP proxies that will let anybody
> > register with them from any domain, as well as STUN and TURN servers),
> > everything is done with existing technologies.
> >
> > Furthermore, only the "individual, fixed" and "individual, mobile" use
> > cases
> > require anything that can't be deployed off the shelf right now, and
> even
> > that is just a repackaging of existing technologies (with the exception
> of
> > the public-resource DHT, if we choose to go that route). Indeed, it's
> not
> > hard to imagine a well-packaged SIP client that has an install wizard
> > saying:
> >
> > What is your SIP address?
> > a) I already have an address assigned to me (eg, dbarrett at yahoo.com)
> > b) I have no address, please help me create one
> >
> > (Choose B) Every address consist of a username and a domain. What domain
> > do
> > you want to use?
> > a) I want to use an existing domain (choose from a list of public
> > providers,
> > eg @yahoo.com)
> > b) I want to buy a new domain (choose from a list of domain providers)
> > c) I want to get a free domain (choose from a list of free providers,
> such
> > as DynDNS)
> > d) I want to use a domain I already own (enter the domain)
> >
> > (Choose D) Testing... dbarrett.quinthar.com is registered with 'DynDNS,
> > Inc',
> > which is a recognized dynamic DNS provider. Enter your DynDNS name and
> > password and we will automatically update dbarrett.quinthar.com to be
> > associated with this computer, even if it changes networks.
> >
> > (Enter name/pw) Testing... we've confirmed our ability to update
> > dbarrett.quinthar.com. Your new SIP address is
> > dbarrett at dbarrett.quinthar.com.
> >
> > Basically, make the client aware of dynamic DNS, use the standard
> dynamic
> > DNS update protocols, and it can be made really easy for the user.
> > Similarly, this SIP client could act as an mDNS server (on OS's that
> don't
> > already support it) and thus autodetect other P2P-SIP clients on an
> ad-hoc
> > network.
> >
> > It can be totally self-configuring and seamlessly scalable from huge to
> > tiny
> > deployments, without requiring a benevolent dictator or inventing a
> > replacement for the IANA.
> >
> > -david
> >
> > ________________________________________
> > From: p2p-sip-bounces at cs.columbia.edu
> > [mailto:p2p-sip-bounces at cs.columbia.edu] On Behalf Of Matthew Kaufman
> > Sent: Tuesday, November 28, 2006 4:30 PM
> > To: 'Adam Fisk'; 'Eunsoo Shim'
> > Cc: p2p-sip at cs.columbia.edu
> > Subject: Re: [p2p-sip] Fwd: what's wrong with DNS?
> >
> > Adam Fisk:
> > David's breakdown described in best, earlier in this thread. Here it is
> > again:
> > 1) Every user chooses a domain name. Mine might be
> > 'dbarrett.p2psip.org'
> > 2) When I start my client up, I choose any non-firewalled peer (how
> > do
> > I find it?) and REGISTER myself with it as something like
> > 'dbarrett at dbarrett.p2psip.org'
> > 3) After registering with the peer/proxy, I contact some
> > highly-redundant set of dynamic DNS servers and say "this is the latest
> IP
> > address of my proxy, point to it whenever anyone does a DNS lookup on
> > dbarrett.p2psip.org"
> > 4) Your off the shelf, totally unaltered SIP client asks DNS for the
> > IP of 'dbarrett.p2psip.org', gets the latest IP of my proxy, and looks
> up
> > the user 'dbarrett at dbarrett.p2psip.org', and then contacts me as normal.
> > 5) If the peer I chose as my proxy goes down, I detect it
> > immediately
> > and choose a new non-firewwalled peer (again, how do I find it?),
> REGISTER
> > with that as in step (2), and update my dynamic DNS binding as in step
> > (3).
> > 6) Repeat.
> >
> > I would add to this that if you are yourself not firewalled, you should
> be
> > able to REGISTER with "yourself" and your machine should be able to
> > directly
> > process inbound calls that arrive.
> >
> > If you are firewalled, clearly a node is needed that can relay call
> setup
> > messages to you, and pretty clearly those should just be normal SIP
> > messages. There's a good argument for allowing one to register with more
> > than one at a time and point the dynamic DNS appropriately.
> >
> > Additionally, you should also answer mDNS queries for you.p2psip.orgwith
> > your local self-proxy, so that machines on your LAN can call you
> directly
> > whether or not the DNS is reachable, if their SIP clients are
> mDNS-aware.
> >
> > As for your open issue of finding a non-firewalled peer to register
> with?
> > Use the DNS to get an address of one inside of p2psip.org (if that's the
> > domain within which you're operating) Note that this covers the case of
> an
> > organization that wants the clients registered with a stable machine or
> > machines that it operates instead of with client machines acting as
> > registrars. The organization can also set policy around what addresses
> can
> > be registered, if credentials are required, etc.
> >
> > The real work that needs doing is coming up with a minimal set of
> > recommended existing standards plus the minimal extensions necessary
> such
> > that a working system can be built. Clearly STUN or better is needed for
> > NAT
> > traversal. Clearly SIP itself is the call signaling, though extensions
> > might
> > enable neat P2P capabilities. There's some work around using mDNS for
> the
> > detached-from-real-Internet cases. There's some work around what the
> > necessary security infrastructure, if any, might need to be. There's
> some
> > work around the canonical way of finding non-firewalled peers and
> > distributing load among them. There might need to be work around how the
> > DNS
> > entries get updated. And there might need to be some work around
> building
> > a
> > big distributed dynamic DNS system for the case where a domain needs
> many
> > thousands of peer entries but can't afford or doesn't want to operate a
> > centralized cluster of those servers. I would point out, however, that
> the
> > last item is already an active research topic, and not really a P2P SIP
> > issue at all.
> >
> > Using DNS as the protocol for the node-finding mechanism (and possibly
> as
> > the credential-lookup mechanism) means that inbound calls from existing
> > SIP
> > clients should "just work", and that each domain can set its own
> policies
> > around whether the node-lookup (DNS) function is centralized or
> > distributed
> > and whether the registrar function itself is centralized or distributed.
> > Then enabling zeroconf mDNS on top of that covers the case where you
> can't
> > talk to anyone but other clients on your ad-hoc network. And I think
> > that's
> > 100% of the enumerated use cases.
> >
> > With no time spent on new signaling messages for outdated DHT
> > architectures
> > as mandatory client components.
> >
> >
> > Matthew Kaufman
> > matthew at matthew.at
> >
> > ps. To reflect on what Adam said about people who've built "real p2p
> > systems", we built a secure P2P VOIP softphone at amicima, and we spent
> a
> > lot of time thinking about per-domain policy for architecture
> (distributed
> > vs. centralized name to address lookup and NAT traversal functionality)
> > and
> > security (centralized vs. distributed trust models, certificate signing,
> > etc.), which our MFPNet protocol supported. We used our secure media
> flow
> > protocol to transport the signaling messages, the audio, and the
> > simultaneous file transfers rather than pure SIP for signaling and RTP
> for
> > the audio (because we think that media prioritization is a big deal and
> > that
> > even RTP is going to wish it had TCP-compatible congestion control some
> > day), but what we see here is essentially the same problem, and I too am
> > looking to make sure that what gets specified as the "P2P extensions to
> > SIP"
> > is the simplest solution that will serve real user needs.
> >
> >
> >
> > _______________________________________________
> > p2p-sip mailing list
> > p2p-sip at cs.columbia.edu
> > https://lists.cs.columbia.edu/cucslists/listinfo/p2p-sip
>
>
> _______________________________________________
> p2p-sip mailing list
> p2p-sip at cs.columbia.edu
> https://lists.cs.columbia.edu/cucslists/listinfo/p2p-sip
>



-- 
-Ed

Mail:       edpimentl[at]gmail[dot]com
VoIP:      edpimentl [Skype | GoogleTalk]
IM:          edpimentl [ AOL | Jabber | Yahoo | MSN ]

Mobile Content Management
http://mobilecentral.ws

Mobile Social Networks(Private Label)
http://GooGaYa.com

Mobile Payment
http://AgilePAY.ws
-------------- next part --------------
An HTML attachment was scrubbed...
URL: https://lists.cs.columbia.edu/pipermail/p2p-sip/attachments/20061129/4e652731/attachment-0001.html

From hgs at cs.columbia.edu  Wed Nov 29 14:47:12 2006
From: hgs at cs.columbia.edu (Henning Schulzrinne)
Date: Wed, 29 Nov 2006 14:47:12 -0500
Subject: [p2p-sip] Fwd: what's wrong with DNS?
In-Reply-To: <00ce01c713eb$dc718d20$0202a8c0@Quinthar>
References: <00ce01c713eb$dc718d20$0202a8c0@Quinthar>
Message-ID: <456DE3C0.9000809@cs.columbia.edu>


> - I heard gasps when I suggested that VoIP isn't useful when people are
> sitting in easy voice and eyeball range of each other and thus that use case
> isn't valuable, but no acknowledgement or rebuttal.

SIP is useful in that case, for text (see the use of Jabber in IETF 
meetings, even within a room). We were never restricted here to just 
classical 'voice', as no RAI groups are.

Henning

From br at brianrosen.net  Wed Nov 29 14:48:50 2006
From: br at brianrosen.net (Brian Rosen)
Date: Wed, 29 Nov 2006 14:48:50 -0500
Subject: [p2p-sip] Fwd: what's wrong with DNS?
In-Reply-To: <00ce01c713eb$dc718d20$0202a8c0@Quinthar>
Message-ID: <02de01c713ef$6628cd50$640fa8c0@cis.neustar.com>

I don't think we have much divergence of opinion.  We have a couple of
pieces that have had scant attention paid thus far:

> - Do we have a centralized service in the sky that does realtime
> authentication/authorization before every call is set up?  Or do we have
> absolutely no servers and anybody can install a client and choose a name
> and
> it magically just works and is totally secure?  I've heard everything
> across
> the board on this, even though they're wildly divergent in terms of
> architecture.
You get a credential from a centralized credentialing service.  Thereafter,
you can use that credential (as long as it's valid) on any client and join
the overlay network.

> 
> - Do we support the "change password" use case (I lost my laptop and my
> competitor is taking my calls)?  I've heard absolutely nothing on this
> front, suggesting the answer is "no" despite it being absolutely crucial.
> Furthermore, I've heard no explanation for how we can support that use
> case
> without a centralized realtime authentication/authorization service.
Here, we have not had much discussion, so we have no consensus.  I don't
THINK anyone is contemplating a centralized real time anything.

If you used a cert as a credential, then you might want to consult the CA's
CRL, or do something equivalent to make sure it was not revoked.

> 
> - I've heard absolute certainty that we'll have a DHT *somewhere*, but no
> clear explanation for what it'll hold.  Credential information?  Does it
> look up SIP proxies?  *Is* it a SIP proxy somehow?  It seems like we're
> scrambling to justify its use.
The DHT holds the data necessary to convert an address within the domain to
where you send an INVITE.  It may (I would say 'would') be the case that
before you can join the peer network and get access to the DHT, your
credential will be checked.  I don't think the DHT holds credentials.  A SIP
proxy may be a peer in the overlay network, so you could do name resolution
to that proxy within the overlay network.  It is NOT a SIP proxy.  We're not
scrambling to justify its use; we've chosen DHT as the mechanism for doing
address resolution.  Read the concepts document.

> 
> - I've heard absolute certainty that DNS/mDNS can't be trusted for some
> critical set of users, though nobody has been able to give a real-world
> situation where that's true, and logically it seems like it will nearly
> always be available.
Not an issue where there would be a consensus opinion within the p2psip
community.  I personally think this is not a big deal; you can get, and
really need DNS in a real network.

> 
> - I heard gasps when I suggested that VoIP isn't useful when people are
> sitting in easy voice and eyeball range of each other and thus that use
> case
> isn't valuable, but no acknowledgement or rebuttal.
Ever watch people IM each other while they were in the same conference room?
No, I don't think p2psip is about voice.  It's about multimedia session
establishment, just like real sip.  No more, no less.  All the media works
the same way, which means IM and presence (and video and real time text)
work.

> 
> Basically, I've heard a wide ranging desire to support every protocol,
> every
> networking environment, every conceivable use case, across every
> conceivable
> timeline and deployment size.  But I've heard no consensus on anything
> that
> we *won't* do.
It's everywhere sip might be used, but where instead of proxies, we use p2p.
We allow a proxy to be a peer, so that UAs connected to that proxy
conventionally can call other p2psip peers/clients and vice versa.

> 
> Except, of course, use DNS.
I expect we'll use DNS.  We haven't talked a lot, for example, about WHAT,
EXACTLY, you get from the DHT.  Is it a domain name or an IP address?  I'd
think it was a domain name, and you use DNS to resolve that to an IP
address.  No consensus on that, and it hasn't come up yet.


> 
> 
> Regardless, like Adam, I've got no real skin in the game and thus only
> have
> a sideline interest in helping P2P-SIP succeed, so I'll step down and let
> you guys continue your work.  Thanks for the vigorous debate!
Thanks.  It's been interesting.

Are we now back to our regularly scheduled program?  Adam? Matthew?


Brian


From ag at ag-projects.com  Wed Nov 29 14:58:24 2006
From: ag at ag-projects.com (Adrian Georgescu)
Date: Wed, 29 Nov 2006 20:58:24 +0100
Subject: [p2p-sip] Fwd: what's wrong with DNS?
In-Reply-To: <02de01c713ef$6628cd50$640fa8c0@cis.neustar.com>
References: <02de01c713ef$6628cd50$640fa8c0@cis.neustar.com>
Message-ID: <0BA3F014-8AC0-4071-B81B-100D28C98E3E@ag-projects.com>

>> Except, of course, use DNS.
> I expect we'll use DNS.  We haven't talked a lot, for example,  
> about WHAT,
> EXACTLY, you get from the DHT.  Is it a domain name or an IP  
> address?  I'd
> think it was a domain name, and you use DNS to resolve that to an IP
> address.  No consensus on that, and it hasn't come up yet.

The DNS can solve only the address resolution of the non-NAT-ed peer.  
One needs
more than dynamic DNS to merge a self-organizing network with SIP
functionality. No DNS server will be able to do parallel and serial
forking, aggregate presence information, NAT traversal, buddy list
synchronization among devices, redirect for not online to name a few
things one needs in the day-to-day communications. DNS is good for
bootstrap and and for providing global addressable URIs.

I think the WG should concentrate on a methodology of how to make a  
generic lookup
for a given resource when you have a list of boot strap nodes.  
Something like:

1. Use DNS or other bootstrap method to find a live peer in the P2P  
network
2. Perform a lookup function to the node to find a resource/service  
for a specific user in the P2P network

Example:

lookup sip_proxy for user at example.com
lookup voicemail for user at example.com
lookup storage for user at example.com
..

Adrian


From hsinnrei at adobe.com  Wed Nov 29 15:22:53 2006
From: hsinnrei at adobe.com (Henry Sinnreich)
Date: Wed, 29 Nov 2006 12:22:53 -0800
Subject: [p2p-sip] Fwd: what's wrong with DNS?
In-Reply-To: <02de01c713ef$6628cd50$640fa8c0@cis.neustar.com>
Message-ID: <24CCCC428EFEA2469BF046DB3C7A8D2214138D@namail5.corp.adobe.com>

>You get a credential from a centralized credentialing service.  
>Thereafter, you can use that credential (as long as it's valid) on any
>client and join the overlay network.

This is correct and some folks also call it an enrollment server.

Note the enrollment server is basically similar to any e-commerce
enrollment server, except the usage mode may be very different depending
on the particular e-commerce scenario and business model. 
Why should the enrollment for Internet communications be different from
any other e-commerce enrollment?

We had discussed including the enrollment server on the list and
discarded it as out of scope. This was the right decision, since
blessing business models is not the business of the IETF (I wish this
would be also applied to IMS, and other walled garden extensions to
SIP).

Henry

-----Original Message-----
From: Brian Rosen [mailto:br at brianrosen.net] 
Sent: Wednesday, November 29, 2006 1:49 PM
To: 'David Barrett'; Henry Sinnreich; matthew at matthew.at; 'Adam Fisk';
'Eunsoo Shim'
Cc: p2p-sip at cs.columbia.edu
Subject: RE: [p2p-sip] Fwd: what's wrong with DNS?

I don't think we have much divergence of opinion.  We have a couple of
pieces that have had scant attention paid thus far:

> - Do we have a centralized service in the sky that does realtime
> authentication/authorization before every call is set up?  Or do we
have
> absolutely no servers and anybody can install a client and choose a
name
> and
> it magically just works and is totally secure?  I've heard everything
> across
> the board on this, even though they're wildly divergent in terms of
> architecture.
You get a credential from a centralized credentialing service.
Thereafter,
you can use that credential (as long as it's valid) on any client and
join
the overlay network.

> 
> - Do we support the "change password" use case (I lost my laptop and
my
> competitor is taking my calls)?  I've heard absolutely nothing on this
> front, suggesting the answer is "no" despite it being absolutely
crucial.
> Furthermore, I've heard no explanation for how we can support that use
> case
> without a centralized realtime authentication/authorization service.
Here, we have not had much discussion, so we have no consensus.  I don't
THINK anyone is contemplating a centralized real time anything.

If you used a cert as a credential, then you might want to consult the
CA's
CRL, or do something equivalent to make sure it was not revoked.

> 
> - I've heard absolute certainty that we'll have a DHT *somewhere*, but
no
> clear explanation for what it'll hold.  Credential information?  Does
it
> look up SIP proxies?  *Is* it a SIP proxy somehow?  It seems like
we're
> scrambling to justify its use.
The DHT holds the data necessary to convert an address within the domain
to
where you send an INVITE.  It may (I would say 'would') be the case that
before you can join the peer network and get access to the DHT, your
credential will be checked.  I don't think the DHT holds credentials.  A
SIP
proxy may be a peer in the overlay network, so you could do name
resolution
to that proxy within the overlay network.  It is NOT a SIP proxy.  We're
not
scrambling to justify its use; we've chosen DHT as the mechanism for
doing
address resolution.  Read the concepts document.

> 
> - I've heard absolute certainty that DNS/mDNS can't be trusted for
some
> critical set of users, though nobody has been able to give a
real-world
> situation where that's true, and logically it seems like it will
nearly
> always be available.
Not an issue where there would be a consensus opinion within the p2psip
community.  I personally think this is not a big deal; you can get, and
really need DNS in a real network.

> 
> - I heard gasps when I suggested that VoIP isn't useful when people
are
> sitting in easy voice and eyeball range of each other and thus that
use
> case
> isn't valuable, but no acknowledgement or rebuttal.
Ever watch people IM each other while they were in the same conference
room?
No, I don't think p2psip is about voice.  It's about multimedia session
establishment, just like real sip.  No more, no less.  All the media
works
the same way, which means IM and presence (and video and real time text)
work.

> 
> Basically, I've heard a wide ranging desire to support every protocol,
> every
> networking environment, every conceivable use case, across every
> conceivable
> timeline and deployment size.  But I've heard no consensus on anything
> that
> we *won't* do.
It's everywhere sip might be used, but where instead of proxies, we use
p2p.
We allow a proxy to be a peer, so that UAs connected to that proxy
conventionally can call other p2psip peers/clients and vice versa.

> 
> Except, of course, use DNS.
I expect we'll use DNS.  We haven't talked a lot, for example, about
WHAT,
EXACTLY, you get from the DHT.  Is it a domain name or an IP address?
I'd
think it was a domain name, and you use DNS to resolve that to an IP
address.  No consensus on that, and it hasn't come up yet.


> 
> 
> Regardless, like Adam, I've got no real skin in the game and thus only
> have
> a sideline interest in helping P2P-SIP succeed, so I'll step down and
let
> you guys continue your work.  Thanks for the vigorous debate!
Thanks.  It's been interesting.

Are we now back to our regularly scheduled program?  Adam? Matthew?


Brian



From br at brianrosen.net  Wed Nov 29 15:57:56 2006
From: br at brianrosen.net (Brian Rosen)
Date: Wed, 29 Nov 2006 15:57:56 -0500
Subject: [p2p-sip] Fwd: what's wrong with DNS?
In-Reply-To: <24CCCC428EFEA2469BF046DB3C7A8D2214138D@namail5.corp.adobe.com>
Message-ID: <02ee01c713f9$0ee4d250$640fa8c0@cis.neustar.com>

Yeah.  I don't see a point in standardizing the enrollment server
interfaces.  I see some point in defining at least one
mandatory-to-implement form of credential tho.

Brian

> -----Original Message-----
> From: Henry Sinnreich [mailto:hsinnrei at adobe.com]
> Sent: Wednesday, November 29, 2006 3:23 PM
> To: Brian Rosen; David Barrett; matthew at matthew.at; Adam Fisk; Eunsoo Shim
> Cc: p2p-sip at cs.columbia.edu
> Subject: RE: [p2p-sip] Fwd: what's wrong with DNS?
> 
> >You get a credential from a centralized credentialing service.
> >Thereafter, you can use that credential (as long as it's valid) on any
> >client and join the overlay network.
> 
> This is correct and some folks also call it an enrollment server.
> 
> Note the enrollment server is basically similar to any e-commerce
> enrollment server, except the usage mode may be very different depending
> on the particular e-commerce scenario and business model.
> Why should the enrollment for Internet communications be different from
> any other e-commerce enrollment?
> 
> We had discussed including the enrollment server on the list and
> discarded it as out of scope. This was the right decision, since
> blessing business models is not the business of the IETF (I wish this
> would be also applied to IMS, and other walled garden extensions to
> SIP).
> 
> Henry
> 
> -----Original Message-----
> From: Brian Rosen [mailto:br at brianrosen.net]
> Sent: Wednesday, November 29, 2006 1:49 PM
> To: 'David Barrett'; Henry Sinnreich; matthew at matthew.at; 'Adam Fisk';
> 'Eunsoo Shim'
> Cc: p2p-sip at cs.columbia.edu
> Subject: RE: [p2p-sip] Fwd: what's wrong with DNS?
> 
> I don't think we have much divergence of opinion.  We have a couple of
> pieces that have had scant attention paid thus far:
> 
> > - Do we have a centralized service in the sky that does realtime
> > authentication/authorization before every call is set up?  Or do we
> have
> > absolutely no servers and anybody can install a client and choose a
> name
> > and
> > it magically just works and is totally secure?  I've heard everything
> > across
> > the board on this, even though they're wildly divergent in terms of
> > architecture.
> You get a credential from a centralized credentialing service.
> Thereafter,
> you can use that credential (as long as it's valid) on any client and
> join
> the overlay network.
> 
> >
> > - Do we support the "change password" use case (I lost my laptop and
> my
> > competitor is taking my calls)?  I've heard absolutely nothing on this
> > front, suggesting the answer is "no" despite it being absolutely
> crucial.
> > Furthermore, I've heard no explanation for how we can support that use
> > case
> > without a centralized realtime authentication/authorization service.
> Here, we have not had much discussion, so we have no consensus.  I don't
> THINK anyone is contemplating a centralized real time anything.
> 
> If you used a cert as a credential, then you might want to consult the
> CA's
> CRL, or do something equivalent to make sure it was not revoked.
> 
> >
> > - I've heard absolute certainty that we'll have a DHT *somewhere*, but
> no
> > clear explanation for what it'll hold.  Credential information?  Does
> it
> > look up SIP proxies?  *Is* it a SIP proxy somehow?  It seems like
> we're
> > scrambling to justify its use.
> The DHT holds the data necessary to convert an address within the domain
> to
> where you send an INVITE.  It may (I would say 'would') be the case that
> before you can join the peer network and get access to the DHT, your
> credential will be checked.  I don't think the DHT holds credentials.  A
> SIP
> proxy may be a peer in the overlay network, so you could do name
> resolution
> to that proxy within the overlay network.  It is NOT a SIP proxy.  We're
> not
> scrambling to justify its use; we've chosen DHT as the mechanism for
> doing
> address resolution.  Read the concepts document.
> 
> >
> > - I've heard absolute certainty that DNS/mDNS can't be trusted for
> some
> > critical set of users, though nobody has been able to give a
> real-world
> > situation where that's true, and logically it seems like it will
> nearly
> > always be available.
> Not an issue where there would be a consensus opinion within the p2psip
> community.  I personally think this is not a big deal; you can get, and
> really need DNS in a real network.
> 
> >
> > - I heard gasps when I suggested that VoIP isn't useful when people
> are
> > sitting in easy voice and eyeball range of each other and thus that
> use
> > case
> > isn't valuable, but no acknowledgement or rebuttal.
> Ever watch people IM each other while they were in the same conference
> room?
> No, I don't think p2psip is about voice.  It's about multimedia session
> establishment, just like real sip.  No more, no less.  All the media
> works
> the same way, which means IM and presence (and video and real time text)
> work.
> 
> >
> > Basically, I've heard a wide ranging desire to support every protocol,
> > every
> > networking environment, every conceivable use case, across every
> > conceivable
> > timeline and deployment size.  But I've heard no consensus on anything
> > that
> > we *won't* do.
> It's everywhere sip might be used, but where instead of proxies, we use
> p2p.
> We allow a proxy to be a peer, so that UAs connected to that proxy
> conventionally can call other p2psip peers/clients and vice versa.
> 
> >
> > Except, of course, use DNS.
> I expect we'll use DNS.  We haven't talked a lot, for example, about
> WHAT,
> EXACTLY, you get from the DHT.  Is it a domain name or an IP address?
> I'd
> think it was a domain name, and you use DNS to resolve that to an IP
> address.  No consensus on that, and it hasn't come up yet.
> 
> 
> >
> >
> > Regardless, like Adam, I've got no real skin in the game and thus only
> > have
> > a sideline interest in helping P2P-SIP succeed, so I'll step down and
> let
> > you guys continue your work.  Thanks for the vigorous debate!
> Thanks.  It's been interesting.
> 
> Are we now back to our regularly scheduled program?  Adam? Matthew?
> 
> 
> Brian


From spencer at mcsr-labs.org  Wed Nov 29 16:07:46 2006
From: spencer at mcsr-labs.org (Spencer Dawkins)
Date: Wed, 29 Nov 2006 15:07:46 -0600
Subject: [p2p-sip] Fwd: what's wrong with DNS?
References: <02ee01c713f9$0ee4d250$640fa8c0@cis.neustar.com>
Message-ID: <02af01c713fa$6b8b6e00$6b01a8c0@china.huawei.com>

> Yeah.  I don't see a point in standardizing the enrollment server
> interfaces.  I see some point in defining at least one
> mandatory-to-implement form of credential tho.
> 
> Brian

Strongly agree with Brian here.

Thanks,

Spencer


From slavitch at gmail.com  Wed Nov 29 16:13:56 2006
From: slavitch at gmail.com (Michael Slavitch)
Date: Wed, 29 Nov 2006 16:13:56 -0500
Subject: [p2p-sip] Fwd: what's wrong with DNS?
In-Reply-To: <02af01c713fa$6b8b6e00$6b01a8c0@china.huawei.com>
References: <02ee01c713f9$0ee4d250$640fa8c0@cis.neustar.com>
	<02af01c713fa$6b8b6e00$6b01a8c0@china.huawei.com>
Message-ID: <1fc32c8f0611291313p25380ecck9007abffb5ad44@mail.gmail.com>

Very much in agreement.

On 11/29/06, Spencer Dawkins <spencer at mcsr-labs.org> wrote:
>
> > Yeah.  I don't see a point in standardizing the enrollment server
> > interfaces.  I see some point in defining at least one
> > mandatory-to-implement form of credential tho.
> >
> > Brian
>
> Strongly agree with Brian here.
>
> Thanks,
>
> Spencer
>
> _______________________________________________
> p2p-sip mailing list
> p2p-sip at cs.columbia.edu
> https://lists.cs.columbia.edu/cucslists/listinfo/p2p-sip
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: https://lists.cs.columbia.edu/pipermail/p2p-sip/attachments/20061129/f2d92a80/attachment.html

From eunsooshim at gmail.com  Wed Nov 29 16:51:27 2006
From: eunsooshim at gmail.com (Eunsoo Shim)
Date: Wed, 29 Nov 2006 16:51:27 -0500
Subject: [p2p-sip] Fwd: what's wrong with DNS?
In-Reply-To: <00ce01c713eb$dc718d20$0202a8c0@Quinthar>
References: <24CCCC428EFEA2469BF046DB3C7A8D22141384@namail5.corp.adobe.com>
	<00ce01c713eb$dc718d20$0202a8c0@Quinthar>
Message-ID: <7b683f3f0611291351o308bb635rf7ab8ef6458ead19@mail.gmail.com>

http://www.p2psip.org/drafts/draft-singh-p2p-sip-00.txt

On 11/29/06, David Barrett <dbarrett at quinthar.com> wrote:
>
> That's totally fair; I agree what we're discussing is far off the normal
> P2P-SIP track.  I've just been trying to nail down precisely what we're
> doing and why, as it seems a little unfocused.  For example, each of the
> following questions has been individually answered different ways by
> different people, and I'm struggling to get a coherent picture in my head
> of
> what the true answers are all together:
>
> - Do we have a centralized service in the sky that does realtime
> authentication/authorization before every call is set up?  Or do we have
> absolutely no servers and anybody can install a client and choose a name
> and
> it magically just works and is totally secure?  I've heard everything
> across
> the board on this, even though they're wildly divergent in terms of
> architecture.
>
> - Do we support the "change password" use case (I lost my laptop and my
> competitor is taking my calls)?  I've heard absolutely nothing on this
> front, suggesting the answer is "no" despite it being absolutely crucial.
> Furthermore, I've heard no explanation for how we can support that use
> case
> without a centralized realtime authentication/authorization service.
>
> - I've heard absolute certainty that we'll have a DHT *somewhere*, but no
> clear explanation for what it'll hold.  Credential information?  Does it
> look up SIP proxies?  *Is* it a SIP proxy somehow?  It seems like we're
> scrambling to justify its use.


Kundan and Henning wrote a nice document that might give some answer to your
question.

http://www.p2psip.org/drafts/draft-singh-p2p-sip-00.txt

BTW, I have almost the same question about DNS for P2PSIP.


- I've heard absolute certainty that DNS/mDNS can't be trusted for some
> critical set of users, though nobody has been able to give a real-world
> situation where that's true, and logically it seems like it will nearly
> always be available.
>
> - I heard gasps when I suggested that VoIP isn't useful when people are
> sitting in easy voice and eyeball range of each other and thus that use
> case
> isn't valuable, but no acknowledgement or rebuttal.
>
> Basically, I've heard a wide ranging desire to support every protocol,
> every
> networking environment, every conceivable use case, across every
> conceivable
> timeline and deployment size.  But I've heard no consensus on anything
> that
> we *won't* do.
>
> Except, of course, use DNS.


I am still looking forward to hearing what is your  proposal exactly.
However, if you want to step down,  you can send me your thought off the
list.
For I don't want to revive the debate about DNS if it is fading out.

Regardless, like Adam, I've got no real skin in the game and thus only have
> a sideline interest in helping P2P-SIP succeed, so I'll step down and let
> you guys continue your work.  Thanks for the vigorous debate!
>
> -david


Thanks.

Eunsoo

> -----Original Message-----
> > From: Henry Sinnreich [mailto:hsinnrei at adobe.com]
> > Sent: Wednesday, November 29, 2006 6:08 AM
> > To: Brian Rosen; David Barrett; matthew at matthew.at; Adam Fisk; Eunsoo
> Shim
> > Cc: p2p-sip at cs.columbia.edu
> > Subject: RE: [p2p-sip] Fwd: what's wrong with DNS?
> >
> > Full agreement.
> >
> > This list was created for P2P SIP as hammered out in the existing
> charter.
> >
> > Other good ideas and approaches must start with an I-D that people can
> > discuss.
> >
> > I am personally using dyndns for my home SIP service and am sure new DNS
> > approaches deserve a better status as IETF WGs. Henning's proposed
> > approaches for such work is right. There is no need to tie up this WG
> with
> > another topic.
> >
> > Thanks, Henry
> >
> > -----Original Message-----
> > From: p2p-sip-bounces at cs.columbia.edu [mailto:p2p-sip-
> > bounces at cs.columbia.edu] On Behalf Of Brian Rosen
> > Sent: Wednesday, November 29, 2006 7:02 AM
> > To: 'David Barrett'; matthew at matthew.at; 'Adam Fisk'; 'Eunsoo Shim'
> > Cc: p2p-sip at cs.columbia.edu
> > Subject: Re: [p2p-sip] Fwd: what's wrong with DNS?
> >
> > <not as chair, although at present, I'm not chair of anything>
> > I suggest that you take Henning's advice, write up an Internet Draft,
> and
> > submit it to sipping.
> > It may be an interesting enough idea to enough folks that you can get an
> > RFC
> > out of it.
> >
> > It does not meet the charter of p2psip as determined by list and BOF
> > consensus.  It really is a different set of ideas.  Those ideas may have
> > merit, they just don't fit here.
> >
> > As a suggestion if you do decide to proceed:
> > Consider defining the idea as a DDDS application
> > Consider that dbarrett at dbarrett.quinthar.com is not really what you
> want,
> > you want dbarrett at quinthar.com.  You want that sip address to resolve to
> a
> > host who is named dbarret.quinthar.com for this purpose.  A DDDS
> > application
> > could start with that string, convert it to dbarrett.quinthar.com, and
> do
> > a
> > DNS query for a NAPTR that got you what you wanted.
> > You might want something in the quinthar.com domain that indicates that
> it
> > participates in the scheme you describe, since a normal sip client
> getting
> > a
> > request for dbarrett at quinthar.com would try to find the sip proxy at
> > quinthar.com.  If quinthar.com doesn't operate a sip proxy, but instead
> > has
> > a resolution scheme like you imagine, you might want it to say so
> > explicitly.
> >
> > Please do consider the security implications of this proposal.  I think
> > those are the real Achilles heel of the idea.  Proxy servers provide a
> > useful capability that controls incoming calls.  Our current ideas work
> > pretty much like a "Closed User Group", using a credential to authorize
> > placement of a call within the CUG.  We may need more controls that
> that.
> >
> > Brian
> >
> >
> > ________________________________________
> > From: p2p-sip-bounces at cs.columbia.edu
> > [mailto:p2p-sip-bounces at cs.columbia.edu] On Behalf Of David Barrett
> > Sent: Wednesday, November 29, 2006 3:49 AM
> > To: matthew at matthew.at; 'Adam Fisk'; 'Eunsoo Shim'
> > Cc: p2p-sip at cs.columbia.edu
> > Subject: Re: [p2p-sip] Fwd: what's wrong with DNS?
> >
> > I completely agree with Matthew here. Basing this on DNS gives the
> domain
> > owner extreme flexibility over how he wants to manage the users on that
> > domain. For example:
> >
> > - A large provider (eg, Yahoo) might offer a standard public SIP proxy
> for
> > millions of general users at *@yahoo.com
> > - A large company (eg, Intel) might manage its own private SIP proxy for
> > thousands of employees *@intel.com
> > - A small company on a DSL connection might have a simple desktop
> computer
> > running a standard SIP proxy a dozen employees at *@smallco.com,
> updating
> > the IP of the proxy using a standard dynamic DNS client.
> > - An individual, fixed user might have a static IP on his cablemodem and
> > thus run both a SIP proxy and SIP client on the same computer for
> > dbarrett at dbarrett.quinthar.com
> > - An individual, mobile user might have a laptop running an integrated
> > SIP-proxy/SIP-client/dynamic-DNS/mDNS client that updates the
> > dbarrett.quinthar.com record to wherever the laptop is right now
> (perhaps
> > using a DHT to find volunteer anonymous SIP proxies and STUN/TURN
> servers
> > if
> > behind a NAT), and that advertises itself on
> > dbarrett at dbarrett.quinthar.com.local for totally infrastructureless, ad-
> > hoc
> > environments
> >
> > This spans the entire range of the largest to the smallest, the most
> fixed
> > to the most mobile environment. With the exception of some kind of DHT
> to
> > find a cloud of public resources (ie, SIP proxies that will let anybody
> > register with them from any domain, as well as STUN and TURN servers),
> > everything is done with existing technologies.
> >
> > Furthermore, only the "individual, fixed" and "individual, mobile" use
> > cases
> > require anything that can't be deployed off the shelf right now, and
> even
> > that is just a repackaging of existing technologies (with the exception
> of
> > the public-resource DHT, if we choose to go that route). Indeed, it's
> not
> > hard to imagine a well-packaged SIP client that has an install wizard
> > saying:
> >
> > What is your SIP address?
> > a) I already have an address assigned to me (eg, dbarrett at yahoo.com)
> > b) I have no address, please help me create one
> >
> > (Choose B) Every address consist of a username and a domain. What domain
> > do
> > you want to use?
> > a) I want to use an existing domain (choose from a list of public
> > providers,
> > eg @yahoo.com)
> > b) I want to buy a new domain (choose from a list of domain providers)
> > c) I want to get a free domain (choose from a list of free providers,
> such
> > as DynDNS)
> > d) I want to use a domain I already own (enter the domain)
> >
> > (Choose D) Testing... dbarrett.quinthar.com is registered with 'DynDNS,
> > Inc',
> > which is a recognized dynamic DNS provider. Enter your DynDNS name and
> > password and we will automatically update dbarrett.quinthar.com to be
> > associated with this computer, even if it changes networks.
> >
> > (Enter name/pw) Testing... we've confirmed our ability to update
> > dbarrett.quinthar.com. Your new SIP address is
> > dbarrett at dbarrett.quinthar.com.
> >
> > Basically, make the client aware of dynamic DNS, use the standard
> dynamic
> > DNS update protocols, and it can be made really easy for the user.
> > Similarly, this SIP client could act as an mDNS server (on OS's that
> don't
> > already support it) and thus autodetect other P2P-SIP clients on an
> ad-hoc
> > network.
> >
> > It can be totally self-configuring and seamlessly scalable from huge to
> > tiny
> > deployments, without requiring a benevolent dictator or inventing a
> > replacement for the IANA.
> >
> > -david
> >
> > ________________________________________
> > From: p2p-sip-bounces at cs.columbia.edu
> > [mailto:p2p-sip-bounces at cs.columbia.edu] On Behalf Of Matthew Kaufman
> > Sent: Tuesday, November 28, 2006 4:30 PM
> > To: 'Adam Fisk'; 'Eunsoo Shim'
> > Cc: p2p-sip at cs.columbia.edu
> > Subject: Re: [p2p-sip] Fwd: what's wrong with DNS?
> >
> > Adam Fisk:
> > David's breakdown described in best, earlier in this thread. Here it is
> > again:
> > 1) Every user chooses a domain name. Mine might be
> > 'dbarrett.p2psip.org'
> > 2) When I start my client up, I choose any non-firewalled peer (how
> > do
> > I find it?) and REGISTER myself with it as something like
> > 'dbarrett at dbarrett.p2psip.org'
> > 3) After registering with the peer/proxy, I contact some
> > highly-redundant set of dynamic DNS servers and say "this is the latest
> IP
> > address of my proxy, point to it whenever anyone does a DNS lookup on
> > dbarrett.p2psip.org"
> > 4) Your off the shelf, totally unaltered SIP client asks DNS for the
> > IP of 'dbarrett.p2psip.org', gets the latest IP of my proxy, and looks
> up
> > the user 'dbarrett at dbarrett.p2psip.org', and then contacts me as normal.
> > 5) If the peer I chose as my proxy goes down, I detect it
> > immediately
> > and choose a new non-firewwalled peer (again, how do I find it?),
> REGISTER
> > with that as in step (2), and update my dynamic DNS binding as in step
> > (3).
> > 6) Repeat.
> >
> > I would add to this that if you are yourself not firewalled, you should
> be
> > able to REGISTER with "yourself" and your machine should be able to
> > directly
> > process inbound calls that arrive.
> >
> > If you are firewalled, clearly a node is needed that can relay call
> setup
> > messages to you, and pretty clearly those should just be normal SIP
> > messages. There's a good argument for allowing one to register with more
> > than one at a time and point the dynamic DNS appropriately.
> >
> > Additionally, you should also answer mDNS queries for you.p2psip.orgwith
> > your local self-proxy, so that machines on your LAN can call you
> directly
> > whether or not the DNS is reachable, if their SIP clients are
> mDNS-aware.
> >
> > As for your open issue of finding a non-firewalled peer to register
> with?
> > Use the DNS to get an address of one inside of p2psip.org (if that's the
> > domain within which you're operating) Note that this covers the case of
> an
> > organization that wants the clients registered with a stable machine or
> > machines that it operates instead of with client machines acting as
> > registrars. The organization can also set policy around what addresses
> can
> > be registered, if credentials are required, etc.
> >
> > The real work that needs doing is coming up with a minimal set of
> > recommended existing standards plus the minimal extensions necessary
> such
> > that a working system can be built. Clearly STUN or better is needed for
> > NAT
> > traversal. Clearly SIP itself is the call signaling, though extensions
> > might
> > enable neat P2P capabilities. There's some work around using mDNS for
> the
> > detached-from-real-Internet cases. There's some work around what the
> > necessary security infrastructure, if any, might need to be. There's
> some
> > work around the canonical way of finding non-firewalled peers and
> > distributing load among them. There might need to be work around how the
> > DNS
> > entries get updated. And there might need to be some work around
> building
> > a
> > big distributed dynamic DNS system for the case where a domain needs
> many
> > thousands of peer entries but can't afford or doesn't want to operate a
> > centralized cluster of those servers. I would point out, however, that
> the
> > last item is already an active research topic, and not really a P2P SIP
> > issue at all.
> >
> > Using DNS as the protocol for the node-finding mechanism (and possibly
> as
> > the credential-lookup mechanism) means that inbound calls from existing
> > SIP
> > clients should "just work", and that each domain can set its own
> policies
> > around whether the node-lookup (DNS) function is centralized or
> > distributed
> > and whether the registrar function itself is centralized or distributed.
> > Then enabling zeroconf mDNS on top of that covers the case where you
> can't
> > talk to anyone but other clients on your ad-hoc network. And I think
> > that's
> > 100% of the enumerated use cases.
> >
> > With no time spent on new signaling messages for outdated DHT
> > architectures
> > as mandatory client components.
> >
> >
> > Matthew Kaufman
> > matthew at matthew.at
> >
> > ps. To reflect on what Adam said about people who've built "real p2p
> > systems", we built a secure P2P VOIP softphone at amicima, and we spent
> a
> > lot of time thinking about per-domain policy for architecture
> (distributed
> > vs. centralized name to address lookup and NAT traversal functionality)
> > and
> > security (centralized vs. distributed trust models, certificate signing,
> > etc.), which our MFPNet protocol supported. We used our secure media
> flow
> > protocol to transport the signaling messages, the audio, and the
> > simultaneous file transfers rather than pure SIP for signaling and RTP
> for
> > the audio (because we think that media prioritization is a big deal and
> > that
> > even RTP is going to wish it had TCP-compatible congestion control some
> > day), but what we see here is essentially the same problem, and I too am
> > looking to make sure that what gets specified as the "P2P extensions to
> > SIP"
> > is the simplest solution that will serve real user needs.
> >
> >
> >
> > _______________________________________________
> > p2p-sip mailing list
> > p2p-sip at cs.columbia.edu
> > https://lists.cs.columbia.edu/cucslists/listinfo/p2p-sip
>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: https://lists.cs.columbia.edu/pipermail/p2p-sip/attachments/20061129/8603c5a7/attachment-0001.html

From hsinnrei at adobe.com  Thu Nov 30 11:13:55 2006
From: hsinnrei at adobe.com (Henry Sinnreich)
Date: Thu, 30 Nov 2006 08:13:55 -0800
Subject: [p2p-sip] Fwd: what's wrong with DNS?
In-Reply-To: <02ee01c713f9$0ee4d250$640fa8c0@cis.neustar.com>
Message-ID: <24CCCC428EFEA2469BF046DB3C7A8D2214138F@namail5.corp.adobe.com>

> I see some point in defining at least one
> mandatory-to-implement form of credential

This is business decision for the owner(s) of the P2P system.

It would help though to have some BCP documents that cover several usage
scenarios.

Henry

-----Original Message-----
From: Brian Rosen [mailto:br at brianrosen.net] 
Sent: Wednesday, November 29, 2006 2:58 PM
To: Henry Sinnreich; 'David Barrett'; matthew at matthew.at; 'Adam Fisk';
'Eunsoo Shim'
Cc: p2p-sip at cs.columbia.edu
Subject: RE: [p2p-sip] Fwd: what's wrong with DNS?

Yeah.  I don't see a point in standardizing the enrollment server
interfaces.  I see some point in defining at least one
mandatory-to-implement form of credential tho.

Brian

> -----Original Message-----
> From: Henry Sinnreich [mailto:hsinnrei at adobe.com]
> Sent: Wednesday, November 29, 2006 3:23 PM
> To: Brian Rosen; David Barrett; matthew at matthew.at; Adam Fisk; Eunsoo
Shim
> Cc: p2p-sip at cs.columbia.edu
> Subject: RE: [p2p-sip] Fwd: what's wrong with DNS?
> 
> >You get a credential from a centralized credentialing service.
> >Thereafter, you can use that credential (as long as it's valid) on
any
> >client and join the overlay network.
> 
> This is correct and some folks also call it an enrollment server.
> 
> Note the enrollment server is basically similar to any e-commerce
> enrollment server, except the usage mode may be very different
depending
> on the particular e-commerce scenario and business model.
> Why should the enrollment for Internet communications be different
from
> any other e-commerce enrollment?
> 
> We had discussed including the enrollment server on the list and
> discarded it as out of scope. This was the right decision, since
> blessing business models is not the business of the IETF (I wish this
> would be also applied to IMS, and other walled garden extensions to
> SIP).
> 
> Henry
> 
> -----Original Message-----
> From: Brian Rosen [mailto:br at brianrosen.net]
> Sent: Wednesday, November 29, 2006 1:49 PM
> To: 'David Barrett'; Henry Sinnreich; matthew at matthew.at; 'Adam Fisk';
> 'Eunsoo Shim'
> Cc: p2p-sip at cs.columbia.edu
> Subject: RE: [p2p-sip] Fwd: what's wrong with DNS?
> 
> I don't think we have much divergence of opinion.  We have a couple of
> pieces that have had scant attention paid thus far:
> 
> > - Do we have a centralized service in the sky that does realtime
> > authentication/authorization before every call is set up?  Or do we
> have
> > absolutely no servers and anybody can install a client and choose a
> name
> > and
> > it magically just works and is totally secure?  I've heard
everything
> > across
> > the board on this, even though they're wildly divergent in terms of
> > architecture.
> You get a credential from a centralized credentialing service.
> Thereafter,
> you can use that credential (as long as it's valid) on any client and
> join
> the overlay network.
> 
> >
> > - Do we support the "change password" use case (I lost my laptop and
> my
> > competitor is taking my calls)?  I've heard absolutely nothing on
this
> > front, suggesting the answer is "no" despite it being absolutely
> crucial.
> > Furthermore, I've heard no explanation for how we can support that
use
> > case
> > without a centralized realtime authentication/authorization service.
> Here, we have not had much discussion, so we have no consensus.  I
don't
> THINK anyone is contemplating a centralized real time anything.
> 
> If you used a cert as a credential, then you might want to consult the
> CA's
> CRL, or do something equivalent to make sure it was not revoked.
> 
> >
> > - I've heard absolute certainty that we'll have a DHT *somewhere*,
but
> no
> > clear explanation for what it'll hold.  Credential information?
Does
> it
> > look up SIP proxies?  *Is* it a SIP proxy somehow?  It seems like
> we're
> > scrambling to justify its use.
> The DHT holds the data necessary to convert an address within the
domain
> to
> where you send an INVITE.  It may (I would say 'would') be the case
that
> before you can join the peer network and get access to the DHT, your
> credential will be checked.  I don't think the DHT holds credentials.
A
> SIP
> proxy may be a peer in the overlay network, so you could do name
> resolution
> to that proxy within the overlay network.  It is NOT a SIP proxy.
We're
> not
> scrambling to justify its use; we've chosen DHT as the mechanism for
> doing
> address resolution.  Read the concepts document.
> 
> >
> > - I've heard absolute certainty that DNS/mDNS can't be trusted for
> some
> > critical set of users, though nobody has been able to give a
> real-world
> > situation where that's true, and logically it seems like it will
> nearly
> > always be available.
> Not an issue where there would be a consensus opinion within the
p2psip
> community.  I personally think this is not a big deal; you can get,
and
> really need DNS in a real network.
> 
> >
> > - I heard gasps when I suggested that VoIP isn't useful when people
> are
> > sitting in easy voice and eyeball range of each other and thus that
> use
> > case
> > isn't valuable, but no acknowledgement or rebuttal.
> Ever watch people IM each other while they were in the same conference
> room?
> No, I don't think p2psip is about voice.  It's about multimedia
session
> establishment, just like real sip.  No more, no less.  All the media
> works
> the same way, which means IM and presence (and video and real time
text)
> work.
> 
> >
> > Basically, I've heard a wide ranging desire to support every
protocol,
> > every
> > networking environment, every conceivable use case, across every
> > conceivable
> > timeline and deployment size.  But I've heard no consensus on
anything
> > that
> > we *won't* do.
> It's everywhere sip might be used, but where instead of proxies, we
use
> p2p.
> We allow a proxy to be a peer, so that UAs connected to that proxy
> conventionally can call other p2psip peers/clients and vice versa.
> 
> >
> > Except, of course, use DNS.
> I expect we'll use DNS.  We haven't talked a lot, for example, about
> WHAT,
> EXACTLY, you get from the DHT.  Is it a domain name or an IP address?
> I'd
> think it was a domain name, and you use DNS to resolve that to an IP
> address.  No consensus on that, and it hasn't come up yet.
> 
> 
> >
> >
> > Regardless, like Adam, I've got no real skin in the game and thus
only
> > have
> > a sideline interest in helping P2P-SIP succeed, so I'll step down
and
> let
> > you guys continue your work.  Thanks for the vigorous debate!
> Thanks.  It's been interesting.
> 
> Are we now back to our regularly scheduled program?  Adam? Matthew?
> 
> 
> Brian



