From eunsoo at research.panasonic.com  Tue Nov  1 02:42:21 2005
From: eunsoo at research.panasonic.com (Eunsoo Shim)
Date: Tue Nov  1 02:43:15 2005
Subject: [p2p-sip] Draft: Requirements for 
	SIP-basedPeer-to-PeerInternetTelephony
Message-ID: <50e667b5.67b550e6@pintlmail.MITL.Research.Panasonic.COM>

Hi, Suresh,

Thanks for the information about draft-ford-behave-app-01.txt.
I realize we should closely look into BEHAVE WG documents about NAT traversal.

Please let us think about how to incorporate your suggestions on section 7.1.

Regards,

Eunsoo



----- Original Message -----
From: Pyda Srisuresh <srisuresh@yahoo.com>
Date: Monday, October 31, 2005 5:08 pm
Subject: RE: [p2p-sip] Draft: Requirements for SIP-based Peer-to-PeerInternetTelephony

> First off, good requirements set. Thanks for putting them 
> together. Couple of
> comemnts on the draft.
> 
> 1. I would suggest including the <draft-ford-behave-app-01.txt> 
> draft as a
> reference in the context of firewall/NAT traversal. The 
> recommendation in the
> draft about how P2P apps could adapt to traverse NATs would be 
> directlyrelevant here. Also talked in the draft are security 
> concerns with P2P NAT
> traversal.
> 
> 2. Usage Scenarios - Please expand on section 7.1 with a diagram - 
> specificallydescrbing a P2P SIP setup as opposed to a traditional 
> SIP. What
> servers/services are prevalent in traditional SIP setup vs a P2P 
> SIP setup.
> 
> Additional comments in line below.
> 
> regards,
> suresh
> 
> --- Salman Abdul Baset <salman@cs.columbia.edu> wrote:
> 
> > Thanks for your comments. Please see my comments inline:
> > 
> > > > 4.5.  NAT and Firewall Traversal
> > > >    R1.  The peer-to-peer system SHOULD distribute the 
> functionality of
> > > >    NAT and firewall traversal servers to the end-points.
> > >
> > > I don't see why this must be a requirement. Your other 
> requirements talk
> > > about being flexible on where services are located (i.e. 
> centralized or
> > > distributed). Given this flexibility, which I agree with, why 
> require> > that
> > > a certain service be distributed? Why not leave the choice up 
> to the
> > > builder of
> > > the particular P2P overlay network? If that builder wants to 
> provide a
> > > centralized NAT and Firewall Traversal scheme, why not let him?
> > 
> 
> [suresh] I too disagree with the requirement, but for a different 
> reason. I
> beleve, the requirement is that P2P nodes should be able to 
> traverese the
> FW/NAT devices enroute. Why should the functionality of 
> NAT/firewall traversal
> servers be distributed to the end-points? And, why is this a 
> requirement?
> I do not see why services required for P2P shoudl be mandated as 
> part of P2P
> end nodes.
> 
> > I agree that the choice should be left to the bundler of the P2P 
> overlay> network for deciding which services should be centralized 
> and which should
> > be distributed. However, at some point the P2P bundler will need 
> to decide
> > whether the overlay network it is building has degenerated into 
> a typical
> > SIP-based network and still being marketed as P2P:)
> > 
> 
> [suresh] I think, it woudl be a good idea to draw a picture to 
> contrast between
> a traditional SIP network and a P2P SIP network. Endpoint Identity 
> and location
> of P2P nodes (and the associated service) is the key distinction, 
> in my
> opinion. 
> 
> > > On the other hand, I would agree that the design should 
> SUPPORT making
> > > the NAT and
> > > Firewall Traversal distributable, but I think you have already 
> covered> > that in 4.1.
> > 
> > We think that distributed NAT and firewall traversal is one of 
> the key
> > motivations to build a P2P overlay, so we decided to keep it as 
> a seperate
> > requirement. 
> 
> [suresh] I agree, NAT and firewall traveral is a key requirement 
> for P2P
> overlay.
> 
> >              There are two requirements in this section. I think one
> > option is to delete R1 and keep R2 which says:
> > 
> > "A peer with NAT and firewall traversal capabilities SHOULD be 
> selected> such that it does not introduce significant delay 
> between the
> > communicating peers."
> > 
> 
> [suresh] The above requirement is disctinct from R1, in my 
> opinion. I woudl not
> suggest eliminating R1. R2 above is similar to Req-3 in
> <draft-ford-behave-app-01.txt>, which talks about design 
> guidelines for P2P
> applications in the context of NAT traversal.  
> 
> > The other option is to include this section as a subsection of 4.1.
> > 
> 
> [suresh] Please see my comment earlier about rewording R1.
> 
> > >
> > > > 5.4.1.  Centralized Lookup
> > > >    Value-added services such as reliable voicemail storage 
> can be
> > > >    provided by a central authority in a P2P network.  The system
> > > SHOULD
> > > >    allow the peers to efficiently locate the centralized service
> > > >    providers or resources.
> > >
> > > This is just a minor comment. The first sentence seems to 
> imply that
> > > voicemail
> > > is preferably implemented as a centralized service. I disagree 
> with that
> > > implication, and suggest that the first sentence be reworded 
> or simply
> > > removed.
> > >
> > > I DO agree with the requirement in the second sentence.
> > 
> > The sentence says '...reliable voicemail storage can be 
> provided...'. I
> > guess the 'can' should be changed to 'might'.
> > 
> 
> [suresh] That sounds good.
> 
> > 
> > > > 6.1.  Support for different DHTs
> > > >    Any protocol specification SHOULD be able to incorporate 
> different> > >    DHT algorithms based on the P2P service provider 
> requirements.> >
> > > Isn't this the same as requirement 4.3??
> > 
> > We wanted to emphasize that a 'P2P Protocol Spec' must support 
> multiple> DHTs. Since this requirement fell under 'Architecture' 
> and 'Protocol Spec
> > Requirements', we decided to keep it in both. I think cross-
> referencing> the sections will be helpful.
> > 
> > regards,
> > --
> > Salman Baset
> > 
> 
> regards,
> suresh
> 
> > 
> > >
> > > - Philip Matthews
> > >   Nimcat Networks (now a part of Avaya)
> > >
> > > PS. Please note that my e-mail address has recently changed to
> > >        matthewsp@avaya.com
> > >     Mail sent to my old address (matthews@nimcatnetworks.com) will
> > >     be forwarded for a period, but will eventually be returned.
> > >
> > >
> > >
> > > -----Original Message-----
> > > From: p2p-sip-bounces@cs.columbia.edu
> > > [p2p-sip-bounces@cs.columbia.edu] On Behalf Of Salman Abdul Baset
> > > Sent: October 28, 2005 13:16
> > > To: sipping@ietf.org
> > > Cc: p2p-sip@cs.columbia.edu
> > > Subject: [p2p-sip] Draft: Requirements for SIP-based Peer-to-Peer
> > > InternetTelephony
> > >
> > > We have attempted to compile the requirements for SIP-based 
> peer-to-peer
> > > Internet telephony in a draft. It is available at:
> > > http://www1.cs.columbia.edu/~salman/drafts/draft-baset-sipping-
> p2preq-00
> > > .txt
> > >
> > > Title: Requirements for SIP-based Peer-to-Peer Internet Telephony
> > > Abstract:
> > > This document defines requirements for designing peer-to-peer 
> voice,> > text, and real-time multimedia communication system 
> protocols.> >
> > > Comments are requested.
> > >
> > > Regards,
> > > Salman, Henning, Eunsoo, Kishore
> > > _______________________________________________
> > > p2p-sip mailing list
> > > p2p-sip@cs.columbia.edu
> > > http://lists.cs.columbia.edu/mailman/listinfo/p2p-sip
> > >
> > >
> > _______________________________________________
> > p2p-sip mailing list
> > p2p-sip@cs.columbia.edu
> > http://lists.cs.columbia.edu/mailman/listinfo/p2p-sip
> > 
> 
> 
> 
> _______________________________________________
> p2p-sip mailing list
> p2p-sip@cs.columbia.edu
> http://lists.cs.columbia.edu/mailman/listinfo/p2p-sip
> 

From eunsoo at research.panasonic.com  Tue Nov  1 02:45:57 2005
From: eunsoo at research.panasonic.com (Eunsoo Shim)
Date: Tue Nov  1 02:46:18 2005
Subject: [p2p-sip] dhts
Message-ID: <65348195.81956534@pintlmail.MITL.Research.Panasonic.COM>

I think you pointed out an important issue - selection of DHT.
There are many DHT mechanisms and I am sure we will see more in the future.
I don't think this community intends to be an judge on DHT mechanisms but we need to identify a very small number of DHTs for specifications (and/or experiments) of P2P SIP while making the protocol flexible enough to incorporate other DHT mechanisms easily. (It is one of the requirements put in draft-baset-sipping-p2preq-00.txt.)

So a question we have is how to select DHT mechanisms we are going to use initially.
I think it is an important meric if a particular DHT is well understood by people who want to implement P2P SIP.
But there are other requirements for DHT as well such as robustness and lookup speed. Thus this community should come up with a procedure for selection of DHTs for P2P SIP.
I am trying to compile a proposal for such a procedure (thanks to Rohan's advice).
Any input on this will be appreciated.
Thanks.

Eunsoo



----- Original Message -----
From: Adam Fisk <afisk@speedymail.org>
Date: Monday, October 31, 2005 3:23 pm
Subject: [p2p-sip] dhts

> I just wanted to pipe in on the DHT question to caution against 
> the 
> Chord bandwagon.  Sure, Chord is nice to reference because most 
> people 
> have heard of it.  That's really just because Chord was the first 
> prominent DHT algorithm, though, and that's still it's primary 
> distinguishing characteristic.  Other algorithms such as Kademlia 
> are 
> significantly more robust and have less rigid routing schemes, 
> though. 
> 
> We should be clear that Chord should only be referenced *because 
> people 
> know it* and should not seriously be considered for real world 
> implementations.  I'd actually prefer it if people stopped 
> referencing 
> Chord in the various drafts simply because it contributed to the 
> misconception that it's a good option.
> 
> Thanks.
> 
> -Adam
> _______________________________________________
> p2p-sip mailing list
> p2p-sip@cs.columbia.edu
> http://lists.cs.columbia.edu/mailman/listinfo/p2p-sip
> 



From matthewsp at avaya.com  Tue Nov  1 07:42:39 2005
From: matthewsp at avaya.com (Matthews, Philip (Philip))
Date: Tue Nov  1 07:43:16 2005
Subject: [p2p-sip] dhts
Message-ID: <DFD0AADB452F8B4E99FB78719CAF0E6101963138@ON0010AVEXU1.global.avaya.com>

In my opinion, if we choose to standardize a DHT
(and I am not completely sure that is the right thing to do right now),
then we should pick a very simple algorithm that is easy to analyse
and implement. Don't worry that it doesn't seem to scale well or work
in all cases. Instead, provide a way to negotiate the use of another
algorithm.

I see this as analogous to the security mechanisms in various
IETF protocols. A number of protocols require the implementation of
simple password protection (or something equally simple) and allow
the negotiation of something more complex.

Part of my reasons for suggesting this is a suspicion (without any
real evidence so far) that we will NOT find a single algorithm
that will work well in all cases, and so different P2P overlay networks
will want to use different DHT algorithms. 

- Philip


> -----Original Message-----
> From: p2p-sip-bounces@cs.columbia.edu 
> [mailto:p2p-sip-bounces@cs.columbia.edu] On Behalf Of Eunsoo Shim
> Sent: November 1, 2005 02:46
> To: Adam Fisk
> Cc: p2p-sip@cs.columbia.edu
> Subject: Re: [p2p-sip] dhts
> 
> I think you pointed out an important issue - selection of DHT.
> There are many DHT mechanisms and I am sure we will see more 
> in the future.
> I don't think this community intends to be an judge on DHT 
> mechanisms but we need to identify a very small number of 
> DHTs for specifications (and/or experiments) of P2P SIP while 
> making the protocol flexible enough to incorporate other DHT 
> mechanisms easily. (It is one of the requirements put in 
> draft-baset-sipping-p2preq-00.txt.)
> 
> So a question we have is how to select DHT mechanisms we are 
> going to use initially.
> I think it is an important meric if a particular DHT is well 
> understood by people who want to implement P2P SIP.
> But there are other requirements for DHT as well such as 
> robustness and lookup speed. Thus this community should come 
> up with a procedure for selection of DHTs for P2P SIP.
> I am trying to compile a proposal for such a procedure 
> (thanks to Rohan's advice).
> Any input on this will be appreciated.
> Thanks.
> 
> Eunsoo
> 
> 
> 
> ----- Original Message -----
> From: Adam Fisk <afisk@speedymail.org>
> Date: Monday, October 31, 2005 3:23 pm
> Subject: [p2p-sip] dhts
> 
> > I just wanted to pipe in on the DHT question to caution against 
> > the 
> > Chord bandwagon.  Sure, Chord is nice to reference because most 
> > people 
> > have heard of it.  That's really just because Chord was the first 
> > prominent DHT algorithm, though, and that's still it's primary 
> > distinguishing characteristic.  Other algorithms such as Kademlia 
> > are 
> > significantly more robust and have less rigid routing schemes, 
> > though. 
> > 
> > We should be clear that Chord should only be referenced *because 
> > people 
> > know it* and should not seriously be considered for real world 
> > implementations.  I'd actually prefer it if people stopped 
> > referencing 
> > Chord in the various drafts simply because it contributed to the 
> > misconception that it's a good option.
> > 
> > Thanks.
> > 
> > -Adam
> > _______________________________________________
> > p2p-sip mailing list
> > p2p-sip@cs.columbia.edu
> > http://lists.cs.columbia.edu/mailman/listinfo/p2p-sip
> > 
> 
> 
> 
> _______________________________________________
> p2p-sip mailing list
> p2p-sip@cs.columbia.edu
> http://lists.cs.columbia.edu/mailman/listinfo/p2p-sip
> 

From bryan at ethernot.org  Tue Nov  1 10:42:06 2005
From: bryan at ethernot.org (David A. Bryan)
Date: Tue Nov  1 10:42:45 2005
Subject: [p2p-sip] dhts
In-Reply-To: <DFD0AADB452F8B4E99FB78719CAF0E6101963138@ON0010AVEXU1.global.avaya.com>
References: <DFD0AADB452F8B4E99FB78719CAF0E6101963138@ON0010AVEXU1.global.avaya.com>
Message-ID: <8b2769930511010742mb4e74een8f586e84664b3acd@mail.gmail.com>

I'm fully in agreement here. I think that having multiple algorithms
-- and in particular very small simple algorithms as well as more
complex algorithms is essential. We are unlikely to be able to choose
one that will work in all cases, and are certain to not be able to
choose one that will be the "best" for any length of time -- there are
always advances in this area. I think Phillip's analogy to security is
a good one here -- devices should be able to negotiate a P2P protocol,
just as they can with security, encryption, or codec choice. We should
possibly even be able to negotiate non-DHT P2P mechanisms -- perhaps
in some very small scale cases we should use flood for simplicity.
Perhaps also in emergency cases?

--David

On 11/1/05, Matthews, Philip (Philip) <matthewsp@avaya.com> wrote:
> In my opinion, if we choose to standardize a DHT
> (and I am not completely sure that is the right thing to do right now),
> then we should pick a very simple algorithm that is easy to analyse
> and implement. Don't worry that it doesn't seem to scale well or work
> in all cases. Instead, provide a way to negotiate the use of another
> algorithm.
>
> I see this as analogous to the security mechanisms in various
> IETF protocols. A number of protocols require the implementation of
> simple password protection (or something equally simple) and allow
> the negotiation of something more complex.
>
> Part of my reasons for suggesting this is a suspicion (without any
> real evidence so far) that we will NOT find a single algorithm
> that will work well in all cases, and so different P2P overlay networks
> will want to use different DHT algorithms.
>
> - Philip
>
>
> > -----Original Message-----
> > From: p2p-sip-bounces@cs.columbia.edu
> > [mailto:p2p-sip-bounces@cs.columbia.edu] On Behalf Of Eunsoo Shim
> > Sent: November 1, 2005 02:46
> > To: Adam Fisk
> > Cc: p2p-sip@cs.columbia.edu
> > Subject: Re: [p2p-sip] dhts
> >
> > I think you pointed out an important issue - selection of DHT.
> > There are many DHT mechanisms and I am sure we will see more
> > in the future.
> > I don't think this community intends to be an judge on DHT
> > mechanisms but we need to identify a very small number of
> > DHTs for specifications (and/or experiments) of P2P SIP while
> > making the protocol flexible enough to incorporate other DHT
> > mechanisms easily. (It is one of the requirements put in
> > draft-baset-sipping-p2preq-00.txt.)
> >
> > So a question we have is how to select DHT mechanisms we are
> > going to use initially.
> > I think it is an important meric if a particular DHT is well
> > understood by people who want to implement P2P SIP.
> > But there are other requirements for DHT as well such as
> > robustness and lookup speed. Thus this community should come
> > up with a procedure for selection of DHTs for P2P SIP.
> > I am trying to compile a proposal for such a procedure
> > (thanks to Rohan's advice).
> > Any input on this will be appreciated.
> > Thanks.
> >
> > Eunsoo
> >
> >
> >
> > ----- Original Message -----
> > From: Adam Fisk <afisk@speedymail.org>
> > Date: Monday, October 31, 2005 3:23 pm
> > Subject: [p2p-sip] dhts
> >
> > > I just wanted to pipe in on the DHT question to caution against
> > > the
> > > Chord bandwagon.  Sure, Chord is nice to reference because most
> > > people
> > > have heard of it.  That's really just because Chord was the first
> > > prominent DHT algorithm, though, and that's still it's primary
> > > distinguishing characteristic.  Other algorithms such as Kademlia
> > > are
> > > significantly more robust and have less rigid routing schemes,
> > > though.
> > >
> > > We should be clear that Chord should only be referenced *because
> > > people
> > > know it* and should not seriously be considered for real world
> > > implementations.  I'd actually prefer it if people stopped
> > > referencing
> > > Chord in the various drafts simply because it contributed to the
> > > misconception that it's a good option.
> > >
> > > Thanks.
> > >
> > > -Adam
> > > _______________________________________________
> > > p2p-sip mailing list
> > > p2p-sip@cs.columbia.edu
> > > http://lists.cs.columbia.edu/mailman/listinfo/p2p-sip
> > >
> >
> >
> >
> > _______________________________________________
> > p2p-sip mailing list
> > p2p-sip@cs.columbia.edu
> > http://lists.cs.columbia.edu/mailman/listinfo/p2p-sip
> >
>
> _______________________________________________
> p2p-sip mailing list
> p2p-sip@cs.columbia.edu
> http://lists.cs.columbia.edu/mailman/listinfo/p2p-sip
>

From eunsoo at research.panasonic.com  Tue Nov  1 11:14:45 2005
From: eunsoo at research.panasonic.com (Eunsoo Shim)
Date: Tue Nov  1 11:18:21 2005
Subject: [p2p-sip] dhts
In-Reply-To: <8b2769930511010742mb4e74een8f586e84664b3acd@mail.gmail.com>
References: <DFD0AADB452F8B4E99FB78719CAF0E6101963138@ON0010AVEXU1.global.av
	aya.com> <8b2769930511010742mb4e74een8f586e84664b3acd@mail.gmail.com>
Message-ID: <43679475.4010607@research.panasonic.com>

Standardizing a DHT sounds somewhat overwhelming to me.
But if we put it as standardizing (or writing an experimental epec of )
a DHT for SIP-based Internet Telephony, Presence and Instant Messaging
(i.e., for specific target applications), the task sounds more
approachable and I feel more comfortable. And I think we have no choice
but to specify one or more DHTs for P2P SIP. Otherwise, we won't be able
to provide interoperability.

I also agree that we should start with simple algorithms. It is a good
idea to support negotiation about P2P algorithms.

Regards,

Eunsoo

David A. Bryan wrote:

>I'm fully in agreement here. I think that having multiple algorithms
>-- and in particular very small simple algorithms as well as more
>complex algorithms is essential. We are unlikely to be able to choose
>one that will work in all cases, and are certain to not be able to
>choose one that will be the "best" for any length of time -- there are
>always advances in this area. I think Phillip's analogy to security is
>a good one here -- devices should be able to negotiate a P2P protocol,
>just as they can with security, encryption, or codec choice. We should
>possibly even be able to negotiate non-DHT P2P mechanisms -- perhaps
>in some very small scale cases we should use flood for simplicity.
>Perhaps also in emergency cases?
>
>--David
>
>On 11/1/05, Matthews, Philip (Philip) <matthewsp@avaya.com> wrote:
>  
>
>>In my opinion, if we choose to standardize a DHT
>>(and I am not completely sure that is the right thing to do right now),
>>then we should pick a very simple algorithm that is easy to analyse
>>and implement. Don't worry that it doesn't seem to scale well or work
>>in all cases. Instead, provide a way to negotiate the use of another
>>algorithm.
>>
>>I see this as analogous to the security mechanisms in various
>>IETF protocols. A number of protocols require the implementation of
>>simple password protection (or something equally simple) and allow
>>the negotiation of something more complex.
>>
>>Part of my reasons for suggesting this is a suspicion (without any
>>real evidence so far) that we will NOT find a single algorithm
>>that will work well in all cases, and so different P2P overlay networks
>>will want to use different DHT algorithms.
>>
>>- Philip
>>
>>
>>    
>>
>>>-----Original Message-----
>>>From: p2p-sip-bounces@cs.columbia.edu
>>>[mailto:p2p-sip-bounces@cs.columbia.edu] On Behalf Of Eunsoo Shim
>>>Sent: November 1, 2005 02:46
>>>To: Adam Fisk
>>>Cc: p2p-sip@cs.columbia.edu
>>>Subject: Re: [p2p-sip] dhts
>>>
>>>I think you pointed out an important issue - selection of DHT.
>>>There are many DHT mechanisms and I am sure we will see more
>>>in the future.
>>>I don't think this community intends to be an judge on DHT
>>>mechanisms but we need to identify a very small number of
>>>DHTs for specifications (and/or experiments) of P2P SIP while
>>>making the protocol flexible enough to incorporate other DHT
>>>mechanisms easily. (It is one of the requirements put in
>>>draft-baset-sipping-p2preq-00.txt.)
>>>
>>>So a question we have is how to select DHT mechanisms we are
>>>going to use initially.
>>>I think it is an important meric if a particular DHT is well
>>>understood by people who want to implement P2P SIP.
>>>But there are other requirements for DHT as well such as
>>>robustness and lookup speed. Thus this community should come
>>>up with a procedure for selection of DHTs for P2P SIP.
>>>I am trying to compile a proposal for such a procedure
>>>(thanks to Rohan's advice).
>>>Any input on this will be appreciated.
>>>Thanks.
>>>
>>>Eunsoo
>>>
>>>
>>>
>>>----- Original Message -----
>>>From: Adam Fisk <afisk@speedymail.org>
>>>Date: Monday, October 31, 2005 3:23 pm
>>>Subject: [p2p-sip] dhts
>>>
>>>      
>>>
>>>>I just wanted to pipe in on the DHT question to caution against
>>>>the
>>>>Chord bandwagon.  Sure, Chord is nice to reference because most
>>>>people
>>>>have heard of it.  That's really just because Chord was the first
>>>>prominent DHT algorithm, though, and that's still it's primary
>>>>distinguishing characteristic.  Other algorithms such as Kademlia
>>>>are
>>>>significantly more robust and have less rigid routing schemes,
>>>>though.
>>>>
>>>>We should be clear that Chord should only be referenced *because
>>>>people
>>>>know it* and should not seriously be considered for real world
>>>>implementations.  I'd actually prefer it if people stopped
>>>>referencing
>>>>Chord in the various drafts simply because it contributed to the
>>>>misconception that it's a good option.
>>>>
>>>>Thanks.
>>>>
>>>>-Adam
>>>>_______________________________________________
>>>>p2p-sip mailing list
>>>>p2p-sip@cs.columbia.edu
>>>>http://lists.cs.columbia.edu/mailman/listinfo/p2p-sip
>>>>
>>>>        
>>>>
>>>
>>>_______________________________________________
>>>p2p-sip mailing list
>>>p2p-sip@cs.columbia.edu
>>>http://lists.cs.columbia.edu/mailman/listinfo/p2p-sip
>>>
>>>      
>>>
>>_______________________________________________
>>p2p-sip mailing list
>>p2p-sip@cs.columbia.edu
>>http://lists.cs.columbia.edu/mailman/listinfo/p2p-sip
>>
>>    
>>

From RoyR at saic-abingdon.com  Tue Nov  1 11:49:24 2005
From: RoyR at saic-abingdon.com (Roy, Radhika (AEAD))
Date: Tue Nov  1 11:46:38 2005
Subject: [p2p-sip] dhts
Message-ID: <2FE23D25B7292B489A217D1965CDA866089915@bh-exchange02.saic-abingdon.com>

So, the logical conclusion seems to be as follows:

1. P2P protocol will be algorithm agnostic

2. P2P protocol will allow capabilities to negotiate any algorithm to be
used among the respective peers

3. P2P scheme will standardize couple of algorithms that can be
negotiated among the peers starting with a simple one first

It appears that P2P protocol scheme should be flexible enough to use
algorithms and others per-(each)-peer basis.

To extend this further, it may also be said for transport (reliable,
unreliable) and security (transport layer, IP layer) that the P2P
protocol should also use the same principle.

Is that the right summary of this thread so far?

Best regards,
Radhika

-----Original Message-----
From: p2p-sip-bounces@cs.columbia.edu
[mailto:p2p-sip-bounces@cs.columbia.edu] On Behalf Of Eunsoo Shim
Sent: Tuesday, November 01, 2005 11:15 AM
To: David A. Bryan
Cc: p2p-sip@cs.columbia.edu
Subject: Re: [p2p-sip] dhts

Standardizing a DHT sounds somewhat overwhelming to me.
But if we put it as standardizing (or writing an experimental epec of )
a DHT for SIP-based Internet Telephony, Presence and Instant Messaging
(i.e., for specific target applications), the task sounds more
approachable and I feel more comfortable. And I think we have no choice
but to specify one or more DHTs for P2P SIP. Otherwise, we won't be able
to provide interoperability.

I also agree that we should start with simple algorithms. It is a good
idea to support negotiation about P2P algorithms.

Regards,

Eunsoo

David A. Bryan wrote:

>I'm fully in agreement here. I think that having multiple algorithms
>-- and in particular very small simple algorithms as well as more
>complex algorithms is essential. We are unlikely to be able to choose
>one that will work in all cases, and are certain to not be able to
>choose one that will be the "best" for any length of time -- there are
>always advances in this area. I think Phillip's analogy to security is
>a good one here -- devices should be able to negotiate a P2P protocol,
>just as they can with security, encryption, or codec choice. We should
>possibly even be able to negotiate non-DHT P2P mechanisms -- perhaps
>in some very small scale cases we should use flood for simplicity.
>Perhaps also in emergency cases?
>
>--David
>

From bryan at ethernot.org  Tue Nov  1 11:59:17 2005
From: bryan at ethernot.org (David A. Bryan)
Date: Tue Nov  1 11:59:36 2005
Subject: [p2p-sip] dhts
In-Reply-To: <2FE23D25B7292B489A217D1965CDA866089915@bh-exchange02.saic-abingdon.com>
References: <2FE23D25B7292B489A217D1965CDA866089915@bh-exchange02.saic-abingdon.com>
Message-ID: <8b2769930511010859s3cee8f29ga300ee6c24c4a368@mail.gmail.com>

I think this is the spirit of the thread, but a warning that in
practice getting this together may be quite difficult, since different
types of p2p implementations exchange different information between
nodes for admission and upkeep of the overlay.

I think if we take this approach, the negotiation of a p2p mechanism
and the implementation thereof (how p2p information is conveyed
between nodes) will have to be two independent things, and I doubt we
can standardize how the information is conveyed between nodes in a way
that doesn't take into account what algotithm is being used. We may
need a negotiation mechanism and then different "standards" for how
the nodes maintain the overlay, since the information contained in
messages and even the pattern and reason for messages exchanged
between nodes can vary greatly for different p2p mechanisms.

David


On 11/1/05, Roy, Radhika (AEAD) <RoyR@saic-abingdon.com> wrote:
> So, the logical conclusion seems to be as follows:
>
> 1. P2P protocol will be algorithm agnostic
>
> 2. P2P protocol will allow capabilities to negotiate any algorithm to be
> used among the respective peers
>
> 3. P2P scheme will standardize couple of algorithms that can be
> negotiated among the peers starting with a simple one first
>
> It appears that P2P protocol scheme should be flexible enough to use
> algorithms and others per-(each)-peer basis.
>
> To extend this further, it may also be said for transport (reliable,
> unreliable) and security (transport layer, IP layer) that the P2P
> protocol should also use the same principle.
>
> Is that the right summary of this thread so far?
>
> Best regards,
> Radhika
>
> -----Original Message-----
> From: p2p-sip-bounces@cs.columbia.edu
> [mailto:p2p-sip-bounces@cs.columbia.edu] On Behalf Of Eunsoo Shim
> Sent: Tuesday, November 01, 2005 11:15 AM
> To: David A. Bryan
> Cc: p2p-sip@cs.columbia.edu
> Subject: Re: [p2p-sip] dhts
>
> Standardizing a DHT sounds somewhat overwhelming to me.
> But if we put it as standardizing (or writing an experimental epec of )
> a DHT for SIP-based Internet Telephony, Presence and Instant Messaging
> (i.e., for specific target applications), the task sounds more
> approachable and I feel more comfortable. And I think we have no choice
> but to specify one or more DHTs for P2P SIP. Otherwise, we won't be able
> to provide interoperability.
>
> I also agree that we should start with simple algorithms. It is a good
> idea to support negotiation about P2P algorithms.
>
> Regards,
>
> Eunsoo
>
> David A. Bryan wrote:
>
> >I'm fully in agreement here. I think that having multiple algorithms
> >-- and in particular very small simple algorithms as well as more
> >complex algorithms is essential. We are unlikely to be able to choose
> >one that will work in all cases, and are certain to not be able to
> >choose one that will be the "best" for any length of time -- there are
> >always advances in this area. I think Phillip's analogy to security is
> >a good one here -- devices should be able to negotiate a P2P protocol,
> >just as they can with security, encryption, or codec choice. We should
> >possibly even be able to negotiate non-DHT P2P mechanisms -- perhaps
> >in some very small scale cases we should use flood for simplicity.
> >Perhaps also in emergency cases?
> >
> >--David
> >
>

From RoyR at saic-abingdon.com  Tue Nov  1 12:09:06 2005
From: RoyR at saic-abingdon.com (Roy, Radhika (AEAD))
Date: Tue Nov  1 12:05:51 2005
Subject: [p2p-sip] dhts
Message-ID: <2FE23D25B7292B489A217D1965CDA866089916@bh-exchange02.saic-abingdon.com>

I am providing comments on this, "... different p2p mechanisms."

My understanding has been so far that we will standardize only ONE p2p
mechanism. That is, only ONE p2p protocol that will need algorithms,
transports, security features, and others.

Please note that only ONE p2p protocol, but many algorithms, many
transports, many security features, etc.

Is this assumption not correct?

Radhika

-----Original Message-----
From: davidbryan@gmail.com [mailto:davidbryan@gmail.com] On Behalf Of
David A. Bryan
Sent: Tuesday, November 01, 2005 11:59 AM
To: Roy, Radhika (AEAD)
Cc: p2p-sip@cs.columbia.edu
Subject: Re: [p2p-sip] dhts

I think this is the spirit of the thread, but a warning that in
practice getting this together may be quite difficult, since different
types of p2p implementations exchange different information between
nodes for admission and upkeep of the overlay.

I think if we take this approach, the negotiation of a p2p mechanism
and the implementation thereof (how p2p information is conveyed
between nodes) will have to be two independent things, and I doubt we
can standardize how the information is conveyed between nodes in a way
that doesn't take into account what algotithm is being used. We may
need a negotiation mechanism and then different "standards" for how
the nodes maintain the overlay, since the information contained in
messages and even the pattern and reason for messages exchanged
between nodes can vary greatly for different p2p mechanisms.

David


On 11/1/05, Roy, Radhika (AEAD) <RoyR@saic-abingdon.com> wrote:
> So, the logical conclusion seems to be as follows:
>
> 1. P2P protocol will be algorithm agnostic
>
> 2. P2P protocol will allow capabilities to negotiate any algorithm to
be
> used among the respective peers
>
> 3. P2P scheme will standardize couple of algorithms that can be
> negotiated among the peers starting with a simple one first
>
> It appears that P2P protocol scheme should be flexible enough to use
> algorithms and others per-(each)-peer basis.
>
> To extend this further, it may also be said for transport (reliable,
> unreliable) and security (transport layer, IP layer) that the P2P
> protocol should also use the same principle.
>
> Is that the right summary of this thread so far?
>
> Best regards,
> Radhika
>
> -----Original Message-----
> From: p2p-sip-bounces@cs.columbia.edu
> [mailto:p2p-sip-bounces@cs.columbia.edu] On Behalf Of Eunsoo Shim
> Sent: Tuesday, November 01, 2005 11:15 AM
> To: David A. Bryan
> Cc: p2p-sip@cs.columbia.edu
> Subject: Re: [p2p-sip] dhts
>
> Standardizing a DHT sounds somewhat overwhelming to me.
> But if we put it as standardizing (or writing an experimental epec of
)
> a DHT for SIP-based Internet Telephony, Presence and Instant Messaging
> (i.e., for specific target applications), the task sounds more
> approachable and I feel more comfortable. And I think we have no
choice
> but to specify one or more DHTs for P2P SIP. Otherwise, we won't be
able
> to provide interoperability.
>
> I also agree that we should start with simple algorithms. It is a good
> idea to support negotiation about P2P algorithms.
>
> Regards,
>
> Eunsoo
>
> David A. Bryan wrote:
>
> >I'm fully in agreement here. I think that having multiple algorithms
> >-- and in particular very small simple algorithms as well as more
> >complex algorithms is essential. We are unlikely to be able to choose
> >one that will work in all cases, and are certain to not be able to
> >choose one that will be the "best" for any length of time -- there
are
> >always advances in this area. I think Phillip's analogy to security
is
> >a good one here -- devices should be able to negotiate a P2P
protocol,
> >just as they can with security, encryption, or codec choice. We
should
> >possibly even be able to negotiate non-DHT P2P mechanisms -- perhaps
> >in some very small scale cases we should use flood for simplicity.
> >Perhaps also in emergency cases?
> >
> >--David
> >
>

From bryan at ethernot.org  Tue Nov  1 12:15:51 2005
From: bryan at ethernot.org (David A. Bryan)
Date: Tue Nov  1 12:16:05 2005
Subject: [p2p-sip] dhts
In-Reply-To: <2FE23D25B7292B489A217D1965CDA866089916@bh-exchange02.saic-abingdon.com>
References: <2FE23D25B7292B489A217D1965CDA866089916@bh-exchange02.saic-abingdon.com>
Message-ID: <8b2769930511010915w5291c73n8d5444d547488bce@mail.gmail.com>

I think we might be having a terminology problem here.

I would use the term different algorithms for p2p to mean one could
use chord, epichord, kademlia, CAN, flood, etc, and I think that
people are proposing we be able to have more than one and negotiate
which one.

I'm not sure what you mean by one protocol but many algorithms. One
SIP-based type protocol that lets you use different p2p algorithms
underneath? That is what I am saying may be very difficult, since the
different protocols communicate such different information between
each other, and I opened the suggestion that if we do this, we may
need different documents describing the date exchanged between
different implementations. Whether that is a different protocol or not
is a debatable point, I suppose one could say it is.

Maybe we can find a single p2p mechanism that meets all needs and it
will be a moot point, but I still think if we do, one could come up in
a few years that is better, and we would need to have the flexibility
to use that new p2p mechanism.

David

On 11/1/05, Roy, Radhika (AEAD) <RoyR@saic-abingdon.com> wrote:
> I am providing comments on this, "... different p2p mechanisms."
>
> My understanding has been so far that we will standardize only ONE p2p
> mechanism. That is, only ONE p2p protocol that will need algorithms,
> transports, security features, and others.
>
> Please note that only ONE p2p protocol, but many algorithms, many
> transports, many security features, etc.
>
> Is this assumption not correct?
>
> Radhika
>
> -----Original Message-----
> From: davidbryan@gmail.com [mailto:davidbryan@gmail.com] On Behalf Of
> David A. Bryan
> Sent: Tuesday, November 01, 2005 11:59 AM
> To: Roy, Radhika (AEAD)
> Cc: p2p-sip@cs.columbia.edu
> Subject: Re: [p2p-sip] dhts
>
> I think this is the spirit of the thread, but a warning that in
> practice getting this together may be quite difficult, since different
> types of p2p implementations exchange different information between
> nodes for admission and upkeep of the overlay.
>
> I think if we take this approach, the negotiation of a p2p mechanism
> and the implementation thereof (how p2p information is conveyed
> between nodes) will have to be two independent things, and I doubt we
> can standardize how the information is conveyed between nodes in a way
> that doesn't take into account what algotithm is being used. We may
> need a negotiation mechanism and then different "standards" for how
> the nodes maintain the overlay, since the information contained in
> messages and even the pattern and reason for messages exchanged
> between nodes can vary greatly for different p2p mechanisms.
>
> David
>
>
> On 11/1/05, Roy, Radhika (AEAD) <RoyR@saic-abingdon.com> wrote:
> > So, the logical conclusion seems to be as follows:
> >
> > 1. P2P protocol will be algorithm agnostic
> >
> > 2. P2P protocol will allow capabilities to negotiate any algorithm to
> be
> > used among the respective peers
> >
> > 3. P2P scheme will standardize couple of algorithms that can be
> > negotiated among the peers starting with a simple one first
> >
> > It appears that P2P protocol scheme should be flexible enough to use
> > algorithms and others per-(each)-peer basis.
> >
> > To extend this further, it may also be said for transport (reliable,
> > unreliable) and security (transport layer, IP layer) that the P2P
> > protocol should also use the same principle.
> >
> > Is that the right summary of this thread so far?
> >
> > Best regards,
> > Radhika
> >
> > -----Original Message-----
> > From: p2p-sip-bounces@cs.columbia.edu
> > [mailto:p2p-sip-bounces@cs.columbia.edu] On Behalf Of Eunsoo Shim
> > Sent: Tuesday, November 01, 2005 11:15 AM
> > To: David A. Bryan
> > Cc: p2p-sip@cs.columbia.edu
> > Subject: Re: [p2p-sip] dhts
> >
> > Standardizing a DHT sounds somewhat overwhelming to me.
> > But if we put it as standardizing (or writing an experimental epec of
> )
> > a DHT for SIP-based Internet Telephony, Presence and Instant Messaging
> > (i.e., for specific target applications), the task sounds more
> > approachable and I feel more comfortable. And I think we have no
> choice
> > but to specify one or more DHTs for P2P SIP. Otherwise, we won't be
> able
> > to provide interoperability.
> >
> > I also agree that we should start with simple algorithms. It is a good
> > idea to support negotiation about P2P algorithms.
> >
> > Regards,
> >
> > Eunsoo
> >
> > David A. Bryan wrote:
> >
> > >I'm fully in agreement here. I think that having multiple algorithms
> > >-- and in particular very small simple algorithms as well as more
> > >complex algorithms is essential. We are unlikely to be able to choose
> > >one that will work in all cases, and are certain to not be able to
> > >choose one that will be the "best" for any length of time -- there
> are
> > >always advances in this area. I think Phillip's analogy to security
> is
> > >a good one here -- devices should be able to negotiate a P2P
> protocol,
> > >just as they can with security, encryption, or codec choice. We
> should
> > >possibly even be able to negotiate non-DHT P2P mechanisms -- perhaps
> > >in some very small scale cases we should use flood for simplicity.
> > >Perhaps also in emergency cases?
> > >
> > >--David
> > >
> >
>

From ag at ag-projects.com  Tue Nov  1 12:24:02 2005
From: ag at ag-projects.com (Adrian Georgescu)
Date: Tue Nov  1 12:25:13 2005
Subject: [p2p-sip] Draft: Requirements for
	SIP-basedPeer-to-PeerInternetTelephony
In-Reply-To: <2FE23D25B7292B489A217D1965CDA866089915@bh-exchange02.saic-abingdon.com>
References: <2FE23D25B7292B489A217D1965CDA866089915@bh-exchange02.saic-abingdon.com>
Message-ID: <1c61e61ab72ac099b6c0a89501daaa6a@ag-projects.com>

I would imagine the purpose of an IETF WG would be to design a simple 
protocol/algorithm that will allow me and my friends to meet at IETF, 
light up our wifi SIP devices and make an ad-hoc network where we can 
connect to each other at SIP application level like 802.X does at layer 
2. For this ad-hoc network a simple standard P2P could be standardized 
among vendors.

The most relevant (for me) requirements for developing SIP P2P would be:

4.5.  NAT and Firewall Traversal

How to solve NAT and allocating a third party media relay in a clever 
manner when necessary and based on some policy. I am not sure one can 
simply assume any P2P SIP device can do NAT traversal for others it 
might not have the resources or desire to do this.

4.1.  Resources for Distribution

Have an ad-hoc P2P location service, ONE and only ONE simple and 
compatible P2P algorithm among peers

5.5.  Interconnect with P2P, non-P2P and PSTN networks

Link the P2P ad-hoc location service with permanent location services 
provided by the DNS

4.3.  DHT Changeability and 6.1.  Support for different DHTs

This seems difficult to achieve and should not be a requirement. 
Imagining a complex scenario where DHT protocol A meets protocol B will 
introduce yet another type of gateway, "The P2P Gateway", we do not 
need yet another gateway we just need a distributed location service 
that interfaces with the DNS for public identities. We have DNS and SIP 
registrar functions and we could use them right and marry them with the 
P2P DHT algorithm.

Regards,
Adrian

On Nov 1, 2005, at 5:49 PM, Roy, Radhika ((AEAD)) wrote:

> So, the logical conclusion seems to be as follows:
>
> 1. P2P protocol will be algorithm agnostic
>
> 2. P2P protocol will allow capabilities to negotiate any algorithm to 
> be
> used among the respective peers
>
> 3. P2P scheme will standardize couple of algorithms that can be
> negotiated among the peers starting with a simple one first
>
> It appears that P2P protocol scheme should be flexible enough to use
> algorithms and others per-(each)-peer basis.
>
> To extend this further, it may also be said for transport (reliable,
> unreliable) and security (transport layer, IP layer) that the P2P
> protocol should also use the same principle.
>
> Is that the right summary of this thread so far?
>
> Best regards,
> Radhika
>
> -----Original Message-----
> From: p2p-sip-bounces@cs.columbia.edu
> [mailto:p2p-sip-bounces@cs.columbia.edu] On Behalf Of Eunsoo Shim
> Sent: Tuesday, November 01, 2005 11:15 AM
> To: David A. Bryan
> Cc: p2p-sip@cs.columbia.edu
> Subject: Re: [p2p-sip] dhts
>
> Standardizing a DHT sounds somewhat overwhelming to me.
> But if we put it as standardizing (or writing an experimental epec of )
> a DHT for SIP-based Internet Telephony, Presence and Instant Messaging
> (i.e., for specific target applications), the task sounds more
> approachable and I feel more comfortable. And I think we have no choice
> but to specify one or more DHTs for P2P SIP. Otherwise, we won't be 
> able
> to provide interoperability.
>
> I also agree that we should start with simple algorithms. It is a good
> idea to support negotiation about P2P algorithms.
>
> Regards,
>
> Eunsoo
>
> David A. Bryan wrote:
>
>> I'm fully in agreement here. I think that having multiple algorithms
>> -- and in particular very small simple algorithms as well as more
>> complex algorithms is essential. We are unlikely to be able to choose
>> one that will work in all cases, and are certain to not be able to
>> choose one that will be the "best" for any length of time -- there are
>> always advances in this area. I think Phillip's analogy to security is
>> a good one here -- devices should be able to negotiate a P2P protocol,
>> just as they can with security, encryption, or codec choice. We should
>> possibly even be able to negotiate non-DHT P2P mechanisms -- perhaps
>> in some very small scale cases we should use flood for simplicity.
>> Perhaps also in emergency cases?
>>
>> --David
>>
>
> _______________________________________________
> p2p-sip mailing list
> p2p-sip@cs.columbia.edu
> http://lists.cs.columbia.edu/mailman/listinfo/p2p-sip

From matthewsp at avaya.com  Tue Nov  1 14:29:09 2005
From: matthewsp at avaya.com (Matthews, Philip (Philip))
Date: Tue Nov  1 14:32:51 2005
Subject: [p2p-sip] dhts
Message-ID: <DFD0AADB452F8B4E99FB78719CAF0E61019633CC@ON0010AVEXU1.global.avaya.com>

I agree with David.

I think we could standardize a negotiation framework
(which is really a capabilities advertisement mechanism)
along with rules for how to select between alternatives.

Along this line, I am thinking that the hierarchical P2P network
discussed in a separate thread on this list (and the referenced paper)
is looking increasingly attractive.
It allows different groups of peers to use different DHT algorithms.
Say for example, one group of peers (group A) wishes to use algorithm X,
while another group (group B) wishes to use algorithm Y, where both X
and Y
are advanced DHT algorithms. If all nodes also speak some basic
P2P algorithm Z, then we have a hierarchy, with group A nodes using
algorithm X
internally, group B nodes using algorithm Y internally, and the two
groups themselves
being treated as two nodes in a second level, which speaks algorithm Z.
Note that algorithm Z does not have to scale to lots of nodes, since
at its level there are only two: the node representing group A and the
node representing
group B. 

How groups A and B form are good questions. Does a node belong to group
A
simply because it speaks protocol X, or does group A has some other
natural
cohesive grouping? For example, do a bunch of nodes speaking David's
SoSimple
P2P protocol find each other through some (external?) mechanism,
and a bunch of nodes speaking the Nimcat P2P protocol find each other
through
another mechanism, and then the two groups talk using protocol Z?

If others agree with me that a hierarchical approach is the way to go,
then I think that we might be able to:
- Standardize a framework for hierarchical P2P networks
- Standardize a negotiation mechanism for the P2P protocol being used in
a group
- Standardize a simple protocol Z as the default P2P protocol

I agree with David's comment that the pattern (and format) of messages
used
in one P2P protocol may differ quite a bit from the messages used in a
different
protocol. I am not sure how successful we would be if we tried to
standardize
(for example) the message format, and leave only the algorithm open.


- Philip



> -----Original Message-----
> From: p2p-sip-bounces@cs.columbia.edu 
> [mailto:p2p-sip-bounces@cs.columbia.edu] On Behalf Of David A. Bryan
> Sent: November 1, 2005 11:59
> To: Roy, Radhika (AEAD)
> Cc: p2p-sip@cs.columbia.edu
> Subject: Re: [p2p-sip] dhts
> 
> I think this is the spirit of the thread, but a warning that in
> practice getting this together may be quite difficult, since different
> types of p2p implementations exchange different information between
> nodes for admission and upkeep of the overlay.
> 
> I think if we take this approach, the negotiation of a p2p mechanism
> and the implementation thereof (how p2p information is conveyed
> between nodes) will have to be two independent things, and I doubt we
> can standardize how the information is conveyed between nodes in a way
> that doesn't take into account what algotithm is being used. We may
> need a negotiation mechanism and then different "standards" for how
> the nodes maintain the overlay, since the information contained in
> messages and even the pattern and reason for messages exchanged
> between nodes can vary greatly for different p2p mechanisms.
> 
> David
> 
> 
> On 11/1/05, Roy, Radhika (AEAD) <RoyR@saic-abingdon.com> wrote:
> > So, the logical conclusion seems to be as follows:
> >
> > 1. P2P protocol will be algorithm agnostic
> >
> > 2. P2P protocol will allow capabilities to negotiate any 
> algorithm to be
> > used among the respective peers
> >
> > 3. P2P scheme will standardize couple of algorithms that can be
> > negotiated among the peers starting with a simple one first
> >
> > It appears that P2P protocol scheme should be flexible enough to use
> > algorithms and others per-(each)-peer basis.
> >
> > To extend this further, it may also be said for transport (reliable,
> > unreliable) and security (transport layer, IP layer) that the P2P
> > protocol should also use the same principle.
> >
> > Is that the right summary of this thread so far?
> >
> > Best regards,
> > Radhika
> >
> > -----Original Message-----
> > From: p2p-sip-bounces@cs.columbia.edu
> > [mailto:p2p-sip-bounces@cs.columbia.edu] On Behalf Of Eunsoo Shim
> > Sent: Tuesday, November 01, 2005 11:15 AM
> > To: David A. Bryan
> > Cc: p2p-sip@cs.columbia.edu
> > Subject: Re: [p2p-sip] dhts
> >
> > Standardizing a DHT sounds somewhat overwhelming to me.
> > But if we put it as standardizing (or writing an 
> experimental epec of )
> > a DHT for SIP-based Internet Telephony, Presence and 
> Instant Messaging
> > (i.e., for specific target applications), the task sounds more
> > approachable and I feel more comfortable. And I think we 
> have no choice
> > but to specify one or more DHTs for P2P SIP. Otherwise, we 
> won't be able
> > to provide interoperability.
> >
> > I also agree that we should start with simple algorithms. 
> It is a good
> > idea to support negotiation about P2P algorithms.
> >
> > Regards,
> >
> > Eunsoo
> >
> > David A. Bryan wrote:
> >
> > >I'm fully in agreement here. I think that having multiple 
> algorithms
> > >-- and in particular very small simple algorithms as well as more
> > >complex algorithms is essential. We are unlikely to be 
> able to choose
> > >one that will work in all cases, and are certain to not be able to
> > >choose one that will be the "best" for any length of time 
> -- there are
> > >always advances in this area. I think Phillip's analogy to 
> security is
> > >a good one here -- devices should be able to negotiate a 
> P2P protocol,
> > >just as they can with security, encryption, or codec 
> choice. We should
> > >possibly even be able to negotiate non-DHT P2P mechanisms 
> -- perhaps
> > >in some very small scale cases we should use flood for simplicity.
> > >Perhaps also in emergency cases?
> > >
> > >--David
> > >
> >
> 
> _______________________________________________
> p2p-sip mailing list
> p2p-sip@cs.columbia.edu
> http://lists.cs.columbia.edu/mailman/listinfo/p2p-sip
> 

From marc1.bailly at rd.francetelecom.com  Tue Nov  1 16:15:04 2005
From: marc1.bailly at rd.francetelecom.com (BAILLY Marc RD-CORE-LAN)
Date: Tue Nov  1 16:15:40 2005
Subject: [p2p-sip] dhts
Message-ID: <BAF83494CE653943A97B9F755016A06601F80325@ftrdmel1.rd.francetelecom.fr>

Hi,

See my comments inline.

Best regards.

Marc

> 
> I agree with David.
I agree too.
> 
> I think we could standardize a negotiation framework
> (which is really a capabilities advertisement mechanism)
> along with rules for how to select between alternatives.
> 
> Along this line, I am thinking that the hierarchical P2P network
> discussed in a separate thread on this list (and the referenced paper)
> is looking increasingly attractive.
> It allows different groups of peers to use different DHT algorithms.
> Say for example, one group of peers (group A) wishes to use 
> algorithm X,
> while another group (group B) wishes to use algorithm Y, where both X
> and Y
> are advanced DHT algorithms. If all nodes also speak some basic
> P2P algorithm Z, then we have a hierarchy, with group A nodes using
> algorithm X
> internally, group B nodes using algorithm Y internally, and the two
> groups themselves
> being treated as two nodes in a second level, which speaks 
> algorithm Z.
> Note that algorithm Z does not have to scale to lots of nodes, since
> at its level there are only two: the node representing group A and the
> node representing
> group B. 
> 
> (...)
> If others agree with me that a hierarchical approach is the way to go,
> then I think that we might be able to:
> - Standardize a framework for hierarchical P2P networks
> - Standardize a negotiation mechanism for the P2P protocol 
> being used in
> a group
> - Standardize a simple protocol Z as the default P2P protocol

I also think that it's ONE of the way to go.
The standardization of protocol Z for interconnection should be possible
because we may have some assumptions about the nodes (stability,
ressources, number of nodes...) which belong to the top level DHT.
But I'm not sure that this protocol Z, if used for interconnection, can
be called the default P2P protocol.
Indeed, if this protcol needs some adapatations as proposed for ewample
in the paper "Hierarchical P2P systems" : "In the top-level overlay
network, each "node" is actually a group of peers".
This adaptations would not be useful if this P2P SIP protocol is used in
a specific domain.



> 
> I agree with David's comment that the pattern (and format) of messages
> used
> in one P2P protocol may differ quite a bit from the messages used in a
> different
> protocol. I am not sure how successful we would be if we tried to
> standardize
> (for example) the message format, and leave only the algorithm open.

I also have some doubts about this point.

> 
> 
> - Philip
> 
>  
> 
> 
> > -----Original Message-----
> > From: p2p-sip-bounces@cs.columbia.edu 
> > [mailto:p2p-sip-bounces@cs.columbia.edu] On Behalf Of David A. Bryan
> > Sent: November 1, 2005 11:59
> > To: Roy, Radhika (AEAD)
> > Cc: p2p-sip@cs.columbia.edu
> > Subject: Re: [p2p-sip] dhts
> > 
> > I think this is the spirit of the thread, but a warning that in
> > practice getting this together may be quite difficult, 
> since different
> > types of p2p implementations exchange different information between
> > nodes for admission and upkeep of the overlay.
> > 
> > I think if we take this approach, the negotiation of a p2p mechanism
> > and the implementation thereof (how p2p information is conveyed
> > between nodes) will have to be two independent things, and 
> I doubt we
> > can standardize how the information is conveyed between 
> nodes in a way
> > that doesn't take into account what algotithm is being used. We may
> > need a negotiation mechanism and then different "standards" for how
> > the nodes maintain the overlay, since the information contained in
> > messages and even the pattern and reason for messages exchanged
> > between nodes can vary greatly for different p2p mechanisms.
> > 
> > David
> > 
> > 
> > On 11/1/05, Roy, Radhika (AEAD) <RoyR@saic-abingdon.com> wrote:
> > > So, the logical conclusion seems to be as follows:
> > >
> > > 1. P2P protocol will be algorithm agnostic
> > >
> > > 2. P2P protocol will allow capabilities to negotiate any 
> > algorithm to be
> > > used among the respective peers
> > >
> > > 3. P2P scheme will standardize couple of algorithms that can be
> > > negotiated among the peers starting with a simple one first
> > >
> > > It appears that P2P protocol scheme should be flexible 
> enough to use
> > > algorithms and others per-(each)-peer basis.
> > >
> > > To extend this further, it may also be said for transport 
> (reliable,
> > > unreliable) and security (transport layer, IP layer) that the P2P
> > > protocol should also use the same principle.
> > >
> > > Is that the right summary of this thread so far?
> > >
> > > Best regards,
> > > Radhika
> > >
> > > -----Original Message-----
> > > From: p2p-sip-bounces@cs.columbia.edu
> > > [mailto:p2p-sip-bounces@cs.columbia.edu] On Behalf Of Eunsoo Shim
> > > Sent: Tuesday, November 01, 2005 11:15 AM
> > > To: David A. Bryan
> > > Cc: p2p-sip@cs.columbia.edu
> > > Subject: Re: [p2p-sip] dhts
> > >
> > > Standardizing a DHT sounds somewhat overwhelming to me.
> > > But if we put it as standardizing (or writing an 
> > experimental epec of )
> > > a DHT for SIP-based Internet Telephony, Presence and 
> > Instant Messaging
> > > (i.e., for specific target applications), the task sounds more
> > > approachable and I feel more comfortable. And I think we 
> > have no choice
> > > but to specify one or more DHTs for P2P SIP. Otherwise, we 
> > won't be able
> > > to provide interoperability.
> > >
> > > I also agree that we should start with simple algorithms. 
> > It is a good
> > > idea to support negotiation about P2P algorithms.
> > >
> > > Regards,
> > >
> > > Eunsoo
> > >
> > > David A. Bryan wrote:
> > >
> > > >I'm fully in agreement here. I think that having multiple 
> > algorithms
> > > >-- and in particular very small simple algorithms as well as more
> > > >complex algorithms is essential. We are unlikely to be 
> > able to choose
> > > >one that will work in all cases, and are certain to not 
> be able to
> > > >choose one that will be the "best" for any length of time 
> > -- there are
> > > >always advances in this area. I think Phillip's analogy to 
> > security is
> > > >a good one here -- devices should be able to negotiate a 
> > P2P protocol,
> > > >just as they can with security, encryption, or codec 
> > choice. We should
> > > >possibly even be able to negotiate non-DHT P2P mechanisms 
> > -- perhaps
> > > >in some very small scale cases we should use flood for 
> simplicity.
> > > >Perhaps also in emergency cases?
> > > >
> > > >--David
> > > >
> > >
> > 
> > _______________________________________________
> > p2p-sip mailing list
> > p2p-sip@cs.columbia.edu
> > http://lists.cs.columbia.edu/mailman/listinfo/p2p-sip
> > 
> 
> _______________________________________________
> p2p-sip mailing list
> p2p-sip@cs.columbia.edu
> http://lists.cs.columbia.edu/mailman/listinfo/p2p-sip
> 

From RoyR at saic-abingdon.com  Tue Nov  1 16:26:08 2005
From: RoyR at saic-abingdon.com (Roy, Radhika (AEAD))
Date: Tue Nov  1 16:23:10 2005
Subject: [p2p-sip] dhts
Message-ID: <2FE23D25B7292B489A217D1965CDA866089918@bh-exchange02.saic-abingdon.com>

Hi, David and Philip:

It is not that I am disagreeing with you. In fact, Philip has answered
my main question that we will standardize a p2p protocol. Like others, I
also definitely agree with Philip as he has proposed, but in reverse
order:

1.	Standardize a simple protocol Z as the default P2P protocol
2.	Standardize a negotiation mechanism for the P2P protocol being
used in a group
3.	Standardize a framework for hierarchical P2P networks

Now, if we start with item 1, we have to see couple of basic items
(something) as follows:

a.	Discovery (as a part of the p2p protocol)
b.	Lookup (as a part of the p2p protocol)
c.	Content Transfer between peers directly (will be triggered or
invoked as a result of p2p [signaling] protocol)
d.	Use of existing transport protocol(s) depending on the needs of
the p2p protocol
e.	Use of existing security protocol(s) depending on the needs of
the p2p protocol

If we agree with the above, we can then see where DHT and other
algorithms will stand, and how we should address all aspects of p2p
protocol in a consistent way taking some little baby steps in the
beginning.

Best regards,
Radhika

-----Original Message-----
From: matthewsp@avaya.com [mailto:matthewsp@avaya.com] 
Sent: Tuesday, November 01, 2005 2:29 PM
To: David A. Bryan; Roy, Radhika (AEAD); Eunsoo Shim
Cc: p2p-sip@cs.columbia.edu
Subject: RE: [p2p-sip] dhts

I agree with David.

I think we could standardize a negotiation framework
(which is really a capabilities advertisement mechanism)
along with rules for how to select between alternatives.

Along this line, I am thinking that the hierarchical P2P network
discussed in a separate thread on this list (and the referenced paper)
is looking increasingly attractive.
It allows different groups of peers to use different DHT algorithms.
Say for example, one group of peers (group A) wishes to use algorithm X,
while another group (group B) wishes to use algorithm Y, where both X
and Y
are advanced DHT algorithms. If all nodes also speak some basic
P2P algorithm Z, then we have a hierarchy, with group A nodes using
algorithm X
internally, group B nodes using algorithm Y internally, and the two
groups themselves
being treated as two nodes in a second level, which speaks algorithm Z.
Note that algorithm Z does not have to scale to lots of nodes, since
at its level there are only two: the node representing group A and the
node representing
group B. 

How groups A and B form are good questions. Does a node belong to group
A
simply because it speaks protocol X, or does group A has some other
natural
cohesive grouping? For example, do a bunch of nodes speaking David's
SoSimple
P2P protocol find each other through some (external?) mechanism,
and a bunch of nodes speaking the Nimcat P2P protocol find each other
through
another mechanism, and then the two groups talk using protocol Z?

If others agree with me that a hierarchical approach is the way to go,
then I think that we might be able to:
- Standardize a framework for hierarchical P2P networks
- Standardize a negotiation mechanism for the P2P protocol being used in
a group
- Standardize a simple protocol Z as the default P2P protocol

I agree with David's comment that the pattern (and format) of messages
used
in one P2P protocol may differ quite a bit from the messages used in a
different
protocol. I am not sure how successful we would be if we tried to
standardize
(for example) the message format, and leave only the algorithm open.


- Philip



> -----Original Message-----
> From: p2p-sip-bounces@cs.columbia.edu 
> [mailto:p2p-sip-bounces@cs.columbia.edu] On Behalf Of David A. Bryan
> Sent: November 1, 2005 11:59
> To: Roy, Radhika (AEAD)
> Cc: p2p-sip@cs.columbia.edu
> Subject: Re: [p2p-sip] dhts
> 
> I think this is the spirit of the thread, but a warning that in
> practice getting this together may be quite difficult, since different
> types of p2p implementations exchange different information between
> nodes for admission and upkeep of the overlay.
> 
> I think if we take this approach, the negotiation of a p2p mechanism
> and the implementation thereof (how p2p information is conveyed
> between nodes) will have to be two independent things, and I doubt we
> can standardize how the information is conveyed between nodes in a way
> that doesn't take into account what algorithm is being used. We may
> need a negotiation mechanism and then different "standards" for how
> the nodes maintain the overlay, since the information contained in
> messages and even the pattern and reason for messages exchanged
> between nodes can vary greatly for different p2p mechanisms.
> 
> David
> 
> 
> On 11/1/05, Roy, Radhika (AEAD) <RoyR@saic-abingdon.com> wrote:
> > So, the logical conclusion seems to be as follows:
> >
> > 1. P2P protocol will be algorithm agnostic
> >
> > 2. P2P protocol will allow capabilities to negotiate any 
> algorithm to be
> > used among the respective peers
> >
> > 3. P2P scheme will standardize couple of algorithms that can be
> > negotiated among the peers starting with a simple one first
> >
> > It appears that P2P protocol scheme should be flexible enough to use
> > algorithms and others per-(each)-peer basis.
> >
> > To extend this further, it may also be said for transport (reliable,
> > unreliable) and security (transport layer, IP layer) that the P2P
> > protocol should also use the same principle.
> >
> > Is that the right summary of this thread so far?
> >
> > Best regards,
> > Radhika
> >
> > -----Original Message-----
> > From: p2p-sip-bounces@cs.columbia.edu
> > [mailto:p2p-sip-bounces@cs.columbia.edu] On Behalf Of Eunsoo Shim
> > Sent: Tuesday, November 01, 2005 11:15 AM
> > To: David A. Bryan
> > Cc: p2p-sip@cs.columbia.edu
> > Subject: Re: [p2p-sip] dhts
> >
> > Standardizing a DHT sounds somewhat overwhelming to me.
> > But if we put it as standardizing (or writing an 
> experimental epec of )
> > a DHT for SIP-based Internet Telephony, Presence and 
> Instant Messaging
> > (i.e., for specific target applications), the task sounds more
> > approachable and I feel more comfortable. And I think we 
> have no choice
> > but to specify one or more DHTs for P2P SIP. Otherwise, we 
> won't be able
> > to provide interoperability.
> >
> > I also agree that we should start with simple algorithms. 
> It is a good
> > idea to support negotiation about P2P algorithms.
> >
> > Regards,
> >
> > Eunsoo
> >
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://lists.cs.columbia.edu/pipermail/p2p-sip/attachments/20051101/98b627d6/attachment-0001.html
From mahesh.govind at samsung.com  Tue Nov  1 19:00:40 2005
From: mahesh.govind at samsung.com (Mahesh Govind)
Date: Tue Nov  1 19:02:25 2005
Subject: [p2p-sip] dhts
In-Reply-To: <2FE23D25B7292B489A217D1965CDA866089918@bh-exchange02.saic-abingdon.com>
Message-ID: <004f01c5df40$762c1a70$fbcbd5a5@mahesh>

Hi,

I agree with this . I think discovery of my network is the first step for
locating my peer and once I am inside my network, I can find the peer,

Using some means which my network understands.

 

With regards

Mahesh

 

 

-----Original Message-----
From: p2p-sip-bounces@cs.columbia.edu
[mailto:p2p-sip-bounces@cs.columbia.edu] On Behalf Of Roy, Radhika (AEAD)
Sent: Wednesday, November 02, 2005 6:26 AM
To: matthewsp@avaya.com; David A. Bryan; Eunsoo Shim
Cc: p2p-sip@cs.columbia.edu
Subject: RE: [p2p-sip] dhts

 

Hi, David and Philip:

It is not that I am disagreeing with you. In fact, Philip has answered my
main question that we will standardize a p2p protocol. Like others, I also
definitely agree with Philip as he has proposed, but in reverse order:

1.      Standardize a simple protocol Z as the default P2P protocol 
2.      Standardize a negotiation mechanism for the P2P protocol being used
in a group 
3.      Standardize a framework for hierarchical P2P networks 

Now, if we start with item 1, we have to see couple of basic items
(something) as follows:

a.      Discovery (as a part of the p2p protocol) 
b.      Lookup (as a part of the p2p protocol) 
c.      Content Transfer between peers directly (will be triggered or
invoked as a result of p2p [signaling] protocol) 
d.      Use of existing transport protocol(s) depending on the needs of the
p2p protocol 
e.      Use of existing security protocol(s) depending on the needs of the
p2p protocol 

If we agree with the above, we can then see where DHT and other algorithms
will stand, and how we should address all aspects of p2p protocol in a
consistent way taking some little baby steps in the beginning.

Best regards,

Radhika

-----Original Message-----
From: matthewsp@avaya.com [mailto:matthewsp@avaya.com]
Sent: Tuesday, November 01, 2005 2:29 PM
To: David A. Bryan; Roy, Radhika (AEAD); Eunsoo Shim
Cc: p2p-sip@cs.columbia.edu
Subject: RE: [p2p-sip] dhts

I agree with David.

I think we could standardize a negotiation framework

(which is really a capabilities advertisement mechanism)

along with rules for how to select between alternatives.

Along this line, I am thinking that the hierarchical P2P network

discussed in a separate thread on this list (and the referenced paper)

is looking increasingly attractive.

It allows different groups of peers to use different DHT algorithms.

Say for example, one group of peers (group A) wishes to use algorithm X,

while another group (group B) wishes to use algorithm Y, where both X

and Y

are advanced DHT algorithms. If all nodes also speak some basic

P2P algorithm Z, then we have a hierarchy, with group A nodes using

algorithm X

internally, group B nodes using algorithm Y internally, and the two

groups themselves

being treated as two nodes in a second level, which speaks algorithm Z.

Note that algorithm Z does not have to scale to lots of nodes, since

at its level there are only two: the node representing group A and the

node representing

group B. 

How groups A and B form are good questions. Does a node belong to group

A

simply because it speaks protocol X, or does group A has some other

natural

cohesive grouping? For example, do a bunch of nodes speaking David's

SoSimple

P2P protocol find each other through some (external?) mechanism,

and a bunch of nodes speaking the Nimcat P2P protocol find each other

through

another mechanism, and then the two groups talk using protocol Z?

If others agree with me that a hierarchical approach is the way to go,

then I think that we might be able to:

- Standardize a framework for hierarchical P2P networks

- Standardize a negotiation mechanism for the P2P protocol being used in

a group

- Standardize a simple protocol Z as the default P2P protocol

I agree with David's comment that the pattern (and format) of messages

used

in one P2P protocol may differ quite a bit from the messages used in a

different

protocol. I am not sure how successful we would be if we tried to

standardize

(for example) the message format, and leave only the algorithm open.

 

- Philip

 

> -----Original Message-----

> From: p2p-sip-bounces@cs.columbia.edu 

> [mailto:p2p-sip-bounces@cs.columbia.edu] On Behalf Of David A. Bryan

> Sent: November 1, 2005 11:59

> To: Roy, Radhika (AEAD)

> Cc: p2p-sip@cs.columbia.edu

> Subject: Re: [p2p-sip] dhts

> 

> I think this is the spirit of the thread, but a warning that in

> practice getting this together may be quite difficult, since different

> types of p2p implementations exchange different information between

> nodes for admission and upkeep of the overlay.

> 

> I think if we take this approach, the negotiation of a p2p mechanism

> and the implementation thereof (how p2p information is conveyed

> between nodes) will have to be two independent things, and I doubt we

> can standardize how the information is conveyed between nodes in a way

> that doesn't take into account what algorithm is being used. We may

> need a negotiation mechanism and then different "standards" for how

> the nodes maintain the overlay, since the information contained in

> messages and even the pattern and reason for messages exchanged

> between nodes can vary greatly for different p2p mechanisms.

> 

> David

> 

> 

> On 11/1/05, Roy, Radhika (AEAD) <RoyR@saic-abingdon.com> wrote:

> > So, the logical conclusion seems to be as follows:

> >

> > 1. P2P protocol will be algorithm agnostic

> >

> > 2. P2P protocol will allow capabilities to negotiate any 

> algorithm to be

> > used among the respective peers

> >

> > 3. P2P scheme will standardize couple of algorithms that can be

> > negotiated among the peers starting with a simple one first

> >

> > It appears that P2P protocol scheme should be flexible enough to use

> > algorithms and others per-(each)-peer basis.

> >

> > To extend this further, it may also be said for transport (reliable,

> > unreliable) and security (transport layer, IP layer) that the P2P

> > protocol should also use the same principle.

> >

> > Is that the right summary of this thread so far?

> >

> > Best regards,

> > Radhika

> >

> > -----Original Message-----

> > From: p2p-sip-bounces@cs.columbia.edu

> > [mailto:p2p-sip-bounces@cs.columbia.edu] On Behalf Of Eunsoo Shim

> > Sent: Tuesday, November 01, 2005 11:15 AM

> > To: David A. Bryan

> > Cc: p2p-sip@cs.columbia.edu

> > Subject: Re: [p2p-sip] dhts

> >

> > Standardizing a DHT sounds somewhat overwhelming to me.

> > But if we put it as standardizing (or writing an 

> experimental epec of )

> > a DHT for SIP-based Internet Telephony, Presence and 

> Instant Messaging

> > (i.e., for specific target applications), the task sounds more

> > approachable and I feel more comfortable. And I think we 

> have no choice

> > but to specify one or more DHTs for P2P SIP. Otherwise, we 

> won't be able

> > to provide interoperability.

> >

> > I also agree that we should start with simple algorithms. 

> It is a good

> > idea to support negotiation about P2P algorithms.

> >

> > Regards,

> >

> > Eunsoo

> >

-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://lists.cs.columbia.edu/pipermail/p2p-sip/attachments/20051102/7eb09a2b/attachment-0001.html
From eunsoo at research.panasonic.com  Wed Nov  2 05:37:02 2005
From: eunsoo at research.panasonic.com (Eunsoo Shim)
Date: Wed Nov  2 05:37:25 2005
Subject: [p2p-sip] dhts
In-Reply-To: <DFD0AADB452F8B4E99FB78719CAF0E61019633CC@ON0010AVEXU1.global.a
	vaya.com>
References: <DFD0AADB452F8B4E99FB78719CAF0E61019633CC@ON0010AVEXU1.global.av
	aya.com>
Message-ID: <436896CE.4070703@research.panasonic.com>

Hmm.... probably another terminology issue.
I think 'hierarchical architecture' can have two meanings in P2P SIP.
One is something like supernodes and non-supernodes where supernodes are
not necessarily related to domain boundaries. In that sense, you can
have a hierarchical P2P network even within a (administrative) domain.
Another is something like federation of domains with inter-domain
networking.
That is, 'hierarchical architecture' does not necessarily mean
multi-domain or federation fo domains.
I guess Philip is using the term as federation of domains.

I think an architecture supporting federation of domains has signficant
merit in that it provides scalability and flexibility in configuring
individual domains and we need to support it eventually. The idea of
algorithm Z for inter-domain operation and algorithm X and Y for
intra-domain operation seems quite natural. It reminds me of
inter-domain routing protocols and intra-domain routing protocols in IP
routing. However, we can see that an algorithm for inter-domain
operation is not necessarily the simplest or default algorithm in IP
routing. This may not be necessarily the case for P2P SIP but I think it
provides enough motivation for us to be cautious. And most likely
dealing with federation of domains will add more complexity to the
architecture and algorithm even though I don't know how much.

I think we agree that we should start with simple algorithms. So I'd say
we look for simple and reasonably good overlay algorithms in general
rather than specifically something for inter-domain operation in the
beginning. That is, I'd suggest we should first look for simple and
reasonably good algorithm X and Y for individual domains.


Regarding one protocol and multiple algorithms, I also agree that
different algorithms are likely to have differences in information in
messages and even in message flows. What I thought was something like a
common message structure as the base and different payloads for
different overlay algorithms. It is debateable whether such a thing is
one protocol or multiple protocols as David pointed. But that approach
looks better to me than creating totally new and different message
structures for each overlay algorithm. It seems feasible at least for
DHTs. What do you think?

Regards,

Eunsoo


Matthews, Philip (Philip) wrote:

>I agree with David.
>
>I think we could standardize a negotiation framework
>(which is really a capabilities advertisement mechanism)
>along with rules for how to select between alternatives.
>
>Along this line, I am thinking that the hierarchical P2P network
>discussed in a separate thread on this list (and the referenced paper)
>is looking increasingly attractive.
>It allows different groups of peers to use different DHT algorithms.
>Say for example, one group of peers (group A) wishes to use algorithm X,
>while another group (group B) wishes to use algorithm Y, where both X
>and Y
>are advanced DHT algorithms. If all nodes also speak some basic
>P2P algorithm Z, then we have a hierarchy, with group A nodes using
>algorithm X
>internally, group B nodes using algorithm Y internally, and the two
>groups themselves
>being treated as two nodes in a second level, which speaks algorithm Z.
>Note that algorithm Z does not have to scale to lots of nodes, since
>at its level there are only two: the node representing group A and the
>node representing
>group B. 
>
>How groups A and B form are good questions. Does a node belong to group
>A
>simply because it speaks protocol X, or does group A has some other
>natural
>cohesive grouping? For example, do a bunch of nodes speaking David's
>SoSimple
>P2P protocol find each other through some (external?) mechanism,
>and a bunch of nodes speaking the Nimcat P2P protocol find each other
>through
>another mechanism, and then the two groups talk using protocol Z?
>
>If others agree with me that a hierarchical approach is the way to go,
>then I think that we might be able to:
>- Standardize a framework for hierarchical P2P networks
>- Standardize a negotiation mechanism for the P2P protocol being used in
>a group
>- Standardize a simple protocol Z as the default P2P protocol
>
>I agree with David's comment that the pattern (and format) of messages
>used
>in one P2P protocol may differ quite a bit from the messages used in a
>different
>protocol. I am not sure how successful we would be if we tried to
>standardize
>(for example) the message format, and leave only the algorithm open.
>
>
>- Philip
>
>
>
>  
>
>>-----Original Message-----
>>From: p2p-sip-bounces@cs.columbia.edu 
>>[mailto:p2p-sip-bounces@cs.columbia.edu] On Behalf Of David A. Bryan
>>Sent: November 1, 2005 11:59
>>To: Roy, Radhika (AEAD)
>>Cc: p2p-sip@cs.columbia.edu
>>Subject: Re: [p2p-sip] dhts
>>
>>I think this is the spirit of the thread, but a warning that in
>>practice getting this together may be quite difficult, since different
>>types of p2p implementations exchange different information between
>>nodes for admission and upkeep of the overlay.
>>
>>I think if we take this approach, the negotiation of a p2p mechanism
>>and the implementation thereof (how p2p information is conveyed
>>between nodes) will have to be two independent things, and I doubt we
>>can standardize how the information is conveyed between nodes in a way
>>that doesn't take into account what algotithm is being used. We may
>>need a negotiation mechanism and then different "standards" for how
>>the nodes maintain the overlay, since the information contained in
>>messages and even the pattern and reason for messages exchanged
>>between nodes can vary greatly for different p2p mechanisms.
>>
>>David
>>
>>
>>On 11/1/05, Roy, Radhika (AEAD) <RoyR@saic-abingdon.com> wrote:
>>    
>>
>>>So, the logical conclusion seems to be as follows:
>>>
>>>1. P2P protocol will be algorithm agnostic
>>>
>>>2. P2P protocol will allow capabilities to negotiate any 
>>>      
>>>
>>algorithm to be
>>    
>>
>>>used among the respective peers
>>>
>>>3. P2P scheme will standardize couple of algorithms that can be
>>>negotiated among the peers starting with a simple one first
>>>
>>>It appears that P2P protocol scheme should be flexible enough to use
>>>algorithms and others per-(each)-peer basis.
>>>
>>>To extend this further, it may also be said for transport (reliable,
>>>unreliable) and security (transport layer, IP layer) that the P2P
>>>protocol should also use the same principle.
>>>
>>>Is that the right summary of this thread so far?
>>>
>>>Best regards,
>>>Radhika
>>>
>>>-----Original Message-----
>>>From: p2p-sip-bounces@cs.columbia.edu
>>>[mailto:p2p-sip-bounces@cs.columbia.edu] On Behalf Of Eunsoo Shim
>>>Sent: Tuesday, November 01, 2005 11:15 AM
>>>To: David A. Bryan
>>>Cc: p2p-sip@cs.columbia.edu
>>>Subject: Re: [p2p-sip] dhts
>>>
>>>Standardizing a DHT sounds somewhat overwhelming to me.
>>>But if we put it as standardizing (or writing an 
>>>      
>>>
>>experimental epec of )
>>    
>>
>>>a DHT for SIP-based Internet Telephony, Presence and 
>>>      
>>>
>>Instant Messaging
>>    
>>
>>>(i.e., for specific target applications), the task sounds more
>>>approachable and I feel more comfortable. And I think we 
>>>      
>>>
>>have no choice
>>    
>>
>>>but to specify one or more DHTs for P2P SIP. Otherwise, we 
>>>      
>>>
>>won't be able
>>    
>>
>>>to provide interoperability.
>>>
>>>I also agree that we should start with simple algorithms. 
>>>      
>>>
>>It is a good
>>    
>>
>>>idea to support negotiation about P2P algorithms.
>>>
>>>Regards,
>>>
>>>Eunsoo
>>>
>>>David A. Bryan wrote:
>>>
>>>      
>>>
>>>>I'm fully in agreement here. I think that having multiple 
>>>>        
>>>>
>>algorithms
>>    
>>
>>>>-- and in particular very small simple algorithms as well as more
>>>>complex algorithms is essential. We are unlikely to be 
>>>>        
>>>>
>>able to choose
>>    
>>
>>>>one that will work in all cases, and are certain to not be able to
>>>>choose one that will be the "best" for any length of time 
>>>>        
>>>>
>>-- there are
>>    
>>
>>>>always advances in this area. I think Phillip's analogy to 
>>>>        
>>>>
>>security is
>>    
>>
>>>>a good one here -- devices should be able to negotiate a 
>>>>        
>>>>
>>P2P protocol,
>>    
>>
>>>>just as they can with security, encryption, or codec 
>>>>        
>>>>
>>choice. We should
>>    
>>
>>>>possibly even be able to negotiate non-DHT P2P mechanisms 
>>>>        
>>>>
>>-- perhaps
>>    
>>
>>>>in some very small scale cases we should use flood for simplicity.
>>>>Perhaps also in emergency cases?
>>>>
>>>>--David
>>>>
>>>>        
>>>>
>>_______________________________________________
>>p2p-sip mailing list
>>p2p-sip@cs.columbia.edu
>>http://lists.cs.columbia.edu/mailman/listinfo/p2p-sip
>>
>>    
>>

From RoyR at saic-abingdon.com  Wed Nov  2 09:54:59 2005
From: RoyR at saic-abingdon.com (Roy, Radhika (AEAD))
Date: Wed Nov  2 09:51:39 2005
Subject: [p2p-sip] WG or not (was: dhts)
Message-ID: <2FE23D25B7292B489A217D1965CDA86608991B@bh-exchange02.saic-abingdon.com>

Philip:

 

Yes, I understand. It is all discussions in the air just to form our
ideas as we go forward.

 

BR/Radhika

 

________________________________

From: matthewsp@avaya.com [mailto:matthewsp@avaya.com] 
Sent: Wednesday, November 02, 2005 9:48 AM
To: Roy, Radhika (AEAD)
Cc: p2p-sip@cs.columbia.edu
Subject: RE: [p2p-sip] WG or not (was: dhts)

 

Radhika:

 

I should just point out that everything we are talking about is
preliminary.

There is, as present, no working group nor is there any official
decision to form one.

We are still trying to figure out what the work items of a possible WG
might look like.

After that, we must hold an official BOF and convince the IESG to form a
WG.

None of this is guaranteed to happen.

 

So when you say "Philip has answered my main question that we will
standardize a p2p protocol",

what is really true is "Philip is throwing out a suggestion that this
(unofficial) group should discuss

the idea of standardizing a simple P2P protocol".

 

- Philip

	
________________________________


	From: RoyR@saic-abingdon.com [mailto:RoyR@saic-abingdon.com] 
	Sent: November 1, 2005 16:26
	To: Matthews, Philip (Philip); David A. Bryan; Eunsoo Shim
	Cc: p2p-sip@cs.columbia.edu
	Subject: RE: [p2p-sip] dhts

	Hi, David and Philip:

	It is not that I am disagreeing with you. In fact, Philip has
answered my main question that we will standardize a p2p protocol. Like
others, I also definitely agree with Philip as he has proposed, but in
reverse order:

	1.      Standardize a simple protocol Z as the default P2P
protocol 
	2.      Standardize a negotiation mechanism for the P2P protocol
being used in a group 
	3.      Standardize a framework for hierarchical P2P networks 

	Now, if we start with item 1, we have to see couple of basic
items (something) as follows:

	a.      Discovery (as a part of the p2p protocol) 
	b.      Lookup (as a part of the p2p protocol) 
	c.      Content Transfer between peers directly (will be
triggered or invoked as a result of p2p [signaling] protocol) 
	d.      Use of existing transport protocol(s) depending on the
needs of the p2p protocol 
	e.      Use of existing security protocol(s) depending on the
needs of the p2p protocol 

	If we agree with the above, we can then see where DHT and other
algorithms will stand, and how we should address all aspects of p2p
protocol in a consistent way taking some little baby steps in the
beginning.

	Best regards,

	Radhika

	-----Original Message-----
	From: matthewsp@avaya.com [mailto:matthewsp@avaya.com]
	Sent: Tuesday, November 01, 2005 2:29 PM
	To: David A. Bryan; Roy, Radhika (AEAD); Eunsoo Shim
	Cc: p2p-sip@cs.columbia.edu
	Subject: RE: [p2p-sip] dhts

	I agree with David.

	I think we could standardize a negotiation framework

	(which is really a capabilities advertisement mechanism)

	along with rules for how to select between alternatives.

	Along this line, I am thinking that the hierarchical P2P network

	discussed in a separate thread on this list (and the referenced
paper)

	is looking increasingly attractive.

	It allows different groups of peers to use different DHT
algorithms.

	Say for example, one group of peers (group A) wishes to use
algorithm X,

	while another group (group B) wishes to use algorithm Y, where
both X

	and Y

	are advanced DHT algorithms. If all nodes also speak some basic

	P2P algorithm Z, then we have a hierarchy, with group A nodes
using

	algorithm X

	internally, group B nodes using algorithm Y internally, and the
two

	groups themselves

	being treated as two nodes in a second level, which speaks
algorithm Z.

	Note that algorithm Z does not have to scale to lots of nodes,
since

	at its level there are only two: the node representing group A
and the

	node representing

	group B. 

	How groups A and B form are good questions. Does a node belong
to group

	A

	simply because it speaks protocol X, or does group A has some
other

	natural

	cohesive grouping? For example, do a bunch of nodes speaking
David's

	SoSimple

	P2P protocol find each other through some (external?) mechanism,

	and a bunch of nodes speaking the Nimcat P2P protocol find each
other

	through

	another mechanism, and then the two groups talk using protocol
Z?

	If others agree with me that a hierarchical approach is the way
to go,

	then I think that we might be able to:

	- Standardize a framework for hierarchical P2P networks

	- Standardize a negotiation mechanism for the P2P protocol being
used in

	a group

	- Standardize a simple protocol Z as the default P2P protocol

	I agree with David's comment that the pattern (and format) of
messages

	used

	in one P2P protocol may differ quite a bit from the messages
used in a

	different

	protocol. I am not sure how successful we would be if we tried
to

	standardize

	(for example) the message format, and leave only the algorithm
open.

	 

	- Philip

	 

	> -----Original Message-----

	> From: p2p-sip-bounces@cs.columbia.edu 

	> [mailto:p2p-sip-bounces@cs.columbia.edu] On Behalf Of David A.
Bryan

	> Sent: November 1, 2005 11:59

	> To: Roy, Radhika (AEAD)

	> Cc: p2p-sip@cs.columbia.edu

	> Subject: Re: [p2p-sip] dhts

	> 

	> I think this is the spirit of the thread, but a warning that
in

	> practice getting this together may be quite difficult, since
different

	> types of p2p implementations exchange different information
between

	> nodes for admission and upkeep of the overlay.

	> 

	> I think if we take this approach, the negotiation of a p2p
mechanism

	> and the implementation thereof (how p2p information is
conveyed

	> between nodes) will have to be two independent things, and I
doubt we

	> can standardize how the information is conveyed between nodes
in a way

	> that doesn't take into account what algorithm is being used.
We may

	> need a negotiation mechanism and then different "standards"
for how

	> the nodes maintain the overlay, since the information
contained in

	> messages and even the pattern and reason for messages
exchanged

	> between nodes can vary greatly for different p2p mechanisms.

	> 

	> David

	> 

	> 

	> On 11/1/05, Roy, Radhika (AEAD) <RoyR@saic-abingdon.com>
wrote:

	> > So, the logical conclusion seems to be as follows:

	> >

	> > 1. P2P protocol will be algorithm agnostic

	> >

	> > 2. P2P protocol will allow capabilities to negotiate any 

	> algorithm to be

	> > used among the respective peers

	> >

	> > 3. P2P scheme will standardize couple of algorithms that can
be

	> > negotiated among the peers starting with a simple one first

	> >

	> > It appears that P2P protocol scheme should be flexible
enough to use

	> > algorithms and others per-(each)-peer basis.

	> >

	> > To extend this further, it may also be said for transport
(reliable,

	> > unreliable) and security (transport layer, IP layer) that
the P2P

	> > protocol should also use the same principle.

	> >

	> > Is that the right summary of this thread so far?

	> >

	> > Best regards,

	> > Radhika

	> >

	> > -----Original Message-----

	> > From: p2p-sip-bounces@cs.columbia.edu

	> > [mailto:p2p-sip-bounces@cs.columbia.edu] On Behalf Of Eunsoo
Shim

	> > Sent: Tuesday, November 01, 2005 11:15 AM

	> > To: David A. Bryan

	> > Cc: p2p-sip@cs.columbia.edu

	> > Subject: Re: [p2p-sip] dhts

	> >

	> > Standardizing a DHT sounds somewhat overwhelming to me.

	> > But if we put it as standardizing (or writing an 

	> experimental epec of )

	> > a DHT for SIP-based Internet Telephony, Presence and 

	> Instant Messaging

	> > (i.e., for specific target applications), the task sounds
more

	> > approachable and I feel more comfortable. And I think we 

	> have no choice

	> > but to specify one or more DHTs for P2P SIP. Otherwise, we 

	> won't be able

	> > to provide interoperability.

	> >

	> > I also agree that we should start with simple algorithms. 

	> It is a good

	> > idea to support negotiation about P2P algorithms.

	> >

	> > Regards,

	> >

	> > Eunsoo

	> >

-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://lists.cs.columbia.edu/pipermail/p2p-sip/attachments/20051102/5dc1ca3a/attachment-0001.html
From matthewsp at avaya.com  Wed Nov  2 09:48:16 2005
From: matthewsp at avaya.com (Matthews, Philip (Philip))
Date: Wed Nov  2 09:52:35 2005
Subject: [p2p-sip] WG or not (was: dhts)
Message-ID: <DFD0AADB452F8B4E99FB78719CAF0E6101963664@ON0010AVEXU1.global.avaya.com>

Radhika:
 
I should just point out that everything we are talking about is
preliminary.
There is, as present, no working group nor is there any official
decision to form one.
We are still trying to figure out what the work items of a possible WG
might look like.
After that, we must hold an official BOF and convince the IESG to form a
WG.
None of this is guaranteed to happen.
 
So when you say "Philip has answered my main question that we will
standardize a p2p protocol",
what is really true is "Philip is throwing out a suggestion that this
(unofficial) group should discuss
the idea of standardizing a simple P2P protocol".
 
- Philip


  _____  

	From: RoyR@saic-abingdon.com [mailto:RoyR@saic-abingdon.com] 
	Sent: November 1, 2005 16:26
	To: Matthews, Philip (Philip); David A. Bryan; Eunsoo Shim
	Cc: p2p-sip@cs.columbia.edu
	Subject: RE: [p2p-sip] dhts
	
	

	Hi, David and Philip:

	It is not that I am disagreeing with you. In fact, Philip has
answered my main question that we will standardize a p2p protocol. Like
others, I also definitely agree with Philip as he has proposed, but in
reverse order:

	1.      Standardize a simple protocol Z as the default P2P
protocol 
	2.      Standardize a negotiation mechanism for the P2P protocol
being used in a group 
	3.      Standardize a framework for hierarchical P2P networks 

	Now, if we start with item 1, we have to see couple of basic
items (something) as follows:

	a.      Discovery (as a part of the p2p protocol) 
	b.      Lookup (as a part of the p2p protocol) 
	c.      Content Transfer between peers directly (will be
triggered or invoked as a result of p2p [signaling] protocol) 
	d.      Use of existing transport protocol(s) depending on the
needs of the p2p protocol 
	e.      Use of existing security protocol(s) depending on the
needs of the p2p protocol 

	If we agree with the above, we can then see where DHT and other
algorithms will stand, and how we should address all aspects of p2p
protocol in a consistent way taking some little baby steps in the
beginning.

	Best regards,

	Radhika

	-----Original Message-----
	From: matthewsp@avaya.com [mailto:matthewsp@avaya.com]
	Sent: Tuesday, November 01, 2005 2:29 PM
	To: David A. Bryan; Roy, Radhika (AEAD); Eunsoo Shim
	Cc: p2p-sip@cs.columbia.edu
	Subject: RE: [p2p-sip] dhts

	I agree with David.

	I think we could standardize a negotiation framework

	(which is really a capabilities advertisement mechanism)

	along with rules for how to select between alternatives.

	Along this line, I am thinking that the hierarchical P2P network

	discussed in a separate thread on this list (and the referenced
paper)

	is looking increasingly attractive.

	It allows different groups of peers to use different DHT
algorithms.

	Say for example, one group of peers (group A) wishes to use
algorithm X,

	while another group (group B) wishes to use algorithm Y, where
both X

	and Y

	are advanced DHT algorithms. If all nodes also speak some basic

	P2P algorithm Z, then we have a hierarchy, with group A nodes
using

	algorithm X

	internally, group B nodes using algorithm Y internally, and the
two

	groups themselves

	being treated as two nodes in a second level, which speaks
algorithm Z.

	Note that algorithm Z does not have to scale to lots of nodes,
since

	at its level there are only two: the node representing group A
and the

	node representing

	group B. 

	How groups A and B form are good questions. Does a node belong
to group

	A

	simply because it speaks protocol X, or does group A has some
other

	natural

	cohesive grouping? For example, do a bunch of nodes speaking
David's

	SoSimple

	P2P protocol find each other through some (external?) mechanism,

	and a bunch of nodes speaking the Nimcat P2P protocol find each
other

	through

	another mechanism, and then the two groups talk using protocol
Z?

	If others agree with me that a hierarchical approach is the way
to go,

	then I think that we might be able to:

	- Standardize a framework for hierarchical P2P networks

	- Standardize a negotiation mechanism for the P2P protocol being
used in

	a group

	- Standardize a simple protocol Z as the default P2P protocol

	I agree with David's comment that the pattern (and format) of
messages

	used

	in one P2P protocol may differ quite a bit from the messages
used in a

	different

	protocol. I am not sure how successful we would be if we tried
to

	standardize

	(for example) the message format, and leave only the algorithm
open.


	- Philip



	> -----Original Message-----

	> From: p2p-sip-bounces@cs.columbia.edu 

	> [mailto:p2p-sip-bounces@cs.columbia.edu] On Behalf Of David A.
Bryan

	> Sent: November 1, 2005 11:59

	> To: Roy, Radhika (AEAD)

	> Cc: p2p-sip@cs.columbia.edu

	> Subject: Re: [p2p-sip] dhts

	> 

	> I think this is the spirit of the thread, but a warning that
in

	> practice getting this together may be quite difficult, since
different

	> types of p2p implementations exchange different information
between

	> nodes for admission and upkeep of the overlay.

	> 

	> I think if we take this approach, the negotiation of a p2p
mechanism

	> and the implementation thereof (how p2p information is
conveyed

	> between nodes) will have to be two independent things, and I
doubt we

	> can standardize how the information is conveyed between nodes
in a way

	> that doesn't take into account what algorithm is being used.
We may

	> need a negotiation mechanism and then different "standards"
for how

	> the nodes maintain the overlay, since the information
contained in

	> messages and even the pattern and reason for messages
exchanged

	> between nodes can vary greatly for different p2p mechanisms.

	> 

	> David

	> 

	> 

	> On 11/1/05, Roy, Radhika (AEAD) <RoyR@saic-abingdon.com>
wrote:

	> > So, the logical conclusion seems to be as follows:

	> >

	> > 1. P2P protocol will be algorithm agnostic

	> >

	> > 2. P2P protocol will allow capabilities to negotiate any 

	> algorithm to be

	> > used among the respective peers

	> >

	> > 3. P2P scheme will standardize couple of algorithms that can
be

	> > negotiated among the peers starting with a simple one first

	> >

	> > It appears that P2P protocol scheme should be flexible
enough to use

	> > algorithms and others per-(each)-peer basis.

	> >

	> > To extend this further, it may also be said for transport
(reliable,

	> > unreliable) and security (transport layer, IP layer) that
the P2P

	> > protocol should also use the same principle.

	> >

	> > Is that the right summary of this thread so far?

	> >

	> > Best regards,

	> > Radhika

	> >

	> > -----Original Message-----

	> > From: p2p-sip-bounces@cs.columbia.edu

	> > [mailto:p2p-sip-bounces@cs.columbia.edu] On Behalf Of Eunsoo
Shim

	> > Sent: Tuesday, November 01, 2005 11:15 AM

	> > To: David A. Bryan

	> > Cc: p2p-sip@cs.columbia.edu

	> > Subject: Re: [p2p-sip] dhts

	> >

	> > Standardizing a DHT sounds somewhat overwhelming to me.

	> > But if we put it as standardizing (or writing an 

	> experimental epec of )

	> > a DHT for SIP-based Internet Telephony, Presence and 

	> Instant Messaging

	> > (i.e., for specific target applications), the task sounds
more

	> > approachable and I feel more comfortable. And I think we 

	> have no choice

	> > but to specify one or more DHTs for P2P SIP. Otherwise, we 

	> won't be able

	> > to provide interoperability.

	> >

	> > I also agree that we should start with simple algorithms. 

	> It is a good

	> > idea to support negotiation about P2P algorithms.

	> >

	> > Regards,

	> >

	> > Eunsoo

	> >

-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://lists.cs.columbia.edu/pipermail/p2p-sip/attachments/20051102/476284d6/attachment-0001.html
From RoyR at saic-abingdon.com  Wed Nov  2 10:40:25 2005
From: RoyR at saic-abingdon.com (Roy, Radhika (AEAD))
Date: Wed Nov  2 10:37:07 2005
Subject: [p2p-sip] dhts
Message-ID: <2FE23D25B7292B489A217D1965CDA86608991E@bh-exchange02.saic-abingdon.com>

Hi, Eunsoo:

I am starting with your statement:

 "... What I thought was something like a common message structure as
the base and different payloads for different overlay algorithms. .."

Yes, if we can agree to create a "common" message structure, and  p2p
entities can use those messages for signaling among themselves this is
what I mean a single p2p protocol. This serves the purpose what I have
been saying.

Definitely, there will be different kinds of "payloads" within the said
p2p protocol to meet different requirements such as algorithms,
discovery, lookup, transport, security, triggering of
contents/applications, etc.

It means that the proposed p2p protocol may invoke/trigger or require or
call upon a host of other protocols to meet the specific requirements.

With respect to your questions related to
domains/inter-domain/node/supper-nodes, I believe, these are
architectural problems, and is NOT so much about protocol problems. I
guess that it could also be taken care of once the base p2p protocol is
developed. 

Of course, if anything needs to be taken care of early on that will
influence the development of the p2p protocol in view of the
/inter-domain/node/supper-nodes, definitely, we have to keep a provision
for those especially for extensibility.

Sure, like David and Philip, you have also made some good points.

Best regards,
Radhika

-----Original Message-----
From: eunsoo@research.panasonic.com
[mailto:eunsoo@research.panasonic.com] 
Sent: Wednesday, November 02, 2005 5:37 AM
To: Matthews, Philip (Philip)
Cc: David A. Bryan; Roy, Radhika (AEAD); p2p-sip@cs.columbia.edu
Subject: Re: [p2p-sip] dhts

Hmm.... probably another terminology issue.
I think 'hierarchical architecture' can have two meanings in P2P SIP.
One is something like supernodes and non-supernodes where supernodes are
not necessarily related to domain boundaries. In that sense, you can
have a hierarchical P2P network even within a (administrative) domain.
Another is something like federation of domains with inter-domain
networking.
That is, 'hierarchical architecture' does not necessarily mean
multi-domain or federation of domains.
I guess Philip is using the term as federation of domains.

I think an architecture supporting federation of domains has significant
merit in that it provides scalability and flexibility in configuring
individual domains and we need to support it eventually. The idea of
algorithm Z for inter-domain operation and algorithm X and Y for
intra-domain operation seems quite natural. It reminds me of
inter-domain routing protocols and intra-domain routing protocols in IP
routing. However, we can see that an algorithm for inter-domain
operation is not necessarily the simplest or default algorithm in IP
routing. This may not be necessarily the case for P2P SIP but I think it
provides enough motivation for us to be cautious. And most likely
dealing with federation of domains will add more complexity to the
architecture and algorithm even though I don't know how much.

I think we agree that we should start with simple algorithms. So I'd say
we look for simple and reasonably good overlay algorithms in general
rather than specifically something for inter-domain operation in the
beginning. That is, I'd suggest we should first look for simple and
reasonably good algorithm X and Y for individual domains.


Regarding one protocol and multiple algorithms, I also agree that
different algorithms are likely to have differences in information in
messages and even in message flows. What I thought was something like a
common message structure as the base and different payloads for
different overlay algorithms. It is debatable whether such a thing is
one protocol or multiple protocols as David pointed. But that approach
looks better to me than creating totally new and different message
structures for each overlay algorithm. It seems feasible at least for
DHTs. What do you think?

Regards,

Eunsoo


-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://lists.cs.columbia.edu/pipermail/p2p-sip/attachments/20051102/05623dd5/attachment.html
From dhara at research.avayalabs.com  Wed Nov  2 11:44:13 2005
From: dhara at research.avayalabs.com (K. Kishore Dhara)
Date: Wed Nov  2 11:50:46 2005
Subject: [p2p-sip] dhts
In-Reply-To: <2FE23D25B7292B489A217D1965CDA86608991E@bh-exchange02.saic-abingdon.com>
References: <2FE23D25B7292B489A217D1965CDA86608991E@bh-exchange02.saic-abingdon.com>
Message-ID: <4368ECDD.6090007@research.avayalabs.com>

Radhika,


Roy, Radhika (AEAD) wrote:

> Yes, if we can agree to create a "common" message structure, and  p2p 
> entities can use those messages for signaling among themselves this is 
> what I mean a single p2p protocol. This serves the purpose what I have 
> been saying.
>

I agree that a common message structure is important, but a) contents of 
the structure and b) the semantics of the structure are key issues as 
well.  Does the common P2P protocol  contain information that is 
required for the protocols (and algorithms) used by various P2P systems 
(say X and Y)? If it does, then should it be intrepreted by X and Y in 
the same way?  I see one  way of trying to do that is in addition to the 
common structure, we should try to achieve an abstraction that can 
contain information needed by various algorithms and is flexible enough 
to let me them interpret in their own way (or may be even have some sort 
of lose semantic framework for interpretation, not sure though).  Is 
this achievable?  Well, I think first we should flush out the requirements.

Regards,
-- Kishore

From RoyR at saic-abingdon.com  Wed Nov  2 12:01:02 2005
From: RoyR at saic-abingdon.com (Roy, Radhika (AEAD))
Date: Wed Nov  2 11:59:43 2005
Subject: [p2p-sip] dhts
Message-ID: <2FE23D25B7292B489A217D1965CDA866089920@bh-exchange02.saic-abingdon.com>



-----Original Message-----
From: dhara@research.avayalabs.com [mailto:dhara@research.avayalabs.com]

Hi, Kishore:

Yes, as you have indicated, the main idea of this thread has been to
flush out the requirements for p2p protocol framework. This discussion
is taking a shape to go there as we are trying to converge after getting
all inputs from all sides.

The key is this: When we write our requirements for p2p protocol
framework, we know exactly where we are going. Rest is only a matter of
time to implement those requirements.

A little more steps in discussions will help us to crystallize our
thoughts in this direction.

Best regards,
Radhika

Sent: Wednesday, November 02, 2005 11:44 AM
To: Roy, Radhika (AEAD)
Cc: p2p-sip@cs.columbia.edu
Subject: Re: [p2p-sip] dhts

Radhika,


Roy, Radhika (AEAD) wrote:

> Yes, if we can agree to create a "common" message structure, and  p2p 
> entities can use those messages for signaling among themselves this is

> what I mean a single p2p protocol. This serves the purpose what I have

> been saying.
>

I agree that a common message structure is important, but a) contents of

the structure and b) the semantics of the structure are key issues as 
well.  Does the common P2P protocol  contain information that is 
required for the protocols (and algorithms) used by various P2P systems 
(say X and Y)? If it does, then should it be interpreted by X and Y in 
the same way?  I see one  way of trying to do that is in addition to the

common structure, we should try to achieve an abstraction that can 
contain information needed by various algorithms and is flexible enough 
to let me them interpret in their own way (or may be even have some sort

of lose semantic framework for interpretation, not sure though).  Is 
this achievable?  Well, I think first we should flush out the
requirements.

Regards,
-- Kishore

-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://lists.cs.columbia.edu/pipermail/p2p-sip/attachments/20051102/65e804f6/attachment-0001.html
From threelions0916 at yahoo.com.cn  Wed Nov  2 22:39:57 2005
From: threelions0916 at yahoo.com.cn (threelions)
Date: Wed Nov  2 22:40:51 2005
Subject: [p2p-sip] Draft: Requirements for SIP-based
	Peer-to-PeerInternetTelephony
References: <DFD0AADB452F8B4E99FB78719CAF0E6101962B94@ON0010AVEXU1.global.avaya.com>
Message-ID: <003d01c5e028$4b037a80$2518080a@cnc.intra>

I noticed that section 7.1 mentioned "A global P2P telephony VoIP system which allows its users located in   different regions and countries to communicate with each other." ,but there was no suggestion on the infrastructure of gloable p2p voip system, In my opnion, a hierarchical p2p system must be built to achieve the goal of high scalability, for there may be millions of nodes in a global p2p voip system, and the heterogeneity of variant nodes determined the system must be hierarchical.

In a hierarchical system, it's very natural that NAT and Firewall Traversal could be distributed among the super nodes,because the super nodes can provide the more bandwidthã€�more cpu power and longer lifetime.

some other comments listed below.

regards
Michael Liu


----- Original Message ----- 
From: "Matthews, Philip (Philip)" <matthewsp@avaya.com>
To: "Salman Abdul Baset" <salman@cs.columbia.edu>; <hgs@cs.columbia.edu>; <eunsoo@research.panasonic.com>; <dhara@avaya.com>
Cc: <p2p-sip@cs.columbia.edu>
Sent: Saturday, October 29, 2005 6:35 AM
Subject: RE: [p2p-sip] Draft: Requirements for SIP-based Peer-to-PeerInternetTelephony


> [I am posting this to the P2P-SIP mailing list only (and not SIPPING),
> since it is my belief that that list provides the most appropriate
> discussion forum.]
> 
> I think this is a great start on the requirements discussion.
> My thanks to the authors for producing this.
> 
> Here are some quick comments. Hope to generate some more in-depth
> comments next week.
> 
> > 4.3.  DHT Changeability
> >    The protocol(s) for maintaining the peer-to-peer network SHOULD be
> >    flexible to accommodate different DHT algorithms.
> 
> I totally agree with this. From my limited investigation of the DHT
> world,
> I see things just changing too fast to have any hope that today's DHT
> will
> still be tomorrow's DHT of choice.
> 

I agree with that.it's very important. 

> 
> > 4.5.  NAT and Firewall Traversal
> >    R1.  The peer-to-peer system SHOULD distribute the functionality of
> >    NAT and firewall traversal servers to the end-points.
> 
> I don't see why this must be a requirement. Your other requirements talk
> about being flexible on where services are located (i.e. centralized or
> distributed). Given this flexibility, which I agree with, why require
> that
> a certain service be distributed? Why not leave the choice up to the
> builder of
> the particular P2P overlay network? If that builder wants to provide a
> centralized NAT and Firewall Traversal scheme, why not let him?
> 
> As I see it, a centralized service is one that is located on just a
> small number
> of peers, while a distributed service is one that is located on a large
> number
> of peers. The same lookup scheme could be used for both, so an arbitrary
> peer doesn't
> really need to know whether a service is centralized or distributed. So
> why say that
> a certain service must be distributed?
> 
> On the other hand, I would agree that the design should SUPPORT making
> the NAT and
> Firewall Traversal distributable, but I think you have already covered
> that in 4.1.
> 

I think in the global p2p voip network, we must distribute the functionality of  NAT and firewall traversal servers to the super node, and in a local network it's prefered to distribute the functionality of NAT and firewall travesal server to the end nodes, but a centralized server can works well too.
as for the Emergency /ad-hoc scenario, functionality of NAT and firewall travesal must be distribute to the end nodes.

> 
> > 5.4.1.  Centralized Lookup
> >    Value-added services such as reliable voicemail storage can be
> >    provided by a central authority in a P2P network.  The system
> SHOULD
> >    allow the peers to efficiently locate the centralized service
> >    providers or resources.
> 
> This is just a minor comment. The first sentence seems to imply that
> voicemail
> is preferably implemented as a centralized service. I disagree with that
> implication, and suggest that the first sentence be reworded or simply
> removed.
> 
> I DO agree with the requirement in the second sentence.
> 
> 
> > 6.1.  Support for different DHTs
> >    Any protocol specification SHOULD be able to incorporate different
> >    DHT algorithms based on the P2P service provider requirements.
> 
> Isn't this the same as requirement 4.3??
> 
> 
> - Philip Matthews
>   Nimcat Networks (now a part of Avaya)
> 
> PS. Please note that my e-mail address has recently changed to
>        matthewsp@avaya.com
>     Mail sent to my old address (matthews@nimcatnetworks.com) will
>     be forwarded for a period, but will eventually be returned.
> 
> 
> 
> -----Original Message-----
> From: p2p-sip-bounces@cs.columbia.edu
> [mailto:p2p-sip-bounces@cs.columbia.edu] On Behalf Of Salman Abdul Baset
> Sent: October 28, 2005 13:16
> To: sipping@ietf.org
> Cc: p2p-sip@cs.columbia.edu
> Subject: [p2p-sip] Draft: Requirements for SIP-based Peer-to-Peer
> InternetTelephony
> 
> We have attempted to compile the requirements for SIP-based peer-to-peer
> Internet telephony in a draft. It is available at:
> http://www1.cs.columbia.edu/~salman/drafts/draft-baset-sipping-p2preq-00
> .txt
> 
> Title: Requirements for SIP-based Peer-to-Peer Internet Telephony
> Abstract:
> This document defines requirements for designing peer-to-peer voice,
> text, and real-time multimedia communication system protocols.
> 
> Comments are requested.
> 
> Regards,
> Salman, Henning, Eunsoo, Kishore
> _______________________________________________
> p2p-sip mailing list
> p2p-sip@cs.columbia.edu
> http://lists.cs.columbia.edu/mailman/listinfo/p2p-sip
> 
> 
> _______________________________________________
> p2p-sip mailing list
> p2p-sip@cs.columbia.edu
> http://lists.cs.columbia.edu/mailman/listinfo/p2p-sip
ÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿðèb‹˜j(ÿÿÿÿÿÿÿñÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿþm§ÿÿræj)ÿÉ¨h£÷(›ÿâ

From srisuresh at yahoo.com  Thu Nov  3 09:41:09 2005
From: srisuresh at yahoo.com (Pyda Srisuresh)
Date: Thu Nov  3 09:41:30 2005
Subject: P2P-SIP NAT traversal (was Re: [p2p-sip] Draft: Requirements for
	SIP-based Peer-to-PeerInternetTelephony)
In-Reply-To: <003d01c5e028$4b037a80$2518080a@cnc.intra>
Message-ID: <20051103144110.21552.qmail@web33308.mail.mud.yahoo.com>

Folks,

I agree with Michael that NAT traversal is a key requirement of P2P
telephony/voip. Prior to initiating data (media) transfer, end nodes must be
able to a) learn peer's endpoint identities, and b) determine the NAT traversal
technique that is appropriate for the pair of nodes.

Of these, item a) is a direct part of P2P SIP and item b) is not. Item b) is a
P2P NAT traversal inependent of the protocol used for discovering the peer and
its endpoints. However, I believe, both these items (P2P-SIP and P2P-NAT) ought
to be part of the proposed new WG charter (P2P VOIP ?) charter. 

Has anyone used ICE and/or TURN for NAT traversal in the context of P2P
telephony? If so, what sort of delays(i.e., latencies) were noticed in setting
up the data session? I ask this because P2P telephony is very sensitive to
latencies.

cheers,
suresh  

--- threelions <threelions0916@yahoo.com.cn> wrote:

> I noticed that section 7.1 mentioned "A global P2P telephony VoIP system
> which allows its users located in   different regions and countries to
> communicate with each other." ,but there was no suggestion on the
> infrastructure of gloable p2p voip system, In my opnion, a hierarchical p2p
> system must be built to achieve the goal of high scalability, for there may
> be millions of nodes in a global p2p voip system, and the heterogeneity of
> variant nodes determined the system must be hierarchical.
> 
> In a hierarchical system, it's very natural that NAT and Firewall Traversal
> could be distributed among the super nodes,because the super nodes can
> provide the more bandwidth???more cpu power and longer lifetime.
> 
> some other comments listed below.
> 
> regards
> Michael Liu
> 
> 
> ----- Original Message ----- 
> From: "Matthews, Philip (Philip)" <matthewsp@avaya.com>
> To: "Salman Abdul Baset" <salman@cs.columbia.edu>; <hgs@cs.columbia.edu>;
> <eunsoo@research.panasonic.com>; <dhara@avaya.com>
> Cc: <p2p-sip@cs.columbia.edu>
> Sent: Saturday, October 29, 2005 6:35 AM
> Subject: RE: [p2p-sip] Draft: Requirements for SIP-based
> Peer-to-PeerInternetTelephony
> 
> 
> > [I am posting this to the P2P-SIP mailing list only (and not SIPPING),
> > since it is my belief that that list provides the most appropriate
> > discussion forum.]
> > 
> > I think this is a great start on the requirements discussion.
> > My thanks to the authors for producing this.
> > 
> > Here are some quick comments. Hope to generate some more in-depth
> > comments next week.
> > 
> > > 4.3.  DHT Changeability
> > >    The protocol(s) for maintaining the peer-to-peer network SHOULD be
> > >    flexible to accommodate different DHT algorithms.
> > 
> > I totally agree with this. From my limited investigation of the DHT
> > world,
> > I see things just changing too fast to have any hope that today's DHT
> > will
> > still be tomorrow's DHT of choice.
> > 
> 
> I agree with that.it's very important. 
> 
> > 
> > > 4.5.  NAT and Firewall Traversal
> > >    R1.  The peer-to-peer system SHOULD distribute the functionality of
> > >    NAT and firewall traversal servers to the end-points.
> > 
> > I don't see why this must be a requirement. Your other requirements talk
> > about being flexible on where services are located (i.e. centralized or
> > distributed). Given this flexibility, which I agree with, why require
> > that
> > a certain service be distributed? Why not leave the choice up to the
> > builder of
> > the particular P2P overlay network? If that builder wants to provide a
> > centralized NAT and Firewall Traversal scheme, why not let him?
> > 
> > As I see it, a centralized service is one that is located on just a
> > small number
> > of peers, while a distributed service is one that is located on a large
> > number
> > of peers. The same lookup scheme could be used for both, so an arbitrary
> > peer doesn't
> > really need to know whether a service is centralized or distributed. So
> > why say that
> > a certain service must be distributed?
> > 
> > On the other hand, I would agree that the design should SUPPORT making
> > the NAT and
> > Firewall Traversal distributable, but I think you have already covered
> > that in 4.1.
> > 
> 
> I think in the global p2p voip network, we must distribute the functionality
> of  NAT and firewall traversal servers to the super node, and in a local
> network it's prefered to distribute the functionality of NAT and firewall
> travesal server to the end nodes, but a centralized server can works well
> too.
> as for the Emergency /ad-hoc scenario, functionality of NAT and firewall
> travesal must be distribute to the end nodes.
> 
> > 
> > > 5.4.1.  Centralized Lookup
> > >    Value-added services such as reliable voicemail storage can be
> > >    provided by a central authority in a P2P network.  The system
> > SHOULD
> > >    allow the peers to efficiently locate the centralized service
> > >    providers or resources.
> > 
> > This is just a minor comment. The first sentence seems to imply that
> > voicemail
> > is preferably implemented as a centralized service. I disagree with that
> > implication, and suggest that the first sentence be reworded or simply
> > removed.
> > 
> > I DO agree with the requirement in the second sentence.
> > 
> > 
> > > 6.1.  Support for different DHTs
> > >    Any protocol specification SHOULD be able to incorporate different
> > >    DHT algorithms based on the P2P service provider requirements.
> > 
> > Isn't this the same as requirement 4.3??
> > 
> > 
> > - Philip Matthews
> >   Nimcat Networks (now a part of Avaya)
> > 
> > PS. Please note that my e-mail address has recently changed to
> >        matthewsp@avaya.com
> >     Mail sent to my old address (matthews@nimcatnetworks.com) will
> >     be forwarded for a period, but will eventually be returned.
> > 
> > 
> > 
> > -----Original Message-----
> > From: p2p-sip-bounces@cs.columbia.edu
> > [mailto:p2p-sip-bounces@cs.columbia.edu] On Behalf Of Salman Abdul Baset
> > Sent: October 28, 2005 13:16
> > To: sipping@ietf.org
> > Cc: p2p-sip@cs.columbia.edu
> > Subject: [p2p-sip] Draft: Requirements for SIP-based Peer-to-Peer
> > InternetTelephony
> > 
> > We have attempted to compile the requirements for SIP-based peer-to-peer
> > Internet telephony in a draft. It is available at:
> > http://www1.cs.columbia.edu/~salman/drafts/draft-baset-sipping-p2preq-00
> > .txt
> > 
> > Title: Requirements for SIP-based Peer-to-Peer Internet Telephony
> > Abstract:
> > This document defines requirements for designing peer-to-peer voice,
> > text, and real-time multimedia communication system protocols.
> > 
> > Comments are requested.
> > 
> > Regards,
> > Salman, Henning, Eunsoo, Kishore
> > _______________________________________________
> > p2p-sip mailing list
> > p2p-sip@cs.columbia.edu
> > http://lists.cs.columbia.edu/mailman/listinfo/p2p-sip
> > 
> > 
> > _______________________________________________
> > p2p-sip mailing list
> > p2p-sip@cs.columbia.edu
> > http://lists.cs.columbia.edu/mailman/listinfo/p2p-sip
>
???????????????????????????????????????b??j(?????????????????????????????????????m???r?j)???h??(???
> 
> _______________________________________________
> p2p-sip mailing list
> p2p-sip@cs.columbia.edu
> http://lists.cs.columbia.edu/mailman/listinfo/p2p-sip
> 



From afisk at speedymail.org  Thu Nov  3 11:38:31 2005
From: afisk at speedymail.org (Adam Fisk)
Date: Thu Nov  3 11:39:48 2005
Subject: P2P-SIP NAT traversal (was Re: [p2p-sip] Draft: Requirements
	for	SIP-based Peer-to-PeerInternetTelephony)
In-Reply-To: <20051103144110.21552.qmail@web33308.mail.mud.yahoo.com>
References: <20051103144110.21552.qmail@web33308.mail.mud.yahoo.com>
Message-ID: <436A3D07.6050302@speedymail.org>

I've used ICE and TURN for p2p SIP, although not specifically for 
telephony. This system is not yet widely deployed, though, so our 
architecture of basically just using available non-firewalled peers as 
SIP proxies and as TURN servers has not been tested as thoroughly as 
would be nice. I don't see the latencies as being particularly crucial, 
though, simply because you're not dealing with much data in the call setup.

The TURN case for actual media transfer is the tough one. One option 
would be to incorporate locality optimizations such as those that the 
"closestnode" folks implement at:

http://www.closestnode.com/

As far as I understand it, this system works specifically in a p2p 
context. It might be interesting to leave room for enhancements that 
would allow nodes to choose nodes closest to them on the network using a 
system like this. This could easily result in media sessions that 
perform better than typical SIP scenarios because you can use more 
optimized network routes, albeit more bandwidth-limited ones in most cases.

Of course, the TURN for media transfer case should be avoided at all 
costs, but it's necessary as a fall back. That's really the only case 
where I see the network latencies being a serious concern, unless a pair 
of nodes get unlucky and are using sluggish proxies from opposite sides 
of the globe. Might be worth taking some preventative steps against this.

Along these lines, though, clearly each node should implement STUN on 
the both the client and the server side. They should probably implement 
STUN for TCP (STUNT) as well, although that's not standardized. It is 
really necessary for TCP nat traversal, though, so should be implemented.

I would imagine that all non-firewalled nodes would act as SIP proxies, 
TURN, and STUN servers, but you guys have probably already discussed 
that. I think UPnP is also pretty crucial here to make as many nodes as 
possible available as "non-firewalled". In fact, I think there are few 
enough non-firewalled nodes out there that it might be necessary to have 
UPnP support at something like SHOULD strength for a system like this to 
work in the real world. I have no experience authoring RFCs, though, so 
I'll let others decide issues like that.

This is getting lengthy, but I'd also like to make a pitch to stay true 
to SIPs roots as a generalized system that's not specific to telephony. 
Implementing the TCP side of life is more important in this context, for 
example.

-Adam


Pyda Srisuresh wrote:

>Folks,
>
>I agree with Michael that NAT traversal is a key requirement of P2P
>telephony/voip. Prior to initiating data (media) transfer, end nodes must be
>able to a) learn peer's endpoint identities, and b) determine the NAT traversal
>technique that is appropriate for the pair of nodes.
>
>Of these, item a) is a direct part of P2P SIP and item b) is not. Item b) is a
>P2P NAT traversal inependent of the protocol used for discovering the peer and
>its endpoints. However, I believe, both these items (P2P-SIP and P2P-NAT) ought
>to be part of the proposed new WG charter (P2P VOIP ?) charter. 
>
>Has anyone used ICE and/or TURN for NAT traversal in the context of P2P
>telephony? If so, what sort of delays(i.e., latencies) were noticed in setting
>up the data session? I ask this because P2P telephony is very sensitive to
>latencies.
>
>cheers,
>suresh  
>
>--- threelions <threelions0916@yahoo.com.cn> wrote:
>
>  
>
>>I noticed that section 7.1 mentioned "A global P2P telephony VoIP system
>>which allows its users located in   different regions and countries to
>>communicate with each other." ,but there was no suggestion on the
>>infrastructure of gloable p2p voip system, In my opnion, a hierarchical p2p
>>system must be built to achieve the goal of high scalability, for there may
>>be millions of nodes in a global p2p voip system, and the heterogeneity of
>>variant nodes determined the system must be hierarchical.
>>
>>In a hierarchical system, it's very natural that NAT and Firewall Traversal
>>could be distributed among the super nodes,because the super nodes can
>>provide the more bandwidthã€�more cpu power and longer lifetime.
>>
>>some other comments listed below.
>>
>>regards
>>Michael Liu
>>
>>
>>----- Original Message ----- 
>>From: "Matthews, Philip (Philip)" <matthewsp@avaya.com>
>>To: "Salman Abdul Baset" <salman@cs.columbia.edu>; <hgs@cs.columbia.edu>;
>><eunsoo@research.panasonic.com>; <dhara@avaya.com>
>>Cc: <p2p-sip@cs.columbia.edu>
>>Sent: Saturday, October 29, 2005 6:35 AM
>>Subject: RE: [p2p-sip] Draft: Requirements for SIP-based
>>Peer-to-PeerInternetTelephony
>>
>>
>>    
>>
>>>[I am posting this to the P2P-SIP mailing list only (and not SIPPING),
>>>since it is my belief that that list provides the most appropriate
>>>discussion forum.]
>>>
>>>I think this is a great start on the requirements discussion.
>>>My thanks to the authors for producing this.
>>>
>>>Here are some quick comments. Hope to generate some more in-depth
>>>comments next week.
>>>
>>>      
>>>
>>>>4.3.  DHT Changeability
>>>>   The protocol(s) for maintaining the peer-to-peer network SHOULD be
>>>>   flexible to accommodate different DHT algorithms.
>>>>        
>>>>
>>>I totally agree with this. From my limited investigation of the DHT
>>>world,
>>>I see things just changing too fast to have any hope that today's DHT
>>>will
>>>still be tomorrow's DHT of choice.
>>>
>>>      
>>>
>>I agree with that.it's very important. 
>>
>>    
>>
>>>>4.5.  NAT and Firewall Traversal
>>>>   R1.  The peer-to-peer system SHOULD distribute the functionality of
>>>>   NAT and firewall traversal servers to the end-points.
>>>>        
>>>>
>>>I don't see why this must be a requirement. Your other requirements talk
>>>about being flexible on where services are located (i.e. centralized or
>>>distributed). Given this flexibility, which I agree with, why require
>>>that
>>>a certain service be distributed? Why not leave the choice up to the
>>>builder of
>>>the particular P2P overlay network? If that builder wants to provide a
>>>centralized NAT and Firewall Traversal scheme, why not let him?
>>>
>>>As I see it, a centralized service is one that is located on just a
>>>small number
>>>of peers, while a distributed service is one that is located on a large
>>>number
>>>of peers. The same lookup scheme could be used for both, so an arbitrary
>>>peer doesn't
>>>really need to know whether a service is centralized or distributed. So
>>>why say that
>>>a certain service must be distributed?
>>>
>>>On the other hand, I would agree that the design should SUPPORT making
>>>the NAT and
>>>Firewall Traversal distributable, but I think you have already covered
>>>that in 4.1.
>>>
>>>      
>>>
>>I think in the global p2p voip network, we must distribute the functionality
>>of  NAT and firewall traversal servers to the super node, and in a local
>>network it's prefered to distribute the functionality of NAT and firewall
>>travesal server to the end nodes, but a centralized server can works well
>>too.
>>as for the Emergency /ad-hoc scenario, functionality of NAT and firewall
>>travesal must be distribute to the end nodes.
>>
>>    
>>
>>>>5.4.1.  Centralized Lookup
>>>>   Value-added services such as reliable voicemail storage can be
>>>>   provided by a central authority in a P2P network.  The system
>>>>        
>>>>
>>>SHOULD
>>>      
>>>
>>>>   allow the peers to efficiently locate the centralized service
>>>>   providers or resources.
>>>>        
>>>>
>>>This is just a minor comment. The first sentence seems to imply that
>>>voicemail
>>>is preferably implemented as a centralized service. I disagree with that
>>>implication, and suggest that the first sentence be reworded or simply
>>>removed.
>>>
>>>I DO agree with the requirement in the second sentence.
>>>
>>>
>>>      
>>>
>>>>6.1.  Support for different DHTs
>>>>   Any protocol specification SHOULD be able to incorporate different
>>>>   DHT algorithms based on the P2P service provider requirements.
>>>>        
>>>>
>>>Isn't this the same as requirement 4.3??
>>>
>>>
>>>- Philip Matthews
>>>  Nimcat Networks (now a part of Avaya)
>>>
>>>PS. Please note that my e-mail address has recently changed to
>>>       matthewsp@avaya.com
>>>    Mail sent to my old address (matthews@nimcatnetworks.com) will
>>>    be forwarded for a period, but will eventually be returned.
>>>
>>>
>>>
>>>-----Original Message-----
>>>From: p2p-sip-bounces@cs.columbia.edu
>>>[mailto:p2p-sip-bounces@cs.columbia.edu] On Behalf Of Salman Abdul Baset
>>>Sent: October 28, 2005 13:16
>>>To: sipping@ietf.org
>>>Cc: p2p-sip@cs.columbia.edu
>>>Subject: [p2p-sip] Draft: Requirements for SIP-based Peer-to-Peer
>>>InternetTelephony
>>>
>>>We have attempted to compile the requirements for SIP-based peer-to-peer
>>>Internet telephony in a draft. It is available at:
>>>http://www1.cs.columbia.edu/~salman/drafts/draft-baset-sipping-p2preq-00
>>>.txt
>>>
>>>Title: Requirements for SIP-based Peer-to-Peer Internet Telephony
>>>Abstract:
>>>This document defines requirements for designing peer-to-peer voice,
>>>text, and real-time multimedia communication system protocols.
>>>
>>>Comments are requested.
>>>
>>>Regards,
>>>Salman, Henning, Eunsoo, Kishore
>>>_______________________________________________
>>>p2p-sip mailing list
>>>p2p-sip@cs.columbia.edu
>>>http://lists.cs.columbia.edu/mailman/listinfo/p2p-sip
>>>
>>>
>>>_______________________________________________
>>>p2p-sip mailing list
>>>p2p-sip@cs.columbia.edu
>>>http://lists.cs.columbia.edu/mailman/listinfo/p2p-sip
>>>      
>>>
>ÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿðèb‹˜j(ÿÿÿÿÿÿÿñÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿþm§ÿÿræj)ÿÉ¨h£÷(›ÿâ
>  
>
>>_______________________________________________
>>p2p-sip mailing list
>>p2p-sip@cs.columbia.edu
>>http://lists.cs.columbia.edu/mailman/listinfo/p2p-sip
>>
>>    
>>
>
>
>
>_______________________________________________
>p2p-sip mailing list
>p2p-sip@cs.columbia.edu
>http://lists.cs.columbia.edu/mailman/listinfo/p2p-sip
>
>  
>
From buford at research.panasonic.com  Wed Nov  9 19:42:18 2005
From: buford at research.panasonic.com (buford@research.panasonic.com)
Date: Wed Nov  9 19:42:38 2005
Subject: [p2p-sip] slides for P2P-SIP Ad Hoc
Message-ID: <2983629e72.29e7229836@pintlmail.MITL.Research.Panasonic.COM>


Here are the slides on P2P overlays for the ad hoc session on Friday

http://www.cs.uml.edu/~buford/irtf-p2prg/p2psip/JBuford-IETF-P2PSIP-Overlay-Systems-v3.pdf


John


From afisk at speedymail.org  Thu Nov 10 14:14:06 2005
From: afisk at speedymail.org (Adam Fisk)
Date: Thu Nov 10 14:15:22 2005
Subject: [p2p-sip] slides for P2P-SIP Ad Hoc
In-Reply-To: <2983629e72.29e7229836@pintlmail.MITL.Research.Panasonic.COM>
References: <2983629e72.29e7229836@pintlmail.MITL.Research.Panasonic.COM>
Message-ID: <43739BFE.5050107@speedymail.org>

Hi John-

The slides look great.  I just wanted to point out that both Gnutella 
and Kazaa's search algorithms are more accurately described as random 
walks than as flooding.  Kazaa has used a massive indexing and random 
walk approach for many years, and Gnutella clients have used a similar 
approach for at about the last 3 years.   Gnutella basically uses bloom 
filters to index all data within 2 hops, and then dynamically queries 
supernodes depending on how many results are coming back over time.  In 
many cases this leads to queries not going beyond the nearest 
supernode.  Kazaa uses something similar, although no one outside of 
their organization knows the details. 

"Random walk" does not quite capture it either, but it's closer than 
"flooding".  Thes algorithms are actually *very* efficient, producing 
light load and broad network reach, although they cannot scale as well 
as DHTs, clearly.

That's not particularly relevant to the purposes of this group, but more 
for the sake of everyone's general knowledge.

On the DHT side, it may be worth talking to Bryan Turner from the p2p 
hackers group, as he's by far the most knowledgeable person on the 
various DHT implementations I've come across.  Are you on this list, 
Bryan?  He's implemented many of them, and he's a strong advocate of 
"The Continuous Discrete Approach".  The DHT world evolves very quickly, 
and things like Chord and CAN are certainly not the state of the art.

I'm ccing Bryan on this.  I don't know you, Bryan, but your posts on p2p 
hackers make your depth of knowledge in this area clear.

-Adam


buford@research.panasonic.com wrote:

>Here are the slides on P2P overlays for the ad hoc session on Friday
>
>http://www.cs.uml.edu/~buford/irtf-p2prg/p2psip/JBuford-IETF-P2PSIP-Overlay-Systems-v3.pdf
>
>
>John
>
>
>_______________________________________________
>p2p-sip mailing list
>p2p-sip@cs.columbia.edu
>http://lists.cs.columbia.edu/mailman/listinfo/p2p-sip
>
>  
>
From buford at research.panasonic.com  Thu Nov 10 15:50:22 2005
From: buford at research.panasonic.com (John Buford)
Date: Thu Nov 10 15:50:56 2005
Subject: [p2p-sip] slides for P2P-SIP Ad Hoc
In-Reply-To: <43739BFE.5050107@speedymail.org>
Message-ID: <MBEDJFICFCDPPFADFCIHOECECDAA.buford@research.panasonic.com>

Hi Adam,
Thanks for the comments.

Looking at the wiki for the Gnutella Developers Forum in the
section "Querying the Network"

"A servent SHOULD forward incoming Query messages to all of its directly
connected servents, except the one that delivered the incoming Query.
Servents using Flow control or Ultrapeers (sections 3.1 and 3.2) will not
always forward every Query over every connection. "  There is also a TTL on
the packet which is decremented.

This sounds like flooding not random walk.

According to the GDF wiki, during bootstrapping, if a node can't
connect using entries in the host cache,
it can randomly select a GWebCache to populate its host cache.

Regarding Kazaa, the protocol is proprietary.  One research group analyzed
pollution in Kazaa (Ross et al. "Pollution in P2P File Sharing Systems")
and built a crawler to crawl the Kazaa network.
>From their paper:
"When a user wants to find files, the user?s ON sends
a query with keywords over the TCP connection to its
SN. For each match in its local index, the SN returns the
metadata and IP addresses corresponding to the match.
When a SN receives a query, it may flood the query
over the overlay network to one or more of the SNs
to which it is connected. A given query will in general
visit a small subset of the SNs, and hence will obtain
the metadata information of a small subset of all the ONs."

More recently this same research group has reported on a detailed
analysis of the Kazaa overlay in the paper
J. Liang, R. Kumar, and K.W. Ross, The Kazaa Overlay: A Measurement Study,
Computer Networks (Special Issue on Overlays), to appear.

In this paper they describe, among other things, "connection shuffling"
which both SNs and ONs use.  It appears that this connection shuffling
allows peers to reach other parts of the overlay over time, and may be
related to work load re-distribution.

So it appears that within a given set of SN-SN and SN-ON connections
Kazaa uses flooding, but these connections change over time.

Yes, CAN and Chord are not "state of the art" depending on what you
are looking for in an overlay.

Again, thanks for the comments.

John

-----Original Message-----
From: p2p-sip-bounces@cs.columbia.edu
[mailto:p2p-sip-bounces@cs.columbia.edu]On Behalf Of Adam Fisk
Sent: Thursday, November 10, 2005 2:14 PM
To: p2p-sip@cs.columbia.edu; bryan.turner@pobox.com
Subject: Re: [p2p-sip] slides for P2P-SIP Ad Hoc


Hi John-

The slides look great.  I just wanted to point out that both Gnutella
and Kazaa's search algorithms are more accurately described as random
walks than as flooding.  Kazaa has used a massive indexing and random
walk approach for many years, and Gnutella clients have used a similar
approach for at about the last 3 years.   Gnutella basically uses bloom
filters to index all data within 2 hops, and then dynamically queries
supernodes depending on how many results are coming back over time.  In
many cases this leads to queries not going beyond the nearest
supernode.  Kazaa uses something similar, although no one outside of
their organization knows the details.

"Random walk" does not quite capture it either, but it's closer than
"flooding".  Thes algorithms are actually *very* efficient, producing
light load and broad network reach, although they cannot scale as well
as DHTs, clearly.

That's not particularly relevant to the purposes of this group, but more
for the sake of everyone's general knowledge.

On the DHT side, it may be worth talking to Bryan Turner from the p2p
hackers group, as he's by far the most knowledgeable person on the
various DHT implementations I've come across.  Are you on this list,
Bryan?  He's implemented many of them, and he's a strong advocate of
"The Continuous Discrete Approach".  The DHT world evolves very quickly,
and things like Chord and CAN are certainly not the state of the art.

I'm ccing Bryan on this.  I don't know you, Bryan, but your posts on p2p
hackers make your depth of knowledge in this area clear.

-Adam


buford@research.panasonic.com wrote:

>Here are the slides on P2P overlays for the ad hoc session on Friday
>
>http://www.cs.uml.edu/~buford/irtf-p2prg/p2psip/JBuford-IETF-P2PSIP-Overlay
-Systems-v3.pdf
>
>
>John
>
>
>_______________________________________________
>p2p-sip mailing list
>p2p-sip@cs.columbia.edu
>http://lists.cs.columbia.edu/mailman/listinfo/p2p-sip
>
>
>
_______________________________________________
p2p-sip mailing list
p2p-sip@cs.columbia.edu
http://lists.cs.columbia.edu/mailman/listinfo/p2p-sip

From adamfisk at gmail.com  Fri Nov 11 19:00:25 2005
From: adamfisk at gmail.com (Adam Fisk)
Date: Fri Nov 11 19:00:40 2005
Subject: [p2p-sip] slides for P2P-SIP Ad Hoc
Message-ID: <f40963db0511111600l731e682dw9538008b778d16bf@mail.gmail.com>

 Hi John-

The wiki is wrong, unfortunately. The relevant documents on search are the
Gnutella search RFCs I authored at the following (you may need an account
for these to work):

http://tinyurl.com/d6mr5
http://tinyurl.com/cuzuf

I'm not sure who wrote the wiki entry, but it should be corrected. All of
the major Gnutella clients implement some version of the above, including
LimeWire, BearShare, and Morpheus, and you'll see all of these headers in
almost all Gnutella connections.

Regarding Kazaa, I think the key phrase in that study is: "it may flood the
query over the overlay network to **one or more** of the SNs to which it is
connected." The "one or more" should be very similar to the way Gnutella
does it, aka "random walk" or "dynamic querying" or whatever you want to
call it. This is where the real action takes place -- the hops beyond that
first one to the SN -- because that's where the vast majority of the search
traffic is. Again, I don't know for sure, but the response patterns of Kazaa
nodes so closely resembles those from the protocols we implemented on
Gnutella that it's hard to imagine they're doing something that different.

Now, granted, Gnutella and Kazaa are incorrectly published all over the
place as "flooding" architectures, and that's very true of their initial
implementations. Both are rapidly evolving networks, though, and long ago
moved beyond these roots. Just holding the p2p SIP group to a higher
standard!

On the Chord/CAN/other DHTs issue, it's just that I don't think there's any
reason you'd ever seriously consider Chord or CAN as realistic options over
others out there, so it seems silly to mention them. As far as I know, there
are other DHTs that simply do everything better, so it's not a matter of
weighing the advantages and disadvantages of Chord/CAN versus the advantages
and disadvantages of another algorithm.

I guess I can see Chord having more existing open source implementations and
possibly being easier to implement than some others, but that's about it.
Would you agree with that statement, Bryan? Thanks for your participation.

All the Best,

-Adam



John Buford wrote:

Hi Adam,
Thanks for the comments.

Looking at the wiki for the Gnutella Developers Forum in the
section "Querying the Network"

"A servent SHOULD forward incoming Query messages to all of its directly
connected servents, except the one that delivered the incoming Query.
Servents using Flow control or Ultrapeers (sections 3.1 and 3.2) will not
always forward every Query over every connection. " There is also a TTL on
the packet which is decremented.

This sounds like flooding not random walk.

According to the GDF wiki, during bootstrapping, if a node can't
connect using entries in the host cache,
it can randomly select a GWebCache to populate its host cache.

Regarding Kazaa, the protocol is proprietary. One research group analyzed
pollution in Kazaa (Ross et al. "Pollution in P2P File Sharing Systems")
and built a crawler to crawl the Kazaa network.

>From their paper:

"When a user wants to find files, the user's ON sends
a query with keywords over the TCP connection to its
SN. For each match in its local index, the SN returns the
metadata and IP addresses corresponding to the match.
When a SN receives a query, it may flood the query
over the overlay network to one or more of the SNs
to which it is connected. A given query will in general
visit a small subset of the SNs, and hence will obtain
the metadata information of a small subset of all the ONs."

More recently this same research group has reported on a detailed
analysis of the Kazaa overlay in the paper
J. Liang, R. Kumar, and K.W. Ross, The Kazaa Overlay: A Measurement Study,
Computer Networks (Special Issue on Overlays), to appear.

In this paper they describe, among other things, "connection shuffling"
which both SNs and ONs use. It appears that this connection shuffling
allows peers to reach other parts of the overlay over time, and may be
related to work load re-distribution.

So it appears that within a given set of SN-SN and SN-ON connections
Kazaa uses flooding, but these connections change over time.

Yes, CAN and Chord are not "state of the art" depending on what you
are looking for in an overlay.

Again, thanks for the comments.

John

-----Original Message-----
From: p2p-sip-bounces@cs.columbia.edu
[mailto:p2p-sip-bounces@cs.columbia.edu <p2p-sip-bounces@cs.columbia.edu>]On
Behalf Of Adam Fisk
Sent: Thursday, November 10, 2005 2:14 PM
To: p2p-sip@cs.columbia.edu; bryan.turner@pobox.com
Subject: Re: [p2p-sip] slides for P2P-SIP Ad Hoc


Hi John-

The slides look great. I just wanted to point out that both Gnutella
and Kazaa's search algorithms are more accurately described as random
walks than as flooding. Kazaa has used a massive indexing and random
walk approach for many years, and Gnutella clients have used a similar
approach for at about the last 3 years. Gnutella basically uses bloom
filters to index all data within 2 hops, and then dynamically queries
supernodes depending on how many results are coming back over time. In
many cases this leads to queries not going beyond the nearest
supernode. Kazaa uses something similar, although no one outside of
their organization knows the details.

"Random walk" does not quite capture it either, but it's closer than
"flooding". Thes algorithms are actually **very** efficient, producing
light load and broad network reach, although they cannot scale as well
as DHTs, clearly.

That's not particularly relevant to the purposes of this group, but more
for the sake of everyone's general knowledge.

On the DHT side, it may be worth talking to Bryan Turner from the p2p
hackers group, as he's by far the most knowledgeable person on the
various DHT implementations I've come across. Are you on this list,
Bryan? He's implemented many of them, and he's a strong advocate of
"The Continuous Discrete Approach". The DHT world evolves very quickly,
and things like Chord and CAN are certainly not the state of the art.

I'm ccing Bryan on this. I don't know you, Bryan, but your posts on p2p
hackers make your depth of knowledge in this area clear.

-Adam


buford@research.panasonic.com wrote:



Here are the slides on P2P overlays for the ad hoc session on Friday

http://www.cs.uml.edu/~buford/irtf-p2prg/p2psip/JBuford-IETF-P2PSIP-Overlay<http://www.cs.uml.edu/%7Ebuford/irtf-p2prg/p2psip/JBuford-IETF-P2PSIP-Overlay>

-Systems-v3.pdf


John


_______________________________________________
p2p-sip mailing list
p2p-sip@cs.columbia.edu
http://lists.cs.columbia.edu/mailman/listinfo/p2p-sip




_______________________________________________
p2p-sip mailing list
p2p-sip@cs.columbia.edu
http://lists.cs.columbia.edu/mailman/listinfo/p2p-sip
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://lists.cs.columbia.edu/pipermail/p2p-sip/attachments/20051111/63644386/attachment.html
From kasthuriram_v at yahoo.com  Wed Nov  9 08:07:43 2005
From: kasthuriram_v at yahoo.com (Kasthuri Ram V)
Date: Sat Nov 12 01:51:13 2005
Subject: [p2p-sip] Model Implementation of draft-bryan-sipping-p2p-01
Message-ID: <20051109130743.77877.qmail@web32802.mail.mud.yahoo.com>

Hi,

I would like to know if there is any model software
implementation of draft-bryan-sipping-p2p-01.


Thanks,
Ram
--


	
		
__________________________________ 
Yahoo! Mail - PC Magazine Editors' Choice 2005 
http://mail.yahoo.com
From brturn at cisco.com  Thu Nov 10 15:45:17 2005
From: brturn at cisco.com (Bryan Turner (brturn))
Date: Sat Nov 12 01:51:17 2005
Subject: [p2p-sip] slides for P2P-SIP Ad Hoc
In-Reply-To: <43739BFE.5050107@speedymail.org>
Message-ID: <200511102045.jAAKjHmi013829@rtp-core-2.cisco.com>

Hello,

	Thanks for inviting me to the discussion, although I'm a bit lost on
the context (not a member of the p2p-sip list).

	As Adam has mentioned I've implemented several of the major DHTs
(Chord, Kademlia, Continuous-Discrete) and researched the rest (Pastry, CAN,
etc).  I'm happy to transmit my experiences on this topic.

	I am not up to speed on the unstructured filesharing systems
(Gnutella, Kazaa, etc).  So Adam's input on that topic is new to me. :)

Let me know if I can help,
--Bryan
bryan.turner@pobox.com

-----Original Message-----
From: Adam Fisk [mailto:afisk@speedymail.org] 
Sent: Thursday, November 10, 2005 2:14 PM
To: p2p-sip@cs.columbia.edu; bryan.turner@pobox.com
Subject: Re: [p2p-sip] slides for P2P-SIP Ad Hoc

Hi John-

The slides look great.  I just wanted to point out that both Gnutella and
Kazaa's search algorithms are more accurately described as random walks than
as flooding.  Kazaa has used a massive indexing and random walk approach for
many years, and Gnutella clients have used a similar 
approach for at about the last 3 years.   Gnutella basically uses bloom 
filters to index all data within 2 hops, and then dynamically queries
supernodes depending on how many results are coming back over time.  In many
cases this leads to queries not going beyond the nearest supernode.  Kazaa
uses something similar, although no one outside of their organization knows
the details. 

"Random walk" does not quite capture it either, but it's closer than
"flooding".  Thes algorithms are actually *very* efficient, producing light
load and broad network reach, although they cannot scale as well as DHTs,
clearly.

That's not particularly relevant to the purposes of this group, but more for
the sake of everyone's general knowledge.

On the DHT side, it may be worth talking to Bryan Turner from the p2p
hackers group, as he's by far the most knowledgeable person on the various
DHT implementations I've come across.  Are you on this list, Bryan?  He's
implemented many of them, and he's a strong advocate of "The Continuous
Discrete Approach".  The DHT world evolves very quickly, and things like
Chord and CAN are certainly not the state of the art.

I'm ccing Bryan on this.  I don't know you, Bryan, but your posts on p2p
hackers make your depth of knowledge in this area clear.

-Adam


buford@research.panasonic.com wrote:

>Here are the slides on P2P overlays for the ad hoc session on Friday
>
>http://www.cs.uml.edu/~buford/irtf-p2prg/p2psip/JBuford-IETF-P2PSIP-Ove
>rlay-Systems-v3.pdf
>
>
>John
>
>
>_______________________________________________
>p2p-sip mailing list
>p2p-sip@cs.columbia.edu
>http://lists.cs.columbia.edu/mailman/listinfo/p2p-sip
>
>  
>
From enrico.marocco at TILAB.COM  Sat Nov 12 20:42:06 2005
From: enrico.marocco at TILAB.COM (Enrico Marocco)
Date: Sat Nov 12 20:42:51 2005
Subject: [p2p-sip] First steps towards P2P-SIP
Message-ID: <437699EE.4050900@tilab.com>

Just some thoughts about the recent meeting.

As Henry suggested, maybe it's time to stop beating about the bush: p2p
for rtc is a fact and the longer propretary solutions (_the_ propretary
solution) arise, the harder it will be for a standard proposal being
adopted.  Probably the most effective way for proceeding is to identify
a few items in a charter list and to focus on them as deliverables:
 + distributed resource location (and call establishment);
 + integration with traditional SIP networks;
 + security.

Also one of the most crucial issues, as Cullen reported, is security.
Some aspects of security could, in fact, be harder to provide than
in traditional centralized networks (e.g. how can we efficently
assert identities, without a central authority?)

On the other hand, a very important feature - killer application for
existing solutions - would be NAT traversal.

Here is my proposal: why not trying to define a very "best effort"
core protocol, developing additonal features as separate applications
and providing an efficient way to find them?  In Henning's and
Cullen(et al.)'s works on services in sipping wg, there is some good
stuff which fits well in this domain, IMHO.

Cheers
Enrico


Gruppo Telecom Italia - Direzione e coordinamento di Telecom Italia S.p.A.

====================================================================
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 send an e_mail to 
MailAdmin@tilab.com. Thank you
====================================================================

From buford at research.panasonic.com  Sun Nov 13 18:04:19 2005
From: buford at research.panasonic.com (buford@research.panasonic.com)
Date: Sun Nov 13 18:05:06 2005
Subject: [p2p-sip] slides for P2P-SIP Ad Hoc
Message-ID: <2c40f31498.314982c40f@pintlmail.MITL.Research.Panasonic.COM>

Hi Adam,
Thanks for the pointers for correct documentation on Gnutella,
I'll take a look at them when I get back.

Some other reasons why Chord is talked about
(and I am not advocating Chord for P2P-SIP) are that
its protocol is public and there is some published
work that shows guaranteed query response for
high levels of churn of peers.  

Deterministic behavior seems important for P2P SIP.

Best,
John

Content-Type: multipart/alternative; 
	boundary="----=_Part_1514_10942674.1131753625665"


------=_Part_1514_10942674.1131753625665
Content-Type: text/plain; charset=ISO-8859-1
Content-Transfer-Encoding: quoted-printable
Content-Disposition: inline

 Hi John-

The wiki is wrong, unfortunately. The relevant documents on search are the
Gnutella search RFCs I authored at the following (you may need an account
for these to work):

http://tinyurl.com/d6mr5
http://tinyurl.com/cuzuf

I'm not sure who wrote the wiki entry, but it should be corrected. All of
the major Gnutella clients implement some version of the above, including
LimeWire, BearShare, and Morpheus, and you'll see all of these headers in
almost all Gnutella connections.

Regarding Kazaa, I think the key phrase in that study is: "it may flood the
query over the overlay network to **one or more** of the SNs to which it is
connected." The "one or more" should be very similar to the way Gnutella
does it, aka "random walk" or "dynamic querying" or whatever you want to
call it. This is where the real action takes place -- the hops beyond that
first one to the SN -- because that's where the vast majority of the search
traffic is. Again, I don't know for sure, but the response patterns of Kaza=
a
nodes so closely resembles those from the protocols we implemented on
Gnutella that it's hard to imagine they're doing something that different.

Now, granted, Gnutella and Kazaa are incorrectly published all over the
place as "flooding" architectures, and that's very true of their initial
implementations. Both are rapidly evolving networks, though, and long ago
moved beyond these roots. Just holding the p2p SIP group to a higher
standard!

On the Chord/CAN/other DHTs issue, it's just that I don't think there's any
reason you'd ever seriously consider Chord or CAN as realistic options over
others out there, so it seems silly to mention them. As far as I know, ther=
e
are other DHTs that simply do everything better, so it's not a matter of
weighing the advantages and disadvantages of Chord/CAN versus the advantage=
s
and disadvantages of another algorithm.

I guess I can see Chord having more existing open source implementations an=
d
possibly being easier to implement than some others, but that's about it.
Would you agree with that statement, Bryan? Thanks for your participation.

All the Best,

-Adam



John Buford wrote:

Hi Adam,
Thanks for the comments.

Looking at the wiki for the Gnutella Developers Forum in the
section "Querying the Network"

"A servent SHOULD forward incoming Query messages to all of its directly
connected servents, except the one that delivered the incoming Query.
Servents using Flow control or Ultrapeers (sections 3.1 and 3.2) will not
always forward every Query over every connection. " There is also a TTL on
the packet which is decremented.

This sounds like flooding not random walk.

According to the GDF wiki, during bootstrapping, if a node can't
connect using entries in the host cache,
it can randomly select a GWebCache to populate its host cache.

Regarding Kazaa, the protocol is proprietary. One research group analyzed
pollution in Kazaa (Ross et al. "Pollution in P2P File Sharing Systems")
and built a crawler to crawl the Kazaa network.

>From their paper:

"When a user wants to find files, the user's ON sends
a query with keywords over the TCP connection to its
SN. For each match in its local index, the SN returns the
metadata and IP addresses corresponding to the match.
When a SN receives a query, it may flood the query
over the overlay network to one or more of the SNs
to which it is connected. A given query will in general
visit a small subset of the SNs, and hence will obtain
the metadata information of a small subset of all the ONs."

More recently this same research group has reported on a detailed
analysis of the Kazaa overlay in the paper
J. Liang, R. Kumar, and K.W. Ross, The Kazaa Overlay: A Measurement Study,
Computer Networks (Special Issue on Overlays), to appear.

In this paper they describe, among other things, "connection shuffling"
which both SNs and ONs use. It appears that this connection shuffling
allows peers to reach other parts of the overlay over time, and may be
related to work load re-distribution.

So it appears that within a given set of SN-SN and SN-ON connections
Kazaa uses flooding, but these connections change over time.

Yes, CAN and Chord are not "state of the art" depending on what you
are looking for in an overlay.

Again, thanks for the comments.

John

-----Original Message-----
From: p2p-sip-bounces@cs.columbia.edu
[mailto:p2p-sip-bounces@cs.columbia.edu <p2p-sip-bounces@cs.columbia.edu>]O=
n
Behalf Of Adam Fisk
Sent: Thursday, November 10, 2005 2:14 PM
To: p2p-sip@cs.columbia.edu; bryan.turner@pobox.com
Subject: Re: [p2p-sip] slides for P2P-SIP Ad Hoc


Hi John-

The slides look great. I just wanted to point out that both Gnutella
and Kazaa's search algorithms are more accurately described as random
walks than as flooding. Kazaa has used a massive indexing and random
walk approach for many years, and Gnutella clients have used a similar
approach for at about the last 3 years. Gnutella basically uses bloom
filters to index all data within 2 hops, and then dynamically queries
supernodes depending on how many results are coming back over time. In
many cases this leads to queries not going beyond the nearest
supernode. Kazaa uses something similar, although no one outside of
their organization knows the details.

"Random walk" does not quite capture it either, but it's closer than
"flooding". Thes algorithms are actually **very** efficient, producing
light load and broad network reach, although they cannot scale as well
as DHTs, clearly.

That's not particularly relevant to the purposes of this group, but more
for the sake of everyone's general knowledge.

On the DHT side, it may be worth talking to Bryan Turner from the p2p
hackers group, as he's by far the most knowledgeable person on the
various DHT implementations I've come across. Are you on this list,
Bryan? He's implemented many of them, and he's a strong advocate of
"The Continuous Discrete Approach". The DHT world evolves very quickly,
and things like Chord and CAN are certainly not the state of the art.

I'm ccing Bryan on this. I don't know you, Bryan, but your posts on p2p
hackers make your depth of knowledge in this area clear.

-Adam


buford@research.panasonic.com wrote:



Here are the slides on P2P overlays for the ad hoc session on Friday

http://www.cs.uml.edu/~buford/irtf-p2prg/p2psip/JBuford-IETF-P2PSIP-Overlay=
<http://www.cs.uml.edu/%7Ebuford/irtf-p2prg/p2psip/JBuford-IETF-P2PSIP-Over=
lay>

-Systems-v3.pdf


John


_______________________________________________
p2p-sip mailing list
p2p-sip@cs.columbia.edu
http://lists.cs.columbia.edu/mailman/listinfo/p2p-sip




_______________________________________________
p2p-sip mailing list
p2p-sip@cs.columbia.edu
http://lists.cs.columbia.edu/mailman/listinfo/p2p-sip

------=_Part_1514_10942674.1131753625665
Content-Type: text/html; charset=ISO-8859-1
Content-Transfer-Encoding: quoted-printable
Content-Disposition: inline


<div class=3D"moz-text-flowed" style=3D"font-family: -moz-fixed; font-size:=
 13px;" lang=3D"x-western">Hi John-
<br>
<br>The wiki is wrong, unfortunately. The relevant documents on search are
the Gnutella search RFCs I authored at the following (you may=20
need an account for these to work):
<br>
<br><a class=3D"moz-txt-link-freetext" href=3D"http://tinyurl.com/d6mr5">ht=
tp://tinyurl.com/d6mr5</a>
<br><a class=3D"moz-txt-link-freetext" href=3D"http://tinyurl.com/cuzuf">ht=
tp://tinyurl.com/cuzuf</a>
<br>
<br>I'm not sure who wrote the wiki entry, but it should be corrected. All
of the major Gnutella clients implement some version of the above,
including LimeWire, BearShare, and Morpheus, and you'll see all of these he=
aders in almost all Gnutella connections. <br>
<br>
Regarding Kazaa, I think
the key phrase in that study is: &quot;it may flood the query over the
overlay network to **one or more** of the SNs to which it is
connected.&quot;&nbsp; The &quot;one or more&quot; should be very similar t=
o the way
Gnutella does it,
aka &quot;random walk&quot; or &quot;dynamic querying&quot; or whatever you=
 want to call
it. This is where the
real action takes place -- the hops beyond that first one to the SN --
because that's where the vast majority of the search traffic is. Again,
I don't know for sure, but the response patterns of Kazaa nodes so
closely resembles those from the protocols we implemented on Gnutella
that it's hard to imagine they're doing something that different.
<br>
<br>
Now, granted, Gnutella and Kazaa are incorrectly published all over the
place as &quot;flooding&quot; architectures, and that's very true of their
initial implementations.&nbsp; Both are rapidly evolving networks,
though, and long ago moved beyond these roots.&nbsp; Just holding the
p2p SIP group to a higher standard!<br>
<br>On the Chord/CAN/other DHTs issue, it's just that I don't think there's
any reason you'd ever seriously consider Chord or CAN as realistic
options over others out there, so it seems silly to mention them. As far
as I know, there are other DHTs that simply do everything better, so
it's not a matter of weighing the advantages and disadvantages of Chord/CAN=
 versus the
advantages and disadvantages of another algorithm. <br>
<br>
I guess I can see Chord having more=20
existing open source implementations and possibly being easier to=20
implement than some others, but that's about it. Would you agree with=20
that statement, Bryan?
Thanks for your participation.<br>
<br>All the Best,<br>
<br>-Adam
<br>
<br>
<br>
<br>John Buford wrote:
<br>
<br><blockquote type=3D"cite">Hi Adam,
<br>Thanks for the comments.
<br>
<br>Looking at the wiki for the Gnutella Developers Forum in the
<br>section &quot;Querying the Network&quot;
<br>
<br>&quot;A servent SHOULD forward incoming Query messages to all of its di=
rectly
<br>connected servents, except the one that delivered the incoming Query.
<br>Servents using Flow control or Ultrapeers (sections 3.1 and 3.2) will n=
ot
<br>always forward every Query over every connection. &quot;&nbsp; There is=
 also a TTL on
<br>the packet which is decremented.
<br>
<br>This sounds like flooding not random walk.
<br>
<br>According to the GDF wiki, during bootstrapping, if a node can't
<br>connect using entries in the host cache,
<br>it can randomly select a GWebCache to populate its host cache.
<br>
<br>Regarding Kazaa, the protocol is proprietary.&nbsp; One research group =
analyzed
<br>pollution in Kazaa (Ross et al. &quot;Pollution in P2P File Sharing Sys=
tems&quot;)
<br>and built a crawler to crawl the Kazaa network.
<br><blockquote type=3D"cite">From their paper:
<br></blockquote>&quot;When a user wants to find files, the user's ON sends
<br>a query with keywords over the TCP connection to its
<br>SN. For each match in its local index, the SN returns the
<br>metadata and IP addresses corresponding to the match.
<br>When a SN receives a query, it may flood the query
<br>over the overlay network to one or more of the SNs<br>to which it is co=
nnected. A given query will in general
<br>visit a small subset of the SNs, and hence will obtain
<br>the metadata information of a small subset of all the ONs.&quot;
<br>
<br>More recently this same research group has reported on a detailed
<br>analysis of the Kazaa overlay in the paper
<br>J. Liang, R. Kumar, and K.W. Ross, The Kazaa Overlay: A Measurement Stu=
dy,
<br>Computer Networks (Special Issue on Overlays), to appear.
<br>
<br>In this paper they describe, among other things, &quot;connection shuff=
ling&quot;
<br>which both SNs and ONs use.&nbsp; It appears that this connection shuff=
ling
<br>allows peers to reach other parts of the overlay over time, and may be
<br>related to work load re-distribution.
<br>
<br>So it appears that within a given set of SN-SN and SN-ON connections
<br>Kazaa uses flooding, but these connections change over time.
<br>
<br>Yes, CAN and Chord are not &quot;state of the art&quot; depending on wh=
at you
<br>are looking for in an overlay.
<br>
<br>Again, thanks for the comments.
<br>
<br>John
<br>
<br>-----Original Message-----
<br>From: <a class=3D"moz-txt-link-abbreviated" href=3D"mailto:p2p-sip-boun=
ces@cs.columbia.edu">p2p-sip-bounces@cs.columbia.edu</a>
<br>[<a class=3D"moz-txt-link-freetext" href=3D"mailto:p2p-sip-bounces@cs.c=
olumbia.edu">mailto:p2p-sip-bounces@cs.columbia.edu</a>]On Behalf Of Adam F=
isk
<br>Sent: Thursday, November 10, 2005 2:14 PM
<br>To: <a class=3D"moz-txt-link-abbreviated" href=3D"mailto:p2p-sip@cs.col=
umbia.edu">p2p-sip@cs.columbia.edu</a>; <a class=3D"moz-txt-link-abbreviate=
d" href=3D"mailto:bryan.turner@pobox.com">bryan.turner@pobox.com</a>
<br>Subject: Re: [p2p-sip] slides for P2P-SIP Ad Hoc<br>
<br>
<br>Hi John-
<br>
<br>The slides look great.&nbsp; I just wanted to point out that both Gnute=
lla<br>and Kazaa's search algorithms are more accurately described as rando=
m
<br>walks than as flooding.&nbsp; Kazaa has used a massive indexing and ran=
dom
<br>walk approach for many years, and Gnutella clients have used a similar
<br>approach for at about the last 3 years.&nbsp;&nbsp; Gnutella basically =
uses bloom
<br>filters to index all data within 2 hops, and then dynamically queries
<br>supernodes depending on how many results are coming back over time.&nbs=
p; In
<br>many cases this leads to queries not going beyond the nearest
<br>supernode.&nbsp; Kazaa uses something similar, although no one outside =
of
<br>their organization knows the details.
<br>
<br>&quot;Random walk&quot; does not quite capture it either, but it's clos=
er than
<br>&quot;flooding&quot;.&nbsp; Thes algorithms are actually <b class=3D"mo=
z-txt-star"><span class=3D"moz-txt-tag">*</span>very<span class=3D"moz-txt-=
tag">*</span></b> efficient, producing
<br>light load and broad network reach, although they cannot scale as well
<br>as DHTs, clearly.
<br>
<br>That's not particularly relevant to the purposes of this group, but mor=
e
<br>for the sake of everyone's general knowledge.
<br>
<br>On the DHT side, it may be worth talking to Bryan Turner from the p2p
<br>hackers group, as he's by far the most knowledgeable person on the
<br>various DHT implementations I've come across.&nbsp; Are you on this lis=
t,
<br>Bryan?&nbsp; He's implemented many of them, and he's a strong advocate =
of
<br>&quot;The Continuous Discrete Approach&quot;.&nbsp; The DHT world evolv=
es very quickly,
<br>and things like Chord and CAN are certainly not the state of the art.
<br>
<br>I'm ccing Bryan on this.&nbsp; I don't know you, Bryan, but your posts =
on p2p
<br>hackers make your depth of knowledge in this area clear.
<br>
<br>-Adam
<br>
<br>
<br><a class=3D"moz-txt-link-abbreviated" href=3D"mailto:buford@research.pa=
nasonic.com">buford@research.panasonic.com</a> wrote:
<br>
<br>&nbsp;
<br>
<br><blockquote type=3D"cite">Here are the slides on P2P overlays for the a=
d hoc session on Friday
<br>
<br><a class=3D"moz-txt-link-freetext" href=3D"http://www.cs.uml.edu/%7Ebuf=
ord/irtf-p2prg/p2psip/JBuford-IETF-P2PSIP-Overlay">http://www.cs.uml.edu/~b=
uford/irtf-p2prg/p2psip/JBuford-IETF-P2PSIP-Overlay</a>
<br>&nbsp;&nbsp;=20

<br></blockquote>-Systems-v3.pdf<br>&nbsp;
<br>
<br><blockquote type=3D"cite">John
<br>
<br>
<br>_______________________________________________
<br>p2p-sip mailing list
<br><a class=3D"moz-txt-link-abbreviated" href=3D"mailto:p2p-sip@cs.columbi=
a.edu">p2p-sip@cs.columbia.edu</a>
<br><a class=3D"moz-txt-link-freetext" href=3D"http://lists.cs.columbia.edu=
/mailman/listinfo/p2p-sip">http://lists.cs.columbia.edu/mailman/listinfo/p2=
p-sip</a>
<br>
<br>
<br>
<br>&nbsp;&nbsp;=20

<br></blockquote>_______________________________________________
<br>p2p-sip mailing list
<br><a class=3D"moz-txt-link-abbreviated" href=3D"mailto:p2p-sip@cs.columbi=
a.edu">p2p-sip@cs.columbia.edu</a>
<br><a class=3D"moz-txt-link-freetext" href=3D"http://lists.cs.columbia.edu=
/mailman/listinfo/p2p-sip">http://lists.cs.columbia.edu/mailman/listinfo/p2=
p-sip</a>
<br>
<br>
<br>&nbsp;
<br>
<br></blockquote>
<br>
<br></div>

------=_Part_1514_10942674.1131753625665--
-------------- next part --------------
_______________________________________________
p2p-sip mailing list
p2p-sip@cs.columbia.edu
http://lists.cs.columbia.edu/mailman/listinfo/p2p-sip
From adamfisk at gmail.com  Sun Nov 13 22:02:40 2005
From: adamfisk at gmail.com (Adam Fisk)
Date: Sun Nov 13 22:02:55 2005
Subject: [p2p-sip] slides for P2P-SIP Ad Hoc
In-Reply-To: <2c40f31498.314982c40f@pintlmail.MITL.Research.Panasonic.COM>
References: <2c40f31498.314982c40f@pintlmail.MITL.Research.Panasonic.COM>
Message-ID: <f40963db0511131902o6ef96f3cyac9f8edefab6ae9@mail.gmail.com>

Hi John-

I think my previous links do require registration. I found a link to it on
the Wiki, though, at:

http://www.the-gdf.org/wiki/index.php?title=Dynamic_Query_Protocol

The "Ultrapeer Query Routing" spec doesn't appear to be in the Wiki yet. If
you're a member of the GDF, though, you can navigate to it at:

http://groups.yahoo.com/group/the_gdf/files/Proposals/Working%20Proposals/search/

Note that I'm using the term "RFC" very loosely here.

-Adam


On 11/13/05, buford@research.panasonic.com <buford@research.panasonic.com>
wrote:
>
> Hi Adam,
> Thanks for the pointers for correct documentation on Gnutella,
> I'll take a look at them when I get back.
>
> Some other reasons why Chord is talked about
> (and I am not advocating Chord for P2P-SIP) are that
> its protocol is public and there is some published
> work that shows guaranteed query response for
> high levels of churn of peers.
>
> Deterministic behavior seems important for P2P SIP.
>
> Best,
> John
>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://lists.cs.columbia.edu/pipermail/p2p-sip/attachments/20051113/0b8afb9f/attachment.html
From Richard.Stastny at oefeg.at  Mon Nov 14 10:12:56 2005
From: Richard.Stastny at oefeg.at (Stastny Richard)
Date: Mon Nov 14 10:10:09 2005
Subject: [p2p-sip] First steps towards P2P-SIP
Message-ID: <32755D354E6B65498C3BD9FD496C7D461B221A@oefeg-s04.oefeg.loc>

I fully agree with Enrico that we need to go forward asap
another round of academic discussion is counterproductive

My take from the meeting was the following:

We need a solution for small networks and one for large networks

The small networks (eg up to 10000 users) have eg the following use
cases:

-SME PBX
- adhoc networks eg for meetings, conferences, buddy groups
and also for disaster theatres, eg to set up an onsite adhoc
communication for rescue teams of different organisations

So why not start with these scenarios and leave the large
networks for later.

Security could be the choice of the network:
eg none
a known key (eg such as a WEP key advertised on-site
similar to the secure IETF)
a centralized administerd solution

In these cases most of the proposed DHT schemes would work

Additional minimum requirements for such networks could be
I must be able to interoperable with client-server SIP networks
and also provide UA that allow to connect normal SIP clients.

I last remark: I would prefer the solution with a separate protocol
for P2P.

Regards
Richard



> -----Original Message-----
> From: p2p-sip-bounces@cs.columbia.edu [mailto:p2p-sip-
> bounces@cs.columbia.edu] On Behalf Of Enrico Marocco
> Sent: Sunday, November 13, 2005 2:42 AM
> To: p2p-sip@cs.columbia.edu
> Subject: [p2p-sip] First steps towards P2P-SIP
> 
> Just some thoughts about the recent meeting.
> 
> As Henry suggested, maybe it's time to stop beating about the bush:
p2p
> for rtc is a fact and the longer propretary solutions (_the_
propretary
> solution) arise, the harder it will be for a standard proposal being
> adopted.  Probably the most effective way for proceeding is to
identify
> a few items in a charter list and to focus on them as deliverables:
>  + distributed resource location (and call establishment);
>  + integration with traditional SIP networks;
>  + security.
> 
> Also one of the most crucial issues, as Cullen reported, is security.
> Some aspects of security could, in fact, be harder to provide than
> in traditional centralized networks (e.g. how can we efficently
> assert identities, without a central authority?)
> 
> On the other hand, a very important feature - killer application for
> existing solutions - would be NAT traversal.
> 
> Here is my proposal: why not trying to define a very "best effort"
> core protocol, developing additonal features as separate applications
> and providing an efficient way to find them?  In Henning's and
> Cullen(et al.)'s works on services in sipping wg, there is some good
> stuff which fits well in this domain, IMHO.
> 
> Cheers
> Enrico
> 
> 
> Gruppo Telecom Italia - Direzione e coordinamento di Telecom Italia
S.p.A.
> 
> ====================================================================
> 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 send an e_mail to
> MailAdmin@tilab.com. Thank you
> ====================================================================
> 
> _______________________________________________
> p2p-sip mailing list
> p2p-sip@cs.columbia.edu
> http://lists.cs.columbia.edu/mailman/listinfo/p2p-sip

From BeckW at t-systems.com  Mon Nov 14 10:43:30 2005
From: BeckW at t-systems.com (Beck01, Wolfgang)
Date: Mon Nov 14 10:44:18 2005
Subject: [p2p-sip] First steps towards P2P-SIP
Message-ID: <97A461F3EE5284469E41ED3728202F41121EE9@S4DE9JSAAMW.ost.t-com.de>


Richard wrote:
> 
> We need a solution for small networks and one for large networks
>
I've counted even five scenarios:

1 living room; consumer electronics, low CPU power, low security demands, high churn, small size
2 meeting venue; adhoc networks, friendly users(?), small size, high churn?
3 PBX; low churn, friendly users, moderate size
4 automatic clustering of SIP proxies; low churn, friendly users (= the other proxies in the cluster), small size
5 Internet, low churn, unfriendly users; requires good security, very large number of nodes

2-4 have more in common and seem easier to standardize than 1 and 5. However, the IETF is
about the Internet with capital "I" so ruling out 5) might be used against P2PSIP in
a BoF discussion.


Wolfgang

From dbarrett at quinthar.com  Mon Nov 14 13:02:34 2005
From: dbarrett at quinthar.com (David Barrett)
Date: Mon Nov 14 13:03:08 2005
Subject: [p2p-sip] STUN/TURN/ICE integration
References: <1131991359.B4B1F7@dg11.dngr.org>
Message-ID: <1131991363.1552CF63@bd12.dngr.org>

I'm not sure if it's inside the mandate of this group, but are we open 
to re-considering the separation between SIP and STUN/TURN/ICE?

Given that anyone doing P2P SIP will need to solve the NAT and firewall 
penertration problems, it would be very nice to do it all within a 
single protocol with consisent grammars, conventions, state machines, 
etc.

Speaking as just a solo developer, I found it so difficult to 
understand, build, and follow the various standards (all of which are 
being continuously updated) I just gave up in order to get iGlance 
(http://www.iglance.com) out the door.  In the end I just expanded SIP 
slightly to encompass STUN/TURN/NAT, thereby allowing me to use a single 
FSM and single parser to manage the complete connection establishment 
process.

-david
From eunsoo at research.panasonic.com  Mon Nov 14 13:20:58 2005
From: eunsoo at research.panasonic.com (Eunsoo Shim)
Date: Mon Nov 14 13:21:28 2005
Subject: [p2p-sip] STUN/TURN/ICE integration
In-Reply-To: <1131991363.1552CF63@bd12.dngr.org>
References: <1131991359.B4B1F7@dg11.dngr.org> 
	<1131991363.1552CF63@bd12.dngr.org>
Message-ID: <4378D58A.5050200@research.panasonic.com>

Certainly NAT/firewall traversal is an important problem for P2P SIP.
But I don't think this group should step on what other groups are 
working on, trying to redo or change whatever they did.
The basic approach about STUN/TURN/ICE will be asking those groups 
working on STUN/TURN/ICE to consider any requirements we might have for 
P2P SIP if the requirements were something those groups did not consider 
in the past.

On the other hand, one thing we should do about NAT Traversal is to 
figure how to use STUN/TURN/ICE for P2P SIP, in particular, how to let 
each peer locate  proper STUN/TURN servers.

Regards,

Eunsoo

David Barrett wrote:

> I'm not sure if it's inside the mandate of this group, but are we open 
> to re-considering the separation between SIP and STUN/TURN/ICE?
>
> Given that anyone doing P2P SIP will need to solve the NAT and 
> firewall penertration problems, it would be very nice to do it all 
> within a single protocol with consisent grammars, conventions, state 
> machines, etc.
>
> Speaking as just a solo developer, I found it so difficult to 
> understand, build, and follow the various standards (all of which are 
> being continuously updated) I just gave up in order to get iGlance 
> (http://www.iglance.com) out the door.  In the end I just expanded SIP 
> slightly to encompass STUN/TURN/NAT, thereby allowing me to use a 
> single FSM and single parser to manage the complete connection 
> establishment process.
>
> -david
> _______________________________________________
> p2p-sip mailing list
> p2p-sip@cs.columbia.edu
> http://lists.cs.columbia.edu/mailman/listinfo/p2p-sip


From royrr at us.saic.com  Mon Nov 14 10:59:07 2005
From: royrr at us.saic.com (Roy, Radhika R.)
Date: Tue Nov 15 09:24:48 2005
Subject: [p2p-sip] First steps towards P2P-SIP
Message-ID: <84E5D7F15F52D34DA84C9C90ACE6554F622BBF@0591-its-exmp01.us.saic.com>

Wolfgang:

Let me rephrase it.

Our goal will be item 5 in BoF.

Items 1-4 will be special applications/use cases of item 5 that can be
implemented now keeping the scale of the solution in a limited environment
without violating the Internet principle..

Please note that items 1-4 are still p2p which is the basic principle of the
Internet.

Let me point out the works of MANET WG. It is still used as an access
network to the Internet, but its solution space is p2p in a limited
networking environment.

Any comments?

Best regards,

Radhika


-----Original Message-----
From: p2p-sip-bounces@cs.columbia.edu
[mailto:p2p-sip-bounces@cs.columbia.edu
<mailto:p2p-sip-bounces@cs.columbia.edu> ] On Behalf Of Beck01, Wolfgang
Sent: Monday, November 14, 2005 10:44 AM
To: p2p-sip@cs.columbia.edu
Subject: Re: [p2p-sip] First steps towards P2P-SIP


Richard wrote:

> 

> We need a solution for small networks and one for large networks

>

I've counted even five scenarios:

1 living room; consumer electronics, low CPU power, low security demands,
high churn, small size

2 meeting venue; adhoc networks, friendly users(?), small size, high churn?

3 PBX; low churn, friendly users, moderate size

4 automatic clustering of SIP proxies; low churn, friendly users (= the
other proxies in the cluster), small size

5 Internet, low churn, unfriendly users; requires good security, very large
number of nodes

2-4 have more in common and seem easier to standardize than 1 and 5.
However, the IETF is

about the Internet with capital "I" so ruling out 5) might be used against
P2PSIP in

a BoF discussion.


Wolfgang

-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://lists.cs.columbia.edu/pipermail/p2p-sip/attachments/20051114/4cdd4071/attachment.html
From royrr at us.saic.com  Mon Nov 14 13:54:03 2005
From: royrr at us.saic.com (Roy, Radhika R.)
Date: Tue Nov 15 09:24:52 2005
Subject: [p2p-sip] STUN/TURN/ICE integration
Message-ID: <84E5D7F15F52D34DA84C9C90ACE6554F622BC3@0591-its-exmp01.us.saic.com>

In the same token, security, QoS support, routing, etc. are also needed to
be addressed in p2p SIP.

Best regards,

Radhika

-----Original Message-----
From: p2p-sip-bounces@cs.columbia.edu
[mailto:p2p-sip-bounces@cs.columbia.edu
<mailto:p2p-sip-bounces@cs.columbia.edu> ] On Behalf Of Eunsoo Shim
Sent: Monday, November 14, 2005 1:21 PM
To: David Barrett
Cc: p2p-sip@cs.columbia.edu
Subject: Re: [p2p-sip] STUN/TURN/ICE integration

Certainly NAT/firewall traversal is an important problem for P2P SIP.

But I don't think this group should step on what other groups are 

working on, trying to redo or change whatever they did.

The basic approach about STUN/TURN/ICE will be asking those groups 

working on STUN/TURN/ICE to consider any requirements we might have for 

P2P SIP if the requirements were something those groups did not consider 

in the past.

On the other hand, one thing we should do about NAT Traversal is to 

figure how to use STUN/TURN/ICE for P2P SIP, in particular, how to let 

each peer locate  proper STUN/TURN servers.

Regards,

Eunsoo

David Barrett wrote:

> I'm not sure if it's inside the mandate of this group, but are we open 

> to re-considering the separation between SIP and STUN/TURN/ICE?

>

> Given that anyone doing P2P SIP will need to solve the NAT and 

> firewall penetration problems, it would be very nice to do it all 

> within a single protocol with consistent grammars, conventions, state 

> machines, etc.

>

> Speaking as just a solo developer, I found it so difficult to 

> understand, build, and follow the various standards (all of which are 

> being continuously updated) I just gave up in order to get iGlance 

> (http://www.iglance.com <http://www.iglance.com> ) out the door.  In the
end I just expanded SIP 

> slightly to encompass STUN/TURN/NAT, thereby allowing me to use a 

> single FSM and single parser to manage the complete connection 

> establishment process.

>

> -david

-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://lists.cs.columbia.edu/pipermail/p2p-sip/attachments/20051114/f4b6cc42/attachment.html
From ssal at intracom.gr  Tue Nov 15 11:58:33 2005
From: ssal at intracom.gr (Sotiris Salloum)
Date: Tue Nov 15 10:59:02 2005
Subject: [p2p-sip] Research project regarding p2p-sip security
Message-ID: <200511151543.jAFFhuq14426@hal.intranet.gr>

Dear all, 

 

I'm a PhD candidate at Aegean University, and I plan to study p2p security
issues. I've been previously active with

SIP and I'm considering also p2p sip as one of my fields since I believe
that security is also an issue with p2p. 

 

I would appreciate if you could tell me your opinions on this issues, I
believe this is the right place to post this question if not

so please accept my apologies, and I would be thankful if you could point to
me an alternative mailing list.

 

Regards

Sotiris

 

-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://lists.cs.columbia.edu/pipermail/p2p-sip/attachments/20051115/70e53281/attachment-0001.html
From eunsoo at research.panasonic.com  Tue Nov 15 11:10:07 2005
From: eunsoo at research.panasonic.com (Eunsoo Shim)
Date: Tue Nov 15 11:10:45 2005
Subject: [p2p-sip] Research project regarding p2p-sip security
In-Reply-To: <200511151543.jAFFhuq14426@hal.intranet.gr>
References: <200511151543.jAFFhuq14426@hal.intranet.gr>
Message-ID: <437A085F.30505@research.panasonic.com>

Hi,

There is no doubt that security is an important issue for P2P SIP.
Cullen gave a good presentation about security issues in the ad-hoc 
meeting in Vancouver.
You might want to take a look at his slides later when they become 
available.
Regards,

Eunsoo

Sotiris Salloum wrote:

> Dear all,
>
> I?m a PhD candidate at Aegean University, and I plan to study p2p 
> security issues. I?ve been previously active with
>
> SIP and I?m considering also p2p sip as one of my fields since I 
> believe that security is also an issue with p2p.
>
> I would appreciate if you could tell me your opinions on this issues, 
> I believe this is the right place to post this question if not
>
> so please accept my apologies, and I would be thankful if you could 
> point to me an alternative mailing list.
>
> Regards
>
> Sotiris
>
>------------------------------------------------------------------------
>
>_______________________________________________
>p2p-sip mailing list
>p2p-sip@cs.columbia.edu
>http://lists.cs.columbia.edu/mailman/listinfo/p2p-sip
>  
>

From matthewsp at avaya.com  Tue Nov 15 15:12:15 2005
From: matthewsp at avaya.com (Matthews, Philip (Philip))
Date: Tue Nov 15 15:14:36 2005
Subject: [p2p-sip] Research project regarding p2p-sip security
Message-ID: <DFD0AADB452F8B4E99FB78719CAF0E6101A1A03E@ON0010AVEXU1.global.avaya.com>

Indeed, security is a big issue.
For now, this list is the appropriate place to discuss it.
If/when we get a WG going, then we may have to move
out-of-scope discussions elsewhere, but for now it is pretty wide open.

I would also note that the P2P-Hackers mailing list is a good place
to discuss security of P2P systems in general (not just P2P SIP
systems).
There was a long thread there recently on Authentication issues.

- Philip 

> -----Original Message-----
> From: p2p-sip-bounces@cs.columbia.edu 
> [mailto:p2p-sip-bounces@cs.columbia.edu] On Behalf Of Eunsoo Shim
> Sent: November 15, 2005 11:10
> To: Sotiris Salloum
> Cc: p2p-sip@cs.columbia.edu
> Subject: Re: [p2p-sip] Research project regarding p2p-sip security
> 
> Hi,
> 
> There is no doubt that security is an important issue for P2P SIP.
> Cullen gave a good presentation about security issues in the ad-hoc 
> meeting in Vancouver.
> You might want to take a look at his slides later when they become 
> available.
> Regards,
> 
> Eunsoo
> 
> Sotiris Salloum wrote:
> 
> > Dear all,
> >
> > I'm a PhD candidate at Aegean University, and I plan to study p2p 
> > security issues. I've been previously active with
> >
> > SIP and I'm considering also p2p sip as one of my fields since I 
> > believe that security is also an issue with p2p.
> >
> > I would appreciate if you could tell me your opinions on 
> this issues, 
> > I believe this is the right place to post this question if not
> >
> > so please accept my apologies, and I would be thankful if you could 
> > point to me an alternative mailing list.
> >
> > Regards
> >
> > Sotiris
> >
> >-------------------------------------------------------------
> -----------
> >
> >_______________________________________________
> >p2p-sip mailing list
> >p2p-sip@cs.columbia.edu
> >http://lists.cs.columbia.edu/mailman/listinfo/p2p-sip
> >  
> >
> 
> _______________________________________________
> p2p-sip mailing list
> p2p-sip@cs.columbia.edu
> http://lists.cs.columbia.edu/mailman/listinfo/p2p-sip
> 

From tme at multicasttech.com  Wed Nov 16 19:25:40 2005
From: tme at multicasttech.com (Marshall Eubanks)
Date: Wed Nov 16 19:25:52 2005
Subject: [p2p-sip] Fwd: mboned: FYI: Relayed Multicast Protocol
References: <20051116151835.GC14523@1-4-5.net>
Message-ID: <FE40E42B-57F6-409F-9C2C-2D90E59DEABF@multicasttech.com>

I thought this might be interesting to the group. The file is at

http://www.americafree.tv/papers/RCMP_H1.pdf

(The list moderator can kill the embargoed version with an attachment).

Regards
Marshall Eubanks

Begin forwarded message:

> From: David Meyer <dmm@1-4-5.net>
> Date: November 16, 2005 10:18:35 AM EST
> To: mboned@lists.uoregon.edu
> Subject: mboned: FYI: Relayed Multicast Protocol
>
> 	pdf attached.
>
> 	Dave
>
>
>>> ITU Standardized Technology Set To Spawn New Era Web Broadcast
>>>  _____
>>>
>>> With the popularity of blogs, podcasting and web-based photo
>>> libraries,
>>> web content has become something much more accessible to the
>>> individual
>>> in the last few years. Now video looks set to be the next media to
>>> gain
>>> popularity with the new generation of home based media moguls.
>>>
>>> Work on a new protocol that may spawn a whole new generation of
>>> independent broadcasters is underway at ITU. Relayed Multicast
>>> Protocol
>>> (RMCP) being developed by the ITU's Study Group 17
>>> <http://www.itu.int/ITU-T/studygroups/com17/mgmt.html>  uses  
>>> something
>>> like a peer-to-peer model meaning that independent broadcasters no
>>> longer have to subscribe to a fat-pipe, instead relying on a
>>> collection
>>> of 'peers' or 'relay agents', in other words other people's  
>>> computers.
>>> Peer to peer type traffic is reckoned to make-up as much as 72 per
>>> cent
>>> of current Internet traffic. And this figure is predicted to rise.
>>>
>>> RMCP allows the live broadcast of video or audio piggy-backing off
>>> other
>>> users (or servers). So in a scenario where 100 people are  
>>> demanding a
>>> live broadcast, instead of serving each one of these clients  
>>> their own
>>> video stream, only one stream has to be provided and each user  
>>> will be
>>> served from another in the network. This has significant  
>>> implications
>>> for instance for businesses broadcasting live events, where a  
>>> previous
>>> scenario demanded 100 users be fed individual feeds, RMCP allows the
>>> broadcast of just one.
>>>
>>> Juyoung Park the editor of the ITU-T Recommendations says that RMCP
>>> allows for the efficient serving of hundreds of thousands of
>>> simultaneous connection requests.
>>>
>>> Park says that the need for this type of protocol was identified by
>>> content providers. Standardization means that a single client can
>>> receive content from any number of suppliers.
>>>
>>> An alternative solution - IP Multicast - is not applicable in  
>>> today's
>>> networks according to Park. For a start the success of IP Multicast
>>> would mean router upgrades throughout a network, something that many
>>> operators would balk at, especially given the unclear benefits of IP
>>> Multicast to their revenue streams.
>>>
>>> Park says that tests by his organization - ETRI - have shown that
>>> speeds
>>> of 2 Mbit/s are possible. This reflects standard broadcast rates.
>>> However he says that typically users will experience something more
>>> like
>>> 640 Kbit/s.
>>>
>>> ITU-T has published one Recommendation (ITU-T Rec. X.603) on the  
>>> topic
>>> outlining requirements, framework etc. The next two Recommendations
>>> due
>>> in 2006 will focus on the technical specifications. One focusing on
>>> one
>>> broadcaster to many clients, and the other on many broadcasters to
>>> many
>>> clients.
>
> ----- End forwarded message -----

From bryan at ethernot.org  Wed Nov 16 21:28:03 2005
From: bryan at ethernot.org (David A. Bryan)
Date: Wed Nov 16 21:28:17 2005
Subject: [p2p-sip] Ad-Hoc Meeting Notes and Slides Available
Message-ID: <8b2769930511161828l496a1fbn62b00df3d6c37ee2@mail.gmail.com>

I have put up Dean's notes and all the slides from the ad hoc meeting
on P2P SIP at p2psip.org.

http://www.p2psip.org/ietf64.php

As always, comments, corrections, reports of omissions and broken
links are appreciated. Many thanks to all who attended.

David Bryan

From enrico.marocco at TILAB.COM  Thu Nov 17 03:41:42 2005
From: enrico.marocco at TILAB.COM (Enrico Marocco)
Date: Thu Nov 17 03:43:02 2005
Subject: [p2p-sip] [Fwd: [Sipping] Ad-Hoc Meeting Notes and Slides Available]
Message-ID: <437C4246.6050504@tilab.com>

People reading this ml only may be interested...

Cheers
Enrico

-------- Original Message --------
Subject: [Sipping] Ad-Hoc Meeting Notes and Slides Available
Date: Wed, 16 Nov 2005 21:28:03 -0500
From: David A. Bryan <bryan@ethernot.org>
To: P2P-SIP <p2p-sip@cs.columbia.edu>, sipping <sipping@ietf.org>

I have put up Dean's notes and all the slides from the ad hoc meeting
on P2P SIP at p2psip.org.

http://www.p2psip.org/ietf64.php

As always, comments, corrections, reports of omissions and broken
links are appreciated. Many thanks to all who attended.

David Bryan

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


Gruppo Telecom Italia - Direzione e coordinamento di Telecom Italia S.p.A.

====================================================================
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 send an e_mail to 
MailAdmin@tilab.com. Thank you
====================================================================

From tme at multicasttech.com  Wed Nov 16 19:08:41 2005
From: tme at multicasttech.com (Marshall Eubanks)
Date: Sun Nov 20 09:52:28 2005
Subject: [p2p-sip] Fwd: mboned: FYI: Relayed Multicast Protocol
References: <20051116151835.GC14523@1-4-5.net>
Message-ID: <702CCE93-43A9-44C8-A484-329B7AC8F9DD@multicasttech.com>

Thought this might be interesting to this  group.

Regards
Marshall Eubanks

Begin forwarded message:

> From: David Meyer <dmm@1-4-5.net>
> Date: November 16, 2005 10:18:35 AM EST
> To: mboned@lists.uoregon.edu
> Subject: mboned: FYI: Relayed Multicast Protocol
>
> 	pdf attached.
>
> 	Dave
>
>
>>> ITU Standardized Technology Set To Spawn New Era Web Broadcast
>>>  _____
>>>
>>> With the popularity of blogs, podcasting and web-based photo
>>> libraries,
>>> web content has become something much more accessible to the
>>> individual
>>> in the last few years. Now video looks set to be the next media to
>>> gain
>>> popularity with the new generation of home based media moguls.
>>>
>>> Work on a new protocol that may spawn a whole new generation of
>>> independent broadcasters is underway at ITU. Relayed Multicast
>>> Protocol
>>> (RMCP) being developed by the ITU's Study Group 17
>>> <http://www.itu.int/ITU-T/studygroups/com17/mgmt.html>  uses  
>>> something
>>> like a peer-to-peer model meaning that independent broadcasters no
>>> longer have to subscribe to a fat-pipe, instead relying on a
>>> collection
>>> of 'peers' or 'relay agents', in other words other people's  
>>> computers.
>>> Peer to peer type traffic is reckoned to make-up as much as 72 per
>>> cent
>>> of current Internet traffic. And this figure is predicted to rise.
>>>
>>> RMCP allows the live broadcast of video or audio piggy-backing off
>>> other
>>> users (or servers). So in a scenario where 100 people are  
>>> demanding a
>>> live broadcast, instead of serving each one of these clients  
>>> their own
>>> video stream, only one stream has to be provided and each user  
>>> will be
>>> served from another in the network. This has significant  
>>> implications
>>> for instance for businesses broadcasting live events, where a  
>>> previous
>>> scenario demanded 100 users be fed individual feeds, RMCP allows the
>>> broadcast of just one.
>>>
>>> Juyoung Park the editor of the ITU-T Recommendations says that RMCP
>>> allows for the efficient serving of hundreds of thousands of
>>> simultaneous connection requests.
>>>
>>> Park says that the need for this type of protocol was identified by
>>> content providers. Standardization means that a single client can
>>> receive content from any number of suppliers.
>>>
>>> An alternative solution - IP Multicast - is not applicable in  
>>> today's
>>> networks according to Park. For a start the success of IP Multicast
>>> would mean router upgrades throughout a network, something that many
>>> operators would balk at, especially given the unclear benefits of IP
>>> Multicast to their revenue streams.
>>>
>>> Park says that tests by his organization - ETRI - have shown that
>>> speeds
>>> of 2 Mbit/s are possible. This reflects standard broadcast rates.
>>> However he says that typically users will experience something more
>>> like
>>> 640 Kbit/s.
>>>
>>> ITU-T has published one Recommendation (ITU-T Rec. X.603) on the  
>>> topic
>>> outlining requirements, framework etc. The next two Recommendations
>>> due
>>> in 2006 will focus on the technical specifications. One focusing on
>>> one
>>> broadcaster to many clients, and the other on many broadcasters to
>>> many
>>> clients.
>
> ----- End forwarded message -----
-------------- next part --------------
A non-text attachment was scrubbed...
Name: H1.pdf
Type: application/pdf
Size: 880503 bytes
Desc: not available
Url : http://lists.cs.columbia.edu/pipermail/p2p-sip/attachments/20051116/efb51129/H1-0001.pdf
-------------- next part --------------

From eunsoo at research.panasonic.com  Sun Nov 20 17:18:53 2005
From: eunsoo at research.panasonic.com (Eunsoo Shim)
Date: Sun Nov 20 17:23:12 2005
Subject: [p2p-sip] Fwd: mboned: FYI: Relayed Multicast Protocol
In-Reply-To: <702CCE93-43A9-44C8-A484-329B7AC8F9DD@multicasttech.com>
References: <20051116151835.GC14523@1-4-5.net> 
	<702CCE93-43A9-44C8-A484-329B7AC8F9DD@multicasttech.com>
Message-ID: <4380F64D.9050206@research.panasonic.com>

Thanks for the information.
I had an impression that ITU-T would be slow in adopting new technologies.
Probably I had a wrong impression.

I notice that the ITU-T folks focus on video/audio broadcast as an
application of the p2p multicast.
Does anyone know what protocol do they use for session management?
Regards,

Eunsoo

Marshall Eubanks wrote:

> Thought this might be interesting to this  group.
>
> Regards
> Marshall Eubanks
>
> Begin forwarded message:
>
>> From: David Meyer <dmm@1-4-5.net>
>> Date: November 16, 2005 10:18:35 AM EST
>> To: mboned@lists.uoregon.edu
>> Subject: mboned: FYI: Relayed Multicast Protocol
>>
>>     pdf attached.
>>
>>     Dave
>>
>>
>>>> ITU Standardized Technology Set To Spawn New Era Web Broadcast
>>>>  _____
>>>>
>>>> With the popularity of blogs, podcasting and web-based photo
>>>> libraries,
>>>> web content has become something much more accessible to the
>>>> individual
>>>> in the last few years. Now video looks set to be the next media to
>>>> gain
>>>> popularity with the new generation of home based media moguls.
>>>>
>>>> Work on a new protocol that may spawn a whole new generation of
>>>> independent broadcasters is underway at ITU. Relayed Multicast
>>>> Protocol
>>>> (RMCP) being developed by the ITU's Study Group 17
>>>> <http://www.itu.int/ITU-T/studygroups/com17/mgmt.html>  uses 
>>>> something
>>>> like a peer-to-peer model meaning that independent broadcasters no
>>>> longer have to subscribe to a fat-pipe, instead relying on a
>>>> collection
>>>> of 'peers' or 'relay agents', in other words other people's 
>>>> computers.
>>>> Peer to peer type traffic is reckoned to make-up as much as 72 per
>>>> cent
>>>> of current Internet traffic. And this figure is predicted to rise.
>>>>
>>>> RMCP allows the live broadcast of video or audio piggy-backing off
>>>> other
>>>> users (or servers). So in a scenario where 100 people are  demanding a
>>>> live broadcast, instead of serving each one of these clients  their
>>>> own
>>>> video stream, only one stream has to be provided and each user 
>>>> will be
>>>> served from another in the network. This has significant  implications
>>>> for instance for businesses broadcasting live events, where a 
>>>> previous
>>>> scenario demanded 100 users be fed individual feeds, RMCP allows the
>>>> broadcast of just one.
>>>>
>>>> Juyoung Park the editor of the ITU-T Recommendations says that RMCP
>>>> allows for the efficient serving of hundreds of thousands of
>>>> simultaneous connection requests.
>>>>
>>>> Park says that the need for this type of protocol was identified by
>>>> content providers. Standardization means that a single client can
>>>> receive content from any number of suppliers.
>>>>
>>>> An alternative solution - IP Multicast - is not applicable in  today's
>>>> networks according to Park. For a start the success of IP Multicast
>>>> would mean router upgrades throughout a network, something that many
>>>> operators would balk at, especially given the unclear benefits of IP
>>>> Multicast to their revenue streams.
>>>>
>>>> Park says that tests by his organization - ETRI - have shown that
>>>> speeds
>>>> of 2 Mbit/s are possible. This reflects standard broadcast rates.
>>>> However he says that typically users will experience something more
>>>> like
>>>> 640 Kbit/s.
>>>>
>>>> ITU-T has published one Recommendation (ITU-T Rec. X.603) on the 
>>>> topic
>>>> outlining requirements, framework etc. The next two Recommendations
>>>> due
>>>> in 2006 will focus on the technical specifications. One focusing on
>>>> one
>>>> broadcaster to many clients, and the other on many broadcasters to
>>>> many
>>>> clients.
>>>
>>
>> ----- End forwarded message -----
>
>
>------------------------------------------------------------------------
>
>_______________________________________________
>p2p-sip mailing list
>p2p-sip@cs.columbia.edu
>http://lists.cs.columbia.edu/mailman/listinfo/p2p-sip
>  
>

From henry at pulver.com  Mon Nov 21 14:56:37 2005
From: henry at pulver.com (Henry Sinnreich)
Date: Mon Nov 21 14:57:16 2005
Subject: [p2p-sip]  Relayed Multicast Protocol
In-Reply-To: <4380F64D.9050206@research.panasonic.com>
Message-ID: <200511211956.jALJuiiP017050@cs.columbia.edu>

Do we know how RMCP compares to BitTorrent and the like?

>I had an impression that ITU-T would be slow in adopting new technologies.
>Probably I had a wrong impression.

I believe you are still right :-). P2P has been around for years and seems
to dominate the Internet traffic. Only the standards organizations,
including the IETF seem to have overlooked P2P, or relegated P2P to the IRG
on P2P.

Thanks, Henry

-----Original Message-----
From: Eunsoo Shim [mailto:eunsoo@research.panasonic.com] 
Sent: Sunday, November 20, 2005 4:19 PM
To: Marshall Eubanks; p2p-sip@cs.columbia.edu
Subject: Re: [p2p-sip] Fwd: mboned: FYI: Relayed Multicast Protocol

Thanks for the information.
I had an impression that ITU-T would be slow in adopting new technologies.
Probably I had a wrong impression.

I notice that the ITU-T folks focus on video/audio broadcast as an
application of the p2p multicast.
Does anyone know what protocol do they use for session management?
Regards,

Eunsoo

Marshall Eubanks wrote:

> Thought this might be interesting to this  group.
>
> Regards
> Marshall Eubanks
>
> Begin forwarded message:
>
>> From: David Meyer <dmm@1-4-5.net>
>> Date: November 16, 2005 10:18:35 AM EST
>> To: mboned@lists.uoregon.edu
>> Subject: mboned: FYI: Relayed Multicast Protocol
>>
>>     pdf attached.
>>
>>     Dave
>>
>>
>>>> ITU Standardized Technology Set To Spawn New Era Web Broadcast
>>>>  _____
>>>>
>>>> With the popularity of blogs, podcasting and web-based photo
>>>> libraries,
>>>> web content has become something much more accessible to the
>>>> individual
>>>> in the last few years. Now video looks set to be the next media to
>>>> gain
>>>> popularity with the new generation of home based media moguls.
>>>>
>>>> Work on a new protocol that may spawn a whole new generation of
>>>> independent broadcasters is underway at ITU. Relayed Multicast
>>>> Protocol
>>>> (RMCP) being developed by the ITU's Study Group 17
>>>> <http://www.itu.int/ITU-T/studygroups/com17/mgmt.html>  uses 
>>>> something
>>>> like a peer-to-peer model meaning that independent broadcasters no
>>>> longer have to subscribe to a fat-pipe, instead relying on a
>>>> collection
>>>> of 'peers' or 'relay agents', in other words other people's 
>>>> computers.
>>>> Peer to peer type traffic is reckoned to make-up as much as 72 per
>>>> cent
>>>> of current Internet traffic. And this figure is predicted to rise.
>>>>
>>>> RMCP allows the live broadcast of video or audio piggy-backing off
>>>> other
>>>> users (or servers). So in a scenario where 100 people are  demanding a
>>>> live broadcast, instead of serving each one of these clients  their
>>>> own
>>>> video stream, only one stream has to be provided and each user 
>>>> will be
>>>> served from another in the network. This has significant  implications
>>>> for instance for businesses broadcasting live events, where a 
>>>> previous
>>>> scenario demanded 100 users be fed individual feeds, RMCP allows the
>>>> broadcast of just one.
>>>>
>>>> Juyoung Park the editor of the ITU-T Recommendations says that RMCP
>>>> allows for the efficient serving of hundreds of thousands of
>>>> simultaneous connection requests.
>>>>
>>>> Park says that the need for this type of protocol was identified by
>>>> content providers. Standardization means that a single client can
>>>> receive content from any number of suppliers.
>>>>
>>>> An alternative solution - IP Multicast - is not applicable in  today's
>>>> networks according to Park. For a start the success of IP Multicast
>>>> would mean router upgrades throughout a network, something that many
>>>> operators would balk at, especially given the unclear benefits of IP
>>>> Multicast to their revenue streams.
>>>>
>>>> Park says that tests by his organization - ETRI - have shown that
>>>> speeds
>>>> of 2 Mbit/s are possible. This reflects standard broadcast rates.
>>>> However he says that typically users will experience something more
>>>> like
>>>> 640 Kbit/s.
>>>>
>>>> ITU-T has published one Recommendation (ITU-T Rec. X.603) on the 
>>>> topic
>>>> outlining requirements, framework etc. The next two Recommendations
>>>> due
>>>> in 2006 will focus on the technical specifications. One focusing on
>>>> one
>>>> broadcaster to many clients, and the other on many broadcasters to
>>>> many
>>>> clients.
>>>
>>
>> ----- End forwarded message -----
>
>
>------------------------------------------------------------------------
>
>_______________________________________________
>p2p-sip mailing list
>p2p-sip@cs.columbia.edu
>http://lists.cs.columbia.edu/mailman/listinfo/p2p-sip
>  
>



From TanyaC at Kgex.com.tw  Tue Nov 22 00:02:25 2005
From: TanyaC at Kgex.com.tw (=?big5?B?TUtULVRhbnlhIENob3WpUMBSqXk=?=)
Date: Tue Nov 22 00:03:47 2005
Subject: [p2p-sip] RE: Welcome to the "p2p-sip" mailing list (Digest mode)
Message-ID: <DC2FBC4FF023C64395DF8ED8EBF97583049BF223@khq-ems.kgex.com.tw>



-----Original Message-----
From: p2p-sip-request@cs.columbia.edu
[mailto:p2p-sip-request@cs.columbia.edu]
Sent: Tuesday, November 22, 2005 1:00 PM
Subject: Welcome to the "p2p-sip" mailing list (Digest mode)


Welcome to the p2p-sip@cs.columbia.edu mailing list!

To post to this list, send your email to:

  p2p-sip@cs.columbia.edu

General information about the mailing list is at:

  http://lists.cs.columbia.edu/mailman/listinfo/p2p-sip

If you ever want to unsubscribe or change your options (eg, switch to
or from digest mode, change your password, etc.), visit your
subscription page at:

  http://lists.cs.columbia.edu/mailman/options/p2p-sip/tanyac%40kgex.com.tw


You can also make such adjustments via email by sending a message to:

  p2p-sip-request@cs.columbia.edu

with the word `help' in the subject or body (don't include the
quotes), and you will get back a message with instructions.

You must know your password to change your options (including changing
the password, itself) or to unsubscribe.  It is:

  11111111

Normally, Mailman will remind you of your cs.columbia.edu mailing list
passwords once every month, although you can disable this if you
prefer.  This reminder will also include instructions on how to
unsubscribe or change your account options.  There is also a button on
your options page that will email your current password to you.
From BeckW at t-systems.com  Tue Nov 22 03:56:52 2005
From: BeckW at t-systems.com (Beck01, Wolfgang)
Date: Tue Nov 22 03:57:36 2005
Subject: [p2p-sip] Fwd: mboned: FYI: Relayed Multicast Protocol
Message-ID: <97A461F3EE5284469E41ED3728202F41121EF2@S4DE9JSAAMW.ost.t-com.de>


After skimming through X.603, I don't think it is very helpful for P2P SIP.
Think of a BitTorrent for real-time streams. Content location is central, content
distribution is decentral.

The multicast application is accessed through a central server, which performs AAA
and returns a list of 'multicast agents' MAs. The MAs are the peers which already
receive and redistribute the multicast traffic (application layer multicast). The
new MA connects to one of them and becomes part of the overlay. RMCP will support
1-n and n-n multicast.

It's an interesting proposal for application layer multicast, but P2P SIP has different
requirements.


Wolfgang

--
T-Systems
Next Generation IP Services and Systems
+49 6151 937 2863
Am Kavalleriesand 3
64295 Darmstadt
Germany 
From enrico.marocco at TILAB.COM  Tue Nov 22 03:57:54 2005
From: enrico.marocco at TILAB.COM (Enrico Marocco)
Date: Tue Nov 22 03:58:09 2005
Subject: [p2p-sip]  Relayed Multicast Protocol
In-Reply-To: <200511211956.jALJuiiP017050@cs.columbia.edu>
References: <200511211956.jALJuiiP017050@cs.columbia.edu>
Message-ID: <4382DD92.2060908@tilab.com>

Henry Sinnreich wrote:
>>I had an impression that ITU-T would be slow in adopting new technologies.
>>Probably I had a wrong impression.
> 
> I believe you are still right :-). P2P has been around for years and seems
> to dominate the Internet traffic. Only the standards organizations,
> including the IETF seem to have overlooked P2P, or relegated P2P to the IRG
> on P2P.

Maybe we are late, but aren't we going in the right direction?
David is collecting use cases in a document (even though everybody
agrees that the one use case could be "Skype") which will be ready in a
few weeks; we have already agreed on a charter and we've seen a lot of
people interested in creating a WG.
Don't you think we will be ready for a bof in Dallas?
For a big standard organization this is the right way to go IMHO.

Cheers
Enrico


Gruppo Telecom Italia - Direzione e coordinamento di Telecom Italia S.p.A.

====================================================================
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 send an e_mail to
MailAdmin@tilab.com. Thank you
====================================================================

From eunsoo at research.panasonic.com  Tue Nov 22 08:13:11 2005
From: eunsoo at research.panasonic.com (Eunsoo Shim)
Date: Tue Nov 22 08:31:31 2005
Subject: [p2p-sip]  Relayed Multicast Protocol
In-Reply-To: <4382DD92.2060908@tilab.com>
References: <200511211956.jALJuiiP017050@cs.columbia.edu> 
	<4382DD92.2060908@tilab.com>
Message-ID: <43831967.5060907@research.panasonic.com>

Enrico Marocco wrote:

>Henry Sinnreich wrote:
>  
>
>>>I had an impression that ITU-T would be slow in adopting new technologies.
>>>Probably I had a wrong impression.
>>>      
>>>
>>I believe you are still right :-). P2P has been around for years and seems
>>to dominate the Internet traffic. Only the standards organizations,
>>including the IETF seem to have overlooked P2P, or relegated P2P to the IRG
>>on P2P.
>>    
>>
>
>Maybe we are late, 
>

>but aren't we going in the right direction?
>  
>
Yes, yes. :-)

>David is collecting use cases in a document (even though everybody
>agrees that the one use case could be "Skype") which will be ready in a
>few weeks; we have already agreed on a charter and we've seen a lot of
>people interested in creating a WG.
>Don't you think we will be ready for a bof in Dallas?
>  
>
At least we should try to become ready for a BOF in Dallas.
I believe we will be ready.
We will see more drafts soon including a use cases draft.

Regards,

Eunsoo
From slavitch at objectworld.com  Tue Nov 22 08:43:49 2005
From: slavitch at objectworld.com (Michael Slavitch)
Date: Tue Nov 22 08:44:59 2005
Subject: [p2p-sip]  Relayed Multicast Protocol
Message-ID: <4E2DC1ACE541B94A946B8D8657E372C5176FA2@photon.objectworld.com>

It seems that everyone is understanding the problem and working towards
common goals using open techniques. This should be encouraged, not
discouraged, as the idea of maintaining independent silos of thought is
what is slowing adoption of new technologies. The IETF has been
increasingly guilty of this of late. Let's not let this chauvinism
pollute P2P-SIP.

 
 
 
It is amazing what you can accomplish if you do not care who gets the
credit. 
 
- Harry S Truman

-----Original Message-----
From: p2p-sip-bounces@cs.columbia.edu
[mailto:p2p-sip-bounces@cs.columbia.edu] On Behalf Of Eunsoo Shim
Sent: November 22, 2005 8:13 AM
To: Enrico Marocco
Cc: p2p-sip@cs.columbia.edu; henry@pulver.com
Subject: Re: [p2p-sip] Relayed Multicast Protocol

Enrico Marocco wrote:

>Henry Sinnreich wrote:
>  
>
>>>I had an impression that ITU-T would be slow in adopting new
technologies.
>>>Probably I had a wrong impression.
>>>      
>>>
>>I believe you are still right :-). P2P has been around for years and 
>>seems to dominate the Internet traffic. Only the standards 
>>organizations, including the IETF seem to have overlooked P2P, or 
>>relegated P2P to the IRG on P2P.
>>    
>>
>
>Maybe we are late,
>

>but aren't we going in the right direction?
>  
>
Yes, yes. :-)

>David is collecting use cases in a document (even though everybody
>agrees that the one use case could be "Skype") which will be ready in a
>few weeks; we have already agreed on a charter and we've seen a lot of
>people interested in creating a WG.
>Don't you think we will be ready for a bof in Dallas?
>  
>
At least we should try to become ready for a BOF in Dallas.
I believe we will be ready.
We will see more drafts soon including a use cases draft.

Regards,

Eunsoo
_______________________________________________
p2p-sip mailing list
p2p-sip@cs.columbia.edu
http://lists.cs.columbia.edu/mailman/listinfo/p2p-sip


-- 
BEGIN-ANTISPAM-VOTING-LINKS
------------------------------------------------------
Teach CanIt if this mail (ID 458910) is spam:
Spam:
http://asteroid.objectworld.com/canit/b.php?c=s&i=458910&m=c0b934115ce3
Not spam:
http://asteroid.objectworld.com/canit/b.php?c=n&i=458910&m=c0b934115ce3
Forget vote:
http://asteroid.objectworld.com/canit/b.php?c=f&i=458910&m=c0b934115ce3
------------------------------------------------------
END-ANTISPAM-VOTING-LINKS


From davidbryan at gmail.com  Fri Nov 25 17:00:40 2005
From: davidbryan at gmail.com (David A. Bryan)
Date: Fri Nov 25 17:00:54 2005
Subject: [p2p-sip] Video from the Ad-Hoc in Vancouver
Message-ID: <8b2769930511251400r1cb18b7dxaa356cfe60cce88b@mail.gmail.com>

Erik Lagerway from Eyeball has the first 1/3 of the video from the P2P
SIP Ad-Hoc Meeting up for folks to download and view, so if you missed
it or want to remember what went on, check out:

http://blog.tmcnet.com/sipthat/p2p-sip-voip-video-im/p2p-sip-at-the-64th-ietf-video-footage.asp

and thanks again to Erik for the footage and editing.

I'll also link to this from p2psip.org.

Dave

From davidbryan at gmail.com  Wed Nov 30 02:10:26 2005
From: davidbryan at gmail.com (David A. Bryan)
Date: Wed Nov 30 02:11:03 2005
Subject: [p2p-sip] New P2P SIP use cases internet-draft
Message-ID: <8b2769930511292310u1f47dd5i1d4064a035908015@mail.gmail.com>

Eunsoo Shim, Bruce Lowekamp, and myself have completed an attempt at
collecting some use cases for P2P SIP and submitted this work as an
informational internet-draft. Until it becomes available in the
archive, the document can be viewed here:

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

We look forward to any comments, corrections, complaints, or additions
from the list.

Thanks,

David Bryan

