From dean.willis at softarmor.com  Sun Dec  3 01:18:11 2006
From: dean.willis at softarmor.com (Dean Willis)
Date: Sun, 3 Dec 2006 00:18:11 -0600
Subject: [p2p-sip] Where are we "getting wrapped around the DNS axle"
Message-ID: <BA8A1D67-8623-497A-B105-51E6AFC7FB0D@softarmor.com>


I think we're arguing at cross-purposes about DNS here.

Let me see if I can clarify things a bit:

Assumptions I think we agree on:

1) IF our rendezvous/bootstrapper/cert-server/whatever-we-start-with  
is listed in DNS, and our UA has a hostname for that whatever, THEN  
we MAY be able to use DNS to turn that hostname into an IP address.

2) If we're using DNS, as in #1, then we MAY use dynamic DNS to  
insert a record associating a hostname with an IP address. We MAY use  
conventional DNS configuration as an alternative.

3) The process of associating a hostname with an IP address in DNS is  
completely disjoint from the process of looking up a hostname in DNS  
to discover an IP address. That is, the lookup process is unrelated  
to and not dependent in any way on the configuration process.


Things we MIGHT agree on:

1) If DNS is available, we MAY use it. However, we presume that DNS  
is not always available, so we MUST NOT depend on it. For example, a  
Skype-like configuration where UAs are configured with a default  
rendezvous IP address should be workable.

2) If multicast is available, we MAY use mDNS. However, we presume  
that multicast will not always be available, so we MUST NOT depend on  
mDNS in all cases.


One MIGHT envision a resolution model that looks something like:

IF configured with a rendezvous/bootstrap/...whatever that looks like  
an IP address,
   THEN use that IP address
ELSE
   IF bootstrap looks like a hostname AND
     IF DNS is available AND returns an answer
       THEN use the address returned by DNS
     ELSE
       IF multicast is available AND returns an answer
         THEN use the address returned by mDNS
       ELSE bootstrap is broken, give up and die
       END IF multicast ...
    END IF DNS
   ELSE bootstrap is broken, give up and die
   END IF bootstrap ...
END IF configured ...



Things we have NO agreement on

1) Mandating DNS as the only resolution method for discovering the IP  
address of an entry point to an overlay

2) Using dynamic DNS as the registration/resolution mechanism for  
finding nodes in an overlay. While this could work, it's just a  
straightforward application of dynamic DNS and plain old SIP,  
consequently requiring no new work. It also doesn't meet many of the  
use cases given for P2PSIP.


--
dean

From matthew at matthew.at  Sun Dec  3 02:07:05 2006
From: matthew at matthew.at (Matthew Kaufman)
Date: Sat, 2 Dec 2006 23:07:05 -0800
Subject: [p2p-sip] Fwd: what's wrong with DNS?
In-Reply-To: <0BA3F014-8AC0-4071-B81B-100D28C98E3E@ag-projects.com>
Message-ID: <01b101c716a9$a469cdf0$029b0a0a@matthewdeskxps>

Adrian Georgescu:
>
> The DNS can solve only the address resolution of the non-NAT-ed peer.  
> One needs
> more than dynamic DNS to merge a self-organizing network with SIP
> functionality. No DNS server will be able to do parallel and serial
> forking, aggregate presence information, NAT traversal, buddy list
> synchronization among devices, redirect for not online to name a few
> things one needs in the day-to-day communications. DNS is good for
> bootstrap and and for providing global addressable URIs.

No DHT-based lookup mechanism does all those things either.

Those are all interesting problems that need solving that are independent of
the "how do we look up information that we don't have" problem. 

Interestingly, for things like "certificates" and "IP addresses", DNS
provides a convenient protocol for this *and*, if the answers given to
certain queries (like "what's an IP address of a thing that can act as a
proxy service for this domain") are compatible, allow for direct
interoperability with non-P2P SIP callers.

Building a thing that answers DNS queries out of a large number of nodes
that communicate using a DHT? That's certainly an option that some domains
might want to pursue, but I think it is silly to mandate a DHT all the way
down to the clients when there's (administrative) domains that might wish to
operate in other ways.

Matthew Kaufman
matthew at matthew.at


From matthew at matthew.at  Sun Dec  3 02:08:19 2006
From: matthew at matthew.at (Matthew Kaufman)
Date: Sat, 2 Dec 2006 23:08:19 -0800
Subject: [p2p-sip] Where are we "getting wrapped around the DNS axle"
In-Reply-To: <BA8A1D67-8623-497A-B105-51E6AFC7FB0D@softarmor.com>
Message-ID: <01b201c716a9$d1cbd8b0$029b0a0a@matthewdeskxps>

Dean Willis:
> 2) Using dynamic DNS as the registration/resolution mechanism for  
> finding nodes in an overlay. While this could work, it's just a  
> straightforward application of dynamic DNS and plain old SIP,  
> consequently requiring no new work. It also doesn't meet many of the  
> use cases given for P2PSIP.

I guess I still fail to see how dynamic DNS + zeroconf misses any of node
(and credential, if there are any) lookup problems presented by the use
cases given for P2PSIP. Meanwhile I do see how there's lots of other "P2P"
problems that *do* need solving for SIP.

Personally, I think that the focus on the DHT up front is a distraction from
those other "P2P" problems that need solving, and that standardizing on one
that's been shown to be insufficient for the actual use cases proposed will
be an additional distraction in the future, but -- given the "charter" and
the "working group" and all -- I'm also perfectly willing to sit on the
sidelines here and focus my energy on solving other interesting problems.


Matthew Kaufman
matthew at matthew.at


From dbarrett at quinthar.com  Sun Dec  3 03:11:35 2006
From: dbarrett at quinthar.com (David Barrett)
Date: Sun, 3 Dec 2006 00:11:35 -0800
Subject: [p2p-sip] Where are we "getting wrapped around the DNS axle"
In-Reply-To: <BA8A1D67-8623-497A-B105-51E6AFC7FB0D@softarmor.com>
Message-ID: <02e401c716b2$a6c86850$0202a8c0@Quinthar>

That's a great breakdown!  I love the bootstrapping logic you use; that
sounds like an excellent approach.

As for the others, I totally agree with what Matthew wrote, and I'd add:

> -----Original Message-----
> From: Dean Willis
> Subject: [p2p-sip] Where are we "getting wrapped around the DNS axle"
> 
> 3) The process of associating a hostname with an IP address in DNS is
> completely disjoint from the process of looking up a hostname in DNS
> to discover an IP address. That is, the lookup process is unrelated
> to and not dependent in any way on the configuration process.

I agree, DNS is wonderful for how universally standard it is to query, but
sucks for how nonstandard it is to configure.  The dynamic DNS guys have
made some good strides in automated DNS configuration on a limited scale
(the DynDNS client, for example, supports many different dynamic DNS
providers, not just DynDNS.com), but more work is required.

As such, while I'm receptive to the argument:

	"Even though DNS does a good job resolving an authoritative
hierarchical namespace into IP addresses, it's only appropriate for servers
because of the manual administration required."

My response would be:

	"Perhaps now is the time to automate that administration so we can
extend the existing DNS system to lookup clients and thus blur the line
between clients and servers (as is the hallmark of P2P)."

Indeed, I think there could be an excellent P2P-DNS sister group that
focuses on precisely this: making DNS useful for P2P networks (rather than
only client/server networks, as current).  P2P-SIP would be one of the major
customers of such a group.


> Things we MIGHT agree on:
> 
> 1) If DNS is available, we MAY use it. However, we presume that DNS
> is not always available, so we MUST NOT depend on it. For example, a
> Skype-like configuration where UAs are configured with a default
> rendezvous IP address should be workable.
> 
> 2) If multicast is available, we MAY use mDNS. However, we presume
> that multicast will not always be available, so we MUST NOT depend on
> mDNS in all cases.

Likewise, I agree that this is the consensus and charter of the group.  My
confusion is why we've made this iron-clad, inviolate decision when nobody
has been able to give any real-world example of any significant set of
real-world users who do not already have DNS, won't get DNS in short order,
and can't use mDNS in the meantime.  We've discussed the entire gamut from
the largest to smallest deployments, and DNS/mDNS fit them all.



> Things we have NO agreement on
> 
> 1) Mandating DNS as the only resolution method for discovering the IP
> address of an entry point to an overlay
> 
> 2) Using dynamic DNS as the registration/resolution mechanism for
> finding nodes in an overlay. While this could work, it's just a
> straightforward application of dynamic DNS and plain old SIP,
> consequently requiring no new work. It also doesn't meet many of the
> use cases given for P2PSIP.

On the first, I didn't mean to suggest DNS should be the only option.  If
you want to hard-code or hand-enter IP addresses, go right ahead.  I'm
merely saying that DNS is everywhere the users are, so it's silly not to use
it when convenient, and wasteful to spend so much energy figuring out how to
live without it when it's right there.

On the second, I'd like to hear about these use cases that can't be met by
using DNS/mDNS/dynamic-DNS to find the proxy for a given UA.


So again, I'm not saying a P2P-SIP protocol using an IP-bootstrapped DHT
wouldn't work.  I'm just saying it's a classic case of overengineering to
solve theoretical problems that don't exist in the real world.  In the real
world, DNS and central directories work exceedingly well.  Nobody would have
bothered "improving" upon the hybrid central/p2p model of Napster had the
RIAA not insisted upon it.  Indeed, Bittorrent has been wildly successful
precisely because it didn't try.

Thus it seems much more realistic to just:

a) Have every SIP address of the form 'name at domain'
b) Resolve 'domain' to the IP address of the proxy managing that address
c) Install regular SIP proxies configured with regular DNS wherever possible
d) Have a cloud of P2P-SIP proxies configured with dynamic-DNS for the rest
e) Have P2P-SIP clients respond to mDNS requests when DNS is unavailable
f) Have P2P-DNS focus on standardizing dynamic DNS and mDNS configuration
g) Have P2P-SIP focus on bridging SIP, P2P-DNS, and infrastructureless peers

Admittedly, it's not as sexy as a non-DNS solution.  It probably doesn't
need a DHT (the cloud of public proxies could be better managed with a
simple central directory).  But it seems to work for all of the P2P-SIP use
cases, and seems technically much simpler to do.

-david


From ag at ag-projects.com  Sun Dec  3 13:03:36 2006
From: ag at ag-projects.com (Adrian Georgescu)
Date: Sun, 3 Dec 2006 19:03:36 +0100
Subject: [p2p-sip] Where are we "getting wrapped around the DNS axle"
In-Reply-To: <BA8A1D67-8623-497A-B105-51E6AFC7FB0D@softarmor.com>
References: <BA8A1D67-8623-497A-B105-51E6AFC7FB0D@softarmor.com>
Message-ID: <D1BA94C8-2B26-4D50-82DB-6A60BBA3B2E5@ag-projects.com>

Dean,

What we miss is:

DNS like query/response mechanism for resources in the  P2P overlay.

Adrian


On Dec 3, 2006, at 7:18 AM, Dean Willis wrote:

>
> I think we're arguing at cross-purposes about DNS here.
>
> Let me see if I can clarify things a bit:
>
> Assumptions I think we agree on:
>
> 1) IF our rendezvous/bootstrapper/cert-server/whatever-we-start-with
> is listed in DNS, and our UA has a hostname for that whatever, THEN
> we MAY be able to use DNS to turn that hostname into an IP address.
>
> 2) If we're using DNS, as in #1, then we MAY use dynamic DNS to
> insert a record associating a hostname with an IP address. We MAY use
> conventional DNS configuration as an alternative.
>
> 3) The process of associating a hostname with an IP address in DNS is
> completely disjoint from the process of looking up a hostname in DNS
> to discover an IP address. That is, the lookup process is unrelated
> to and not dependent in any way on the configuration process.
>
>
> Things we MIGHT agree on:
>
> 1) If DNS is available, we MAY use it. However, we presume that DNS
> is not always available, so we MUST NOT depend on it. For example, a
> Skype-like configuration where UAs are configured with a default
> rendezvous IP address should be workable.
>
> 2) If multicast is available, we MAY use mDNS. However, we presume
> that multicast will not always be available, so we MUST NOT depend on
> mDNS in all cases.
>
>
> One MIGHT envision a resolution model that looks something like:
>
> IF configured with a rendezvous/bootstrap/...whatever that looks like
> an IP address,
>    THEN use that IP address
> ELSE
>    IF bootstrap looks like a hostname AND
>      IF DNS is available AND returns an answer
>        THEN use the address returned by DNS
>      ELSE
>        IF multicast is available AND returns an answer
>          THEN use the address returned by mDNS
>        ELSE bootstrap is broken, give up and die
>        END IF multicast ...
>     END IF DNS
>    ELSE bootstrap is broken, give up and die
>    END IF bootstrap ...
> END IF configured ...
>
>
>
> Things we have NO agreement on
>
> 1) Mandating DNS as the only resolution method for discovering the IP
> address of an entry point to an overlay
>
> 2) Using dynamic DNS as the registration/resolution mechanism for
> finding nodes in an overlay. While this could work, it's just a
> straightforward application of dynamic DNS and plain old SIP,
> consequently requiring no new work. It also doesn't meet many of the
> use cases given for P2PSIP.
>
>
> --
> dean
> _______________________________________________
> p2p-sip mailing list
> p2p-sip at cs.columbia.edu
> https://lists.cs.columbia.edu/cucslists/listinfo/p2p-sip


From bryan at ethernot.org  Sun Dec  3 14:40:19 2006
From: bryan at ethernot.org (David A. Bryan)
Date: Sun, 3 Dec 2006 14:40:19 -0500
Subject: [p2p-sip] Where are we "getting wrapped around the DNS axle"
In-Reply-To: <02e401c716b2$a6c86850$0202a8c0@Quinthar>
References: <BA8A1D67-8623-497A-B105-51E6AFC7FB0D@softarmor.com>
	<02e401c716b2$a6c86850$0202a8c0@Quinthar>
Message-ID: <8b2769930612031140y1be88b95n611187a17bf3d8aa@mail.gmail.com>

On 12/3/06, David Barrett <dbarrett at quinthar.com> wrote:
> Thus it seems much more realistic to just:
>
> a) Have every SIP address of the form 'name at domain'
> b) Resolve 'domain' to the IP address of the proxy managing that address
> c) Install regular SIP proxies configured with regular DNS wherever possible
> d) Have a cloud of P2P-SIP proxies configured with dynamic-DNS for the rest
> e) Have P2P-SIP clients respond to mDNS requests when DNS is unavailable
> f) Have P2P-DNS focus on standardizing dynamic DNS and mDNS configuration
> g) Have P2P-SIP focus on bridging SIP, P2P-DNS, and infrastructureless peers

The big question here is who is maintaining these proxies and dynamic
dns servers? This proposal isn't really any different than a
conventional SIP service in terms of the centralization model. Someone
still needs to provide the proxies and someone is providing the
dynamic DNS system, which basically starts to look like a centralized
registrar that someone has to own, you just moved it from speaking SIP
to speaking DNS.

I also still haven't heard a good response to handling mobility. I
want to be able to go from home to my mobile, drive 5 minutes, and
then go to the office, and want to be reachable at all points, which I
don't see a dynamic DNS approach doing.

David

From dbarrett at quinthar.com  Sun Dec  3 16:23:27 2006
From: dbarrett at quinthar.com (David Barrett)
Date: Sun, 3 Dec 2006 13:23:27 -0800
Subject: [p2p-sip] Where are we "getting wrapped around the DNS axle"
In-Reply-To: <8b2769930612031140y1be88b95n611187a17bf3d8aa@mail.gmail.com>
Message-ID: <035201c71721$46701a60$0202a8c0@Quinthar>

> -----Original Message-----
> From: David A. Bryan
> Subject: Re: [p2p-sip] Where are we "getting wrapped around the DNS axle"
> 
> The big question here is who is maintaining these proxies and dynamic
> dns servers? 

A lot of people, as would be expected from any highly distributed system:

- On one end of the spectrum, a domain owner (eg, @intel.com) can choose to
take total responsibility for all calls into that domain and just install a
central SIP server that works in a classic sense.

- On the other end of the spectrum, a random user can trust the kindness of
any number of benevolent "free" networks (each identified by its own domain,
like @p2psip.org or @myfreebiesip.com, with DNS used to find the
bootstrapping servers), leveraging the donated resources of volunteer nodes.

As for maintenance of specific systems:

- Standard SIP proxies would be maintained by either standard SIP providers
or large corporations (eg, intel.com might offer a robust SIP deployment for
its employees). 

- Dynamic DNS servers would be maintained by the dynamic DNS providers, of
which there are dozens (http://www.technopagan.org/dynamic/).  A Dynamic DNS
provider is nothing more than a standard DNS registrar that happens to
support an automated configuration interface and fast-propagating bindings.
I'd expect that in the future, *every* DNS registrar would support dynamic
DNS.

- Distributed SIP proxies (eg P2P-SIP proxies on non-NAT'd boxes registered
into some public cloud) would be maintained in the hardware sense by the
owners of the non-NAT'd boxes themselves.  However the "cloud" in a more
general sense is maintained by the bootstrapping services.  Some might
bootstrap into a DHT, others might just maintain a big central list.  Either
way, any kind soul with time and resources on his hand can register a
domain, create a new cloud (using protocols to be defined by P2P-SIP) and
make sure its good.

DNS is far from enslaving the world under central rule; it's the only system
with *truly* decentralized ownership, and I want that benefit in SIP.  


> This proposal isn't really any different than a
> conventional SIP service in terms of the centralization model. Someone
> still needs to provide the proxies and someone is providing the
> dynamic DNS system, which basically starts to look like a centralized
> registrar that someone has to own, you just moved it from speaking SIP
> to speaking DNS.

For users who like conventional SIP, yes, it's precisely no different, by
design.

For users who don't like conventional SIP they:
1) Get a domain from any of dozens of dynamic DNS providers
2) Find an existing bootstrapping service, or create a new one and host it
3) Enter the following into their P2P-SIP client:
	. Dynamic-DNS credentials (the only password they have)
	. Bootstrapping server IP address or domain name

After which their P2P-SIP client finds public resources using the chosen
bootstrapping service (either asking "give me a STUN server" directly or
"give me a DHT of STUN servers", depending on what the bootstrapping service
provides), registers with the SIP proxy built-into the P2P-SIP proxy of some
non-NAT'd client (or multiple, for redundancy), uses dynamic DNS to update
his record to point to his current SIP proxy(ies), and everything else just
works as normal.  To the outside world, a random, infrastructure-less
P2P-SIP client looks *exactly no different* than a totally client-server SIP
user.  Both have a normal SIP address that is resolved to a SIP proxy using
normal DNS, and whether that SIP proxy is a big server in the sky or a
random non-firewalled client, it makes no difference and nobody is the
wiser.


> I also still haven't heard a good response to handling mobility. I
> want to be able to go from home to my mobile, drive 5 minutes, and
> then go to the office, and want to be reachable at all points, which I
> don't see a dynamic DNS approach doing.

Yes, it can do that too.  Let's assume you had some device with two wireless
adapters that started connected to your home network (IPhome), then
connected to some kind of cellular network (IPcar) as you walked out the
door and dropped IPhome, and then picked up your work network (IPwork) as
you walked into the office and dropped IPcar.

DNS supports multiple IP addresses per domain.  To start, your P2P-SIP
client would register IPhome, and then it would for a brief period register
both IPhome and IPcar, and then when it went out of home range it'd register
only IPcar, and then as you got into the office it's register IPcar and
IPwork, and then when you went into the faraday cage (which we all have,
right?) it'd register just IPwork.  Regardless, every time you switch
networks, there are two options:

1) You have connectivity, and thus can update your DNS record
2) You do not have connectivity, and thus cannot update your DNS record

So long as you can connect to the internet, you can update the relevant
records to ensure people on the internet can call you.  And if you can't
connect to the internet, it's not a problem.

Furthermore, we haven't discussed the mDNS case where even if you walk into
a faraday cage and your company internet connection goes out, you can still
talk with everyone else in the office because they'll do an mDNS lookup for
your domain (using multicast over the LAN, which is very widely supported)
and will find you just fine.

Again, DNS, dynamic-DNS, and mDNS handle *all* these cases as good as
anything else, and likely much better.  I apologize that it's not sexy, but
it's entirely adequate and indeed probably superior. 

There are still plenty of challenges to work out and details to standardize,
and the P2P-SIP WG is ideally suited to do it.  Sure it's not inventing any
new class of fun DHT technology (or standardizing an antiquated, obsolete
DHT), but the IETF isn't supposed to invent: it's supposed to standardize
what is already there.  DNS, dynamic DNS, mDNS -- they're here and here to
stay.  P2P-SIP should be figuring out and standardizing best practices, not
inventing replacements.

-david


From bryan at ethernot.org  Sun Dec  3 22:21:33 2006
From: bryan at ethernot.org (David A. Bryan)
Date: Sun, 3 Dec 2006 22:21:33 -0500
Subject: [p2p-sip] Where are we "getting wrapped around the DNS axle"
In-Reply-To: <035201c71721$46701a60$0202a8c0@Quinthar>
References: <8b2769930612031140y1be88b95n611187a17bf3d8aa@mail.gmail.com>
	<035201c71721$46701a60$0202a8c0@Quinthar>
Message-ID: <8b2769930612031921y6cc68183h802df07893d4fdee@mail.gmail.com>

I still fail to see how having volunteer proxies is P2P. Just because
"a lot people" are participating, it isn't necessarily decentralized,
you've simply moved the central people you are dependant on from being
a company to be one or more volunteers. The individual users using the
service aren't participating (at least not all of them) in providing
the services, they are only using them. You are still relying on
centralized devices that are maintained and provided by others, in
this case, some group of volunteer proxies and some dynamic DNS
servers that may or may not be provided on a volunteer basis. Such a
system is thus susceptible to their whims and desires of those
providing it to stop at any time.

I simply don't see that it is a truly decentralized system, where one
(or a few) entities deciding to stop volunteering or being unable to
volunteer can mean the system fails. Nothing wrong with it of course,
it could work, and certainly would be interesting work to go do (as
others pointed out, the work could likely be submitted as a draft to
SIPPING) I don't think it is really in line with the type of system
that at least many of us who have been working to get this group off
the ground really had in mind.

There certainly are also scenarios (the impeded access, security
conscious small organization, and anonymous communications -- all of
3.2 in draft-bryan-sipping-p2p-usecases seem to fall into that boat)
where I can't see the approach you propose working, either.

I'm sorry I was a bit unclear with my question about how to do
mobility. It is more of a question of handling the update period in a
cached DNS-like system with sufficient speed. 5 minutes is a
reasonable time period to change a phone location, but I am not at all
convinced that conventional dynamic DNS as it is today can push
updates that fast, or it would scale if you tried. DNS does a heck of
a lot of caching. Not sure, but I would think there would be stale
caches causing all sorts of problems. It could work very well for a
permanent or semi-permanent desk phone of course, just not practical
in all the scenarios we have discussed, at least in my mind.

David

On 12/3/06, David Barrett <dbarrett at quinthar.com> wrote:
> > -----Original Message-----
> > From: David A. Bryan
> > Subject: Re: [p2p-sip] Where are we "getting wrapped around the DNS axle"
> >
> > The big question here is who is maintaining these proxies and dynamic
> > dns servers?
>
> A lot of people, as would be expected from any highly distributed system:
>
> - On one end of the spectrum, a domain owner (eg, @intel.com) can choose to
> take total responsibility for all calls into that domain and just install a
> central SIP server that works in a classic sense.
>
> - On the other end of the spectrum, a random user can trust the kindness of
> any number of benevolent "free" networks (each identified by its own domain,
> like @p2psip.org or @myfreebiesip.com, with DNS used to find the
> bootstrapping servers), leveraging the donated resources of volunteer nodes.
>
> As for maintenance of specific systems:
>
> - Standard SIP proxies would be maintained by either standard SIP providers
> or large corporations (eg, intel.com might offer a robust SIP deployment for
> its employees).
>
> - Dynamic DNS servers would be maintained by the dynamic DNS providers, of
> which there are dozens (http://www.technopagan.org/dynamic/).  A Dynamic DNS
> provider is nothing more than a standard DNS registrar that happens to
> support an automated configuration interface and fast-propagating bindings.
> I'd expect that in the future, *every* DNS registrar would support dynamic
> DNS.
>
> - Distributed SIP proxies (eg P2P-SIP proxies on non-NAT'd boxes registered
> into some public cloud) would be maintained in the hardware sense by the
> owners of the non-NAT'd boxes themselves.  However the "cloud" in a more
> general sense is maintained by the bootstrapping services.  Some might
> bootstrap into a DHT, others might just maintain a big central list.  Either
> way, any kind soul with time and resources on his hand can register a
> domain, create a new cloud (using protocols to be defined by P2P-SIP) and
> make sure its good.
>
> DNS is far from enslaving the world under central rule; it's the only system
> with *truly* decentralized ownership, and I want that benefit in SIP.
>
>
> > This proposal isn't really any different than a
> > conventional SIP service in terms of the centralization model. Someone
> > still needs to provide the proxies and someone is providing the
> > dynamic DNS system, which basically starts to look like a centralized
> > registrar that someone has to own, you just moved it from speaking SIP
> > to speaking DNS.
>
> For users who like conventional SIP, yes, it's precisely no different, by
> design.
>
> For users who don't like conventional SIP they:
> 1) Get a domain from any of dozens of dynamic DNS providers
> 2) Find an existing bootstrapping service, or create a new one and host it
> 3) Enter the following into their P2P-SIP client:
>         . Dynamic-DNS credentials (the only password they have)
>         . Bootstrapping server IP address or domain name
>
> After which their P2P-SIP client finds public resources using the chosen
> bootstrapping service (either asking "give me a STUN server" directly or
> "give me a DHT of STUN servers", depending on what the bootstrapping service
> provides), registers with the SIP proxy built-into the P2P-SIP proxy of some
> non-NAT'd client (or multiple, for redundancy), uses dynamic DNS to update
> his record to point to his current SIP proxy(ies), and everything else just
> works as normal.  To the outside world, a random, infrastructure-less
> P2P-SIP client looks *exactly no different* than a totally client-server SIP
> user.  Both have a normal SIP address that is resolved to a SIP proxy using
> normal DNS, and whether that SIP proxy is a big server in the sky or a
> random non-firewalled client, it makes no difference and nobody is the
> wiser.
>
>
> > I also still haven't heard a good response to handling mobility. I
> > want to be able to go from home to my mobile, drive 5 minutes, and
> > then go to the office, and want to be reachable at all points, which I
> > don't see a dynamic DNS approach doing.
>
> Yes, it can do that too.  Let's assume you had some device with two wireless
> adapters that started connected to your home network (IPhome), then
> connected to some kind of cellular network (IPcar) as you walked out the
> door and dropped IPhome, and then picked up your work network (IPwork) as
> you walked into the office and dropped IPcar.
>
> DNS supports multiple IP addresses per domain.  To start, your P2P-SIP
> client would register IPhome, and then it would for a brief period register
> both IPhome and IPcar, and then when it went out of home range it'd register
> only IPcar, and then as you got into the office it's register IPcar and
> IPwork, and then when you went into the faraday cage (which we all have,
> right?) it'd register just IPwork.  Regardless, every time you switch
> networks, there are two options:
>
> 1) You have connectivity, and thus can update your DNS record
> 2) You do not have connectivity, and thus cannot update your DNS record
>
> So long as you can connect to the internet, you can update the relevant
> records to ensure people on the internet can call you.  And if you can't
> connect to the internet, it's not a problem.
>
> Furthermore, we haven't discussed the mDNS case where even if you walk into
> a faraday cage and your company internet connection goes out, you can still
> talk with everyone else in the office because they'll do an mDNS lookup for
> your domain (using multicast over the LAN, which is very widely supported)
> and will find you just fine.
>
> Again, DNS, dynamic-DNS, and mDNS handle *all* these cases as good as
> anything else, and likely much better.  I apologize that it's not sexy, but
> it's entirely adequate and indeed probably superior.
>
> There are still plenty of challenges to work out and details to standardize,
> and the P2P-SIP WG is ideally suited to do it.  Sure it's not inventing any
> new class of fun DHT technology (or standardizing an antiquated, obsolete
> DHT), but the IETF isn't supposed to invent: it's supposed to standardize
> what is already there.  DNS, dynamic DNS, mDNS -- they're here and here to
> stay.  P2P-SIP should be figuring out and standardizing best practices, not
> inventing replacements.
>
> -david
>
>

From dbarrett at quinthar.com  Mon Dec  4 04:48:29 2006
From: dbarrett at quinthar.com (David Barrett)
Date: Mon, 4 Dec 2006 01:48:29 -0800
Subject: [p2p-sip] Where are we "getting wrapped around the DNS axle"
In-Reply-To: <8b2769930612031921y6cc68183h802df07893d4fdee@mail.gmail.com>
Message-ID: <03c901c71789$5c6fd5e0$0202a8c0@Quinthar>

> -----Original Message-----
> From: David A. Bryan
> Subject: Re: [p2p-sip] Where are we "getting wrapped around the DNS axle"
> 
> I still fail to see how having volunteer proxies is P2P. ... The
> individual users using the
> service aren't participating (at least not all of them) in providing
> the services, they are only using them.

I'm sorry, I was unclear -- I assumed it was apparent the P2P-SIP client
would "auto-volunteer" anyone who was able (ie, not NAT'd or firewalled).  I
agree, a system where people must explicitly "opt-in" will probably fail.
So please rethink my proposal with the new assumption that "all P2P-SIP
clients will auto-volunteer to one or more clouds".

To clarify, I'm using the term "volunteer" to describe the classic P2P
relationship where you donate resources for the use of others, in exchange
for others to donate them when you need them.  If there's a better, more
intuitive term for this, please let me know.  

I'm also using the term "cloud" to describe "a group of P2P-SIP clients who
cooperatively volunteer resources".  I've assumed there will be multiple
"clouds" of volunteer nodes, and that anybody can start a new one by just
putting up a bootstrapping server and asking people type that IP/domain into
their client.  However, if it helps things for the sake of argument, we can
simplify the model and assume there's only one cloud, only one bootstrapping
service (provided by p2psip.org), and every P2P-SIP client is hard-coded to
connect to it on startup.


> You are still relying on
> centralized devices that are maintained and provided by others, in
> this case, some group of volunteer proxies and some dynamic DNS
> servers that may or may not be provided on a volunteer basis. Such a
> system is thus susceptible to their whims and desires of those
> providing it to stop at any time.

Hm, this seems to be overstating it:

If you are behind a NAT or firewall, you only need to find at least one
bootstrapping service that can provide you with at least one "volunteer"
node.  Given that there can be any number of bootstrapping services (aka
"clouds"), and each of these can contain any number of nodes (aka
"volunteers") -- and that any node can and probably should volunteer into
multiple clouds -- it'd take a rather large conspiracy to deny you service.
Indeed, how is this any different than bootstrapping into a DHT?

(And if you are not NAT'd and not firewalled then you don't need a
bootstrapping service and don't need any other volunteers, though I'd expect
the client would be written to have you in turn volunteer for others.)

Furthermore, yes, I'm assuming whoever you registered your DNS name with
would continue to maintain it, else you'd move it to a more reliable
provider.  This doesn't seem at all an unreasonable or unusual demand.  With
few exceptions, DNS providers generally aren't capricious.


> I simply don't see that it is a truly decentralized system, where one
> (or a few) entities deciding to stop volunteering or being unable to
> volunteer can mean the system fails.

Have I addressed your concerns here with the above clarifications?


> There certainly are also scenarios (the impeded access, security
> conscious small organization, and anonymous communications -- all of
> 3.2 in draft-bryan-sipping-p2p-usecases seem to fall into that boat)
> where I can't see the approach you propose working, either.

Well, in the absolute, no human technology (including DNS) can achieve that
level of security -- it's like requiring total invisibility to the NSA and
total penetration of the great firewall of China, now and forever.

But while they are impossible requirements in the absolute, it's possible to
be "good enough" in each respect.  Unfortunately, "good enough" isn't well
defined at this point, so it's hard to argue whether DNS can or cannot meet
that actual standard (versus the absolute ideal presented in your draft).
Tell me to what degree you intend to satisfy these requirements in a non-DNS
world, and then we can figure out if DNS can meet or exceed that level.


> I'm sorry I was a bit unclear with my question about how to do
> mobility. It is more of a question of handling the update period in a
> cached DNS-like system with sufficient speed. 5 minutes is a
> reasonable time period to change a phone location, but I am not at all
> convinced that conventional dynamic DNS as it is today can push
> updates that fast, or it would scale if you tried. DNS does a heck of
> a lot of caching. Not sure, but I would think there would be stale
> caches causing all sorts of problems.

Set the TTL to be near-zero and all the caches will expire, causing DNS to
resolve your name from scratch every time.  It'd be inefficient to do that
for a high-volume webserver that never changes IPs, but it's totally
sensible to do that for an extremely low-volume P2P-SIP client that changes
IPs all the time.

Akamai does this.  UltraDNS does this.  This isn't some crazy theory that
pushes DNS in a totally new and explored area.  Lots of people are doing
this today.


So in summary, I've tried to address the following concerns:

Concern: Opt-in volunteering won't work.
Response: Not opt-in; everyone who can volunteers by default.

Concern: Still relying upon others.
Response: It relies upon DNS providers to provide DNS.  Otherwise it relies
on volunteers no more or less than DHT.

Concern: Can't achieve absolute security.
Response: Nothing can.  However, DNS can give "good enough" security.

Concern: DNS can't propagate fast enough.
Response: Set tiny TTLs and it can.

Were there other concerns that I missed?  Are the above responses
inadequate?

Matthew -- You know a lot more about the inner workings of DNS than me; did
I misstate anything?

-david


From enrico.marocco at telecomitalia.it  Mon Dec  4 05:26:01 2006
From: enrico.marocco at telecomitalia.it (Enrico Marocco)
Date: Mon, 04 Dec 2006 11:26:01 +0100
Subject: [p2p-sip] Where are we "getting wrapped around the DNS axle"
In-Reply-To: <03c901c71789$5c6fd5e0$0202a8c0@Quinthar>
References: <03c901c71789$5c6fd5e0$0202a8c0@Quinthar>
Message-ID: <4573F7B9.2080305@telecomitalia.it>

David Barrett wrote:
> Furthermore, yes, I'm assuming whoever you registered your DNS name with
> would continue to maintain it, else you'd move it to a more reliable
> provider.  This doesn't seem at all an unreasonable or unusual demand.  With
> few exceptions, DNS providers generally aren't capricious.

That's the whole point.  We have been figuring a system that doesn't
depend on any provider since the very beginning.  If some are available,
it can make use of them (e.g. for bootstrap), otherwise it must work
anyway, even if with limited (but still required, at least in some
"uncommon" scenarios) functionalities.

This was the very first assumption and unfortunately a solution which
depends on a service "provided" by someone doesn't meet it.  Nobody's
asserting that such a solution wouldn't be better than p2p-sip; it's
just that it isn't p2p-sip.  Period.

-- 
Ciao,
Enrico

From hsinnrei at adobe.com  Mon Dec  4 10:21:31 2006
From: hsinnrei at adobe.com (Henry Sinnreich)
Date: Mon, 4 Dec 2006 07:21:31 -0800
Subject: [p2p-sip] Where are we "getting wrapped around the DNS axle"
In-Reply-To: <4573F7B9.2080305@telecomitalia.it>
Message-ID: <24CCCC428EFEA2469BF046DB3C7A8D221413A5@namail5.corp.adobe.com>

>We have been figuring a system that doesn't depend on any provider
since >the very beginning.  If some are available, it can make use of
them (e.g. >for bootstrap), otherwise it must work anyway

This is a very powerful model shift indeed.
Such a P2P SIP enabled devices could have multiple accounts, so that
users could configure it for their own community networks ("providers"
if you will), but also use providers of _their own choice_.

Maybe the overlay Internet, the openDHT can interconnect all P2P SIP
devices for an access fee, similar to the access fee paid today for
broadband IP. Think of Overlay Access Service (OAS) similar to the IP
broadband access existing today.

If the above gets implemented, there will be a huge market for P2P SIP
communication devices, based on a common global standard and on open
communications under user control. 

Such devices are the opposite of today's locked down mobile phones that
work only for one single provider.

Don't expect though this new business model to have smooth sailing...

Thanks, Henry

-----Original Message-----
From: p2p-sip-bounces at cs.columbia.edu
[mailto:p2p-sip-bounces at cs.columbia.edu] On Behalf Of Enrico Marocco
Sent: Monday, December 04, 2006 4:26 AM
To: David Barrett
Cc: 'P2P-SIP'; 'David A. Bryan'
Subject: Re: [p2p-sip] Where are we "getting wrapped around the DNS
axle"

David Barrett wrote:
> Furthermore, yes, I'm assuming whoever you registered your DNS name
with
> would continue to maintain it, else you'd move it to a more reliable
> provider.  This doesn't seem at all an unreasonable or unusual demand.
With
> few exceptions, DNS providers generally aren't capricious.

That's the whole point.  We have been figuring a system that doesn't
depend on any provider since the very beginning.  If some are available,
it can make use of them (e.g. for bootstrap), otherwise it must work
anyway, even if with limited (but still required, at least in some
"uncommon" scenarios) functionalities.

This was the very first assumption and unfortunately a solution which
depends on a service "provided" by someone doesn't meet it.  Nobody's
asserting that such a solution wouldn't be better than p2p-sip; it's
just that it isn't p2p-sip.  Period.

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


From hsinnrei at adobe.com  Mon Dec  4 10:21:25 2006
From: hsinnrei at adobe.com (Henry Sinnreich)
Date: Mon, 4 Dec 2006 07:21:25 -0800
Subject: [p2p-sip] Where are we "getting wrapped around the DNS axle"
In-Reply-To: <D1BA94C8-2B26-4D50-82DB-6A60BBA3B2E5@ag-projects.com>
Message-ID: <24CCCC428EFEA2469BF046DB3C7A8D221413A6@namail5.corp.adobe.com>

>What we miss is:

>DNS like query/response mechanism for resources in the P2P overlay.

DNS like query/response is not missing if there is a primary bootstrap
mechanism, such as the service mode in the I-D by K. Singh and H.
Schulzrinne. This fits into Dean Willis model, which is right:

IF configured with a rendezvous/bootstrap/...whatever that looks like an
IP address,
   THEN use that IP address
ELSE
   IF bootstrap looks like a hostname AND
     IF DNS is available AND returns an answer
       THEN use the address returned by DNS
     ELSE
       IF multicast is available AND returns an answer
         THEN use the address returned by mDNS
       ELSE bootstrap is broken, give up and die
       END IF multicast ...
    END IF DNS
   ELSE bootstrap is broken, give up and die
   END IF bootstrap ...
END IF configured ...

Henry

-----Original Message-----
From: p2p-sip-bounces at cs.columbia.edu
[mailto:p2p-sip-bounces at cs.columbia.edu] On Behalf Of Adrian Georgescu
Sent: Sunday, December 03, 2006 12:04 PM
To: Dean Willis
Cc: P2P-SIP
Subject: Re: [p2p-sip] Where are we "getting wrapped around the DNS
axle"

Dean,

What we miss is:

DNS like query/response mechanism for resources in the  P2P overlay.

Adrian


On Dec 3, 2006, at 7:18 AM, Dean Willis wrote:

>
> I think we're arguing at cross-purposes about DNS here.
>
> Let me see if I can clarify things a bit:
>
> Assumptions I think we agree on:
>
> 1) IF our rendezvous/bootstrapper/cert-server/whatever-we-start-with
> is listed in DNS, and our UA has a hostname for that whatever, THEN
> we MAY be able to use DNS to turn that hostname into an IP address.
>
> 2) If we're using DNS, as in #1, then we MAY use dynamic DNS to
> insert a record associating a hostname with an IP address. We MAY use
> conventional DNS configuration as an alternative.
>
> 3) The process of associating a hostname with an IP address in DNS is
> completely disjoint from the process of looking up a hostname in DNS
> to discover an IP address. That is, the lookup process is unrelated
> to and not dependent in any way on the configuration process.
>
>
> Things we MIGHT agree on:
>
> 1) If DNS is available, we MAY use it. However, we presume that DNS
> is not always available, so we MUST NOT depend on it. For example, a
> Skype-like configuration where UAs are configured with a default
> rendezvous IP address should be workable.
>
> 2) If multicast is available, we MAY use mDNS. However, we presume
> that multicast will not always be available, so we MUST NOT depend on
> mDNS in all cases.
>
>
> One MIGHT envision a resolution model that looks something like:
>
> IF configured with a rendezvous/bootstrap/...whatever that looks like
> an IP address,
>    THEN use that IP address
> ELSE
>    IF bootstrap looks like a hostname AND
>      IF DNS is available AND returns an answer
>        THEN use the address returned by DNS
>      ELSE
>        IF multicast is available AND returns an answer
>          THEN use the address returned by mDNS
>        ELSE bootstrap is broken, give up and die
>        END IF multicast ...
>     END IF DNS
>    ELSE bootstrap is broken, give up and die
>    END IF bootstrap ...
> END IF configured ...
>
>
>
> Things we have NO agreement on
>
> 1) Mandating DNS as the only resolution method for discovering the IP
> address of an entry point to an overlay
>
> 2) Using dynamic DNS as the registration/resolution mechanism for
> finding nodes in an overlay. While this could work, it's just a
> straightforward application of dynamic DNS and plain old SIP,
> consequently requiring no new work. It also doesn't meet many of the
> use cases given for P2PSIP.
>
>
> --
> dean
> _______________________________________________
> p2p-sip mailing list
> p2p-sip at cs.columbia.edu
> https://lists.cs.columbia.edu/cucslists/listinfo/p2p-sip

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


From matthew at matthew.at  Mon Dec  4 11:35:31 2006
From: matthew at matthew.at (Matthew Kaufman)
Date: Mon, 4 Dec 2006 08:35:31 -0800
Subject: [p2p-sip] Where are we "getting wrapped around the DNS axle"
In-Reply-To: <4573F7B9.2080305@telecomitalia.it>
Message-ID: <003701c717c2$3700c4c0$029b0a0a@matthewdeskxps>

Enrico Marocco:
>
> That's the whole point.  We have been figuring a system that doesn't
> depend on any provider since the very beginning.  If some are available,
> it can make use of them (e.g. for bootstrap), otherwise it must work
> anyway, even if with limited (but still required, at least in some
> "uncommon" scenarios) functionalities.

I would argue that doing a DNS-protocol lookup given the fixed IP addresses
of the canonical root DNS servers (and, if you like, the fixed IP addresses
of some non-traditional root DNS servers) is *exactly the same* as doing a
DHT-protocol lookup given the fixed IP addresses of some seed nodes for the
DHT. And I would argue that if one can find the node at which an answer
would be stored, one can send updated information to that node as often as
bandwidth permits.

In both cases a new client requires the IP address of a bootstrap node, and
in both cases that information can redirect you to a (possibly infinite)
number of other nodes that actually contain the information.

The DNS solution requires more clever thought for the highly-distributed
cases, but can be set up today to just work for the less-distributed cases,
of which there are many. It also means that work on solving all the other
interesting P2P problems could start immediately.

The DHT solution requires more work on the part of the DHT-participating
clients to join the DHT and then requires a DHT algorithm that is stable in
the face of the expected level of client churn in order to keep providing
correct answers (of which Chord is almost certainly not an example).

And in both cases, if the configured fixed IP addresses are unreachable,
then the only way to find other clients on the LAN is multicast. Zeroconf
provides its mDNS mechanism for doing this in a DNS-like way, which is a
reasonable standard to adopt for this use case.

What it really comes down to is that when I hear "P2P SIP" I think of the
actual P2P problems with the SIP part, the P2P *signaling* issues... not the
P2P problems of node discovery and lookup. If I know the IP address of a SIP
client that my buddy is using and a way to traverse the NAT it is behind (if
any), how can I make that client ring if there's no proxy to talk to? What
if I know the IP addresses of her three SIP clients... can I make all three
of them ring without involving a proxy? Can I use SIMPLE to get presence
information from all three of them directly? In general, can I replace the
trapezoid with a straight line from my client to her client and still get
all (or most of) the features I'd like from SIP?

Neither DNS nor a DHT provides solutions to these "Peer to Peer Session
Initiation" problems... they're just two different ways of tackling the
"given a key like a name, how do I get a (list of) value(s) like an IP
address or certificate" problem.

I agree that my client being able to look things up about another client is
a precondition of doing "P2P SIP" with that client, but I don't even think
that the lookup itself *is* "P2P SIP", much less that "P2P SIP" should
mandate a non-standard mechanism for doing it.

Matthew Kaufman
matthew at matthew.at



From drage at alcatel-lucent.com  Mon Dec  4 13:25:44 2006
From: drage at alcatel-lucent.com (Drage, Keith (Keith))
Date: Mon, 4 Dec 2006 19:25:44 +0100
Subject: [p2p-sip] Fwd: what's wrong with DNS?
In-Reply-To: <02ba01c7133a$980b5960$0200a8c0@Quinthar>
Message-ID: <5D1A7985295922448D5550C94DE29180919334@DEEXC1U01.de.lucent.com>

To the best of my knowledge, the draft charter encompasses anything that
SIP can do.

It discusses nothing about the application of SIP, merely provides an
alternative mechanisms for reaching the address of record in SIP.

What SIP can do is of course in the purview of the SIP or SIPPING WG.

Regards

Keith 

> -----Original Message-----
> From: p2p-sip-bounces at cs.columbia.edu 
> [mailto:p2p-sip-bounces at cs.columbia.edu] On Behalf Of David Barrett
> Sent: 28 November 2006 22:15
> To: 'Sean Olson'; 'Henry Sinnreich'; 'Brian Rosen'; 'Matthew 
> Kaufman'; slavitch at gmail.com
> Cc: p2p-sip at cs.columbia.edu; 'Kundan Singh'
> Subject: Re: [p2p-sip] Fwd: what's wrong with DNS?
> 
> Ah, perhaps this is my mistake.  Isn't it limited to the 
> general class of audio/video/realtime communications?  For 
> example, is background swarming file-transfers out of scope?  
> I apologize; I clearly don't know what I'm talking about here 
> regarding scope, I'm sorry for the confusion.
> 
> To take it from another angle, what is out of scope?  Are we 
> competing head to head with UPnP for generalized resource 
> advertisement and discovery?
> 
> -david
> 
> > -----Original Message-----
> > From: Sean Olson [mailto:seanol at exchange.microsoft.com]
> > Sent: Tuesday, November 28, 2006 1:57 PM
> > To: David Barrett; 'Henry Sinnreich'; 'Brian Rosen'; 'Matthew 
> > Kaufman'; slavitch at gmail.com
> > Cc: p2p-sip at cs.columbia.edu; 'Kundan Singh'
> > Subject: RE: [p2p-sip] Fwd: what's wrong with DNS?
> > 
> > I don't understand -- we didn't limit the charter of the SIP WG to 
> > just voice. Why would it be different for P2PSIP?  Why 
> would you get 
> > less service from the P2P network than a traditional trapezoid? IM, 
> > file transfer, video, etc. are all reasonable modalities 
> for P2P. Am I 
> > missing something that would make these harder than voice?
> > 
> > For multicast, the issue is not technical support but admin 
> enablement.
> > Some do; some don't. I've got nothing against using 
> multicast as one 
> > option. But if it is the only option, that would be a deployment 
> > blocker in many real-world customers.
> > 
> > 
> > 
> > -----Original Message-----
> > From: David Barrett [mailto:dbarrett at quinthar.com]
> > Sent: Tuesday, November 28, 2006 1:53 PM
> > To: Sean Olson; 'Henry Sinnreich'; 'Brian Rosen'; 'Matthew 
> Kaufman'; 
> > slavitch at gmail.com
> > Cc: p2p-sip at cs.columbia.edu; 'Kundan Singh'
> > Subject: RE: [p2p-sip] Fwd: what's wrong with DNS?
> > 
> > Incidentally, I believe ad hoc networks *do* support multicast.  
> > Indeed, multicast on the LAN is actually quite common (it's 
> multicast 
> > on the WAN that's generally not available).
> > 
> > Furthermore, it's my impression that the scope of our charter *is* 
> > limited to voice.  If we don't draw the line there, then 
> where *do* we 
> > draw the line?
> > 
> > -david
> > 
> > > -----Original Message-----
> > > From: Sean Olson [mailto:seanol at exchange.microsoft.com]
> > > Sent: Tuesday, November 28, 2006 1:45 PM
> > > To: David Barrett; 'Henry Sinnreich'; 'Brian Rosen'; 'Matthew 
> > > Kaufman'; slavitch at gmail.com
> > > Cc: p2p-sip at cs.columbia.edu; 'Kundan Singh'
> > > Subject: RE: [p2p-sip] Fwd: what's wrong with DNS?
> > >
> > > I'm not advocating limiting this to voice. mDNS is great 
> except when 
> > > multicast is not available. But my real question is not so much 
> > > about technology but the scenario.
> > >
> > > Is this a scenario that is being considered?
> > >
> > >
> > > -----Original Message-----
> > > From: David Barrett [mailto:dbarrett at quinthar.com]
> > > Sent: Tuesday, November 28, 2006 1:43 PM
> > > To: Sean Olson; 'Henry Sinnreich'; 'Brian Rosen'; 
> 'Matthew Kaufman'; 
> > > slavitch at gmail.com
> > > Cc: p2p-sip at cs.columbia.edu; 'Kundan Singh'
> > > Subject: RE: [p2p-sip] Fwd: what's wrong with DNS?
> > >
> > > If they're in the same room, why do they need VoIP?  The range of 
> > > wifi isn't much wider than the range of the spoken voice.
> > >
> > > Even if they did somehow have wifi but not voice range, mDNS (as 
> > > Matthew has repeatedly pointed out) is exactly suited for this 
> > > problem.  If they
> > want
> > > VoIP, they'll probably also want file transfers, IM, and all the 
> > > other accoutrements of internet life.  Why would we have one 
> > > solution for
> > VoIP,
> > > and use something totally different for everything else?
> > >
> > > Why not just make mDNS really easy to use, and then use that for 
> > > these fringe VoIP cases?
> > >
> > > -david
> > >
> > > > -----Original Message-----
> > > > From: Sean Olson [mailto:seanol at exchange.microsoft.com]
> > > > Sent: Tuesday, November 28, 2006 1:36 PM
> > > > To: David Barrett; 'Henry Sinnreich'; 'Brian Rosen'; 'Matthew
> > Kaufman';
> > > > slavitch at gmail.com
> > > > Cc: p2p-sip at cs.columbia.edu; 'Kundan Singh'
> > > > Subject: RE: [p2p-sip] Fwd: what's wrong with DNS?
> > > >
> > > > It would be nice to support an ad-hoc WiFi network 
> among peers and 
> > > > use P2PSIP for voice, chat, etc. This is a scenario 
> where DNS does 
> > > > not
> > exist
> > > > necessarily and it is actually fairly common.  Take for 
> example, a 
> > > > set
> > > of
> > > > consultants working on site with a customer. They 
> probably don't 
> > > > have direct access to the customer's network and maybe 
> not to the 
> > > > broader Internet either. They will however likely have 
> WiFi and be 
> > > > able to
> > setup
> > > a
> > > > small ad-hoc network amongst themselves.
> > > >
> > > > Is this scenario out-of-scope for this WG?
> > > >
> > > >
> > > > -----Original Message-----
> > > > From: p2p-sip-bounces at cs.columbia.edu [mailto:p2p-sip- 
> > > > bounces at cs.columbia.edu] On Behalf Of David Barrett
> > > > Sent: Tuesday, November 28, 2006 1:30 PM
> > > > To: 'Henry Sinnreich'; 'Brian Rosen'; 'Matthew Kaufman'; 
> > > > slavitch at gmail.com
> > > > Cc: p2p-sip at cs.columbia.edu; 'Kundan Singh'
> > > > Subject: Re: [p2p-sip] Fwd: what's wrong with DNS?
> > > >
> > > > > -----Original Message-----
> > > > > From: Henry Sinnreich [mailto:hsinnrei at adobe.com]
> > > > > Subject: RE: [p2p-sip] Fwd: what's wrong with DNS?
> > > > >
> > > > > BTW: Existing P2P traffic is independent of the DNS and it 
> > > > > dominates
> > > the
> > > > > Internet traffic. Numbers between 60% and 80% have 
> been published.
> > > >
> > > > I apologize, but I think that's just simply false and 
> misleading.
> > > >
> > > > First, virtually all of that traffic is file-sharing traffic 
> > > > designed
> > to
> > > > preserve anonymity, which is the direct opposite of our problem.
> > Using
> > > > that
> > > > as justification to keep our system "clean" of DNS is a 
> red herring.
> > > >
> > > > But even ignoring that, most of that traffic is 
> Bittorrent, which
> > *does*
> > > > use
> > > > DNS not only for search (to find torrents on btjunkie.org and so
> > forth),
> > > > but
> > > > to find peers (to contact the tracker).
> > > >
> > > > At the end of the day, DNS is a fundamental part of the 
> internet, 
> > > > used
> > > by
> > > > virtually very user not only ever day, but almost every 
> minute.  
> > > > If
> > DNS
> > > > went
> > > > down globally, the vast majority of internet traffic would also 
> > > > slow
> > to
> > > a
> > > > stop (web, email, bittorrent, etc).  There would be a few 
> > > > survivors -- Skype with its hard-coded IPs would 
> persevere, along 
> > > > with eDonkey, Kazaa and other anonymous or totally 
> closed/private 
> > > > services, but about everything
> > else
> > > > would grind to a halt.  Planning on how to keep working without 
> > > > DNS is like planning on how to work without power -- you just 
> > > > don't.  If DNS goes down, we have bigger problems to 
> worry about.
> > > >
> > > > We certainly *can* reinvent the wheel and figure out how to keep
> > working
> > > > in
> > > > this apocalypse scenario, I just don't see why we should bother.
> > > >
> > > > -david
> > > >
> > > > _______________________________________________
> > > > p2p-sip mailing list
> > > > p2p-sip at cs.columbia.edu
> > > > https://lists.cs.columbia.edu/cucslists/listinfo/p2p-sip
> 
> _______________________________________________
> p2p-sip mailing list
> p2p-sip at cs.columbia.edu
> https://lists.cs.columbia.edu/cucslists/listinfo/p2p-sip
> 


From dbarrett at quinthar.com  Mon Dec  4 13:30:18 2006
From: dbarrett at quinthar.com (David Barrett)
Date: Mon, 4 Dec 2006 10:30:18 -0800
Subject: [p2p-sip] Where are we "getting wrapped around the DNS axle"
In-Reply-To: <4573F7B9.2080305@telecomitalia.it>
Message-ID: <046701c717d2$40c98c70$0202a8c0@Quinthar>

I think this is the mental leap I haven't been able to make.  "Doesn't
depend on any provider" seems impossible.  At the very least, you need
someone to "provide" network access and electricity.  But don't we also need
someone to "provide" a namespace?

In other words, do you agree:

	1) P2P-SIP must guarantee that each name resolves to a single user

	2) Multiple users want the same name

If both of those are answered yes, then how can we avoid a "namespace
provider"?  DNS already does this today on a massive scale, and is available
everywhere real users are.  Why are we so desperate to avoid it?

What is the alternative namespace provider being presented, and why is that
preferable?  Remember: even certificate authorities require someone to
maintain the CA, and that's even more centralized than DNS.

-david

> -----Original Message-----
> From: Enrico Marocco [mailto:enrico.marocco at telecomitalia.it]
> Sent: Monday, December 04, 2006 2:26 AM
> To: David Barrett
> Cc: 'David A. Bryan'; 'P2P-SIP'
> Subject: Re: [p2p-sip] Where are we "getting wrapped around the DNS axle"
> 
> David Barrett wrote:
> > Furthermore, yes, I'm assuming whoever you registered your DNS name with
> > would continue to maintain it, else you'd move it to a more reliable
> > provider.  This doesn't seem at all an unreasonable or unusual demand.
> With
> > few exceptions, DNS providers generally aren't capricious.
> 
> That's the whole point.  We have been figuring a system that doesn't
> depend on any provider since the very beginning.  If some are available,
> it can make use of them (e.g. for bootstrap), otherwise it must work
> anyway, even if with limited (but still required, at least in some
> "uncommon" scenarios) functionalities.
> 
> This was the very first assumption and unfortunately a solution which
> depends on a service "provided" by someone doesn't meet it.  Nobody's
> asserting that such a solution wouldn't be better than p2p-sip; it's
> just that it isn't p2p-sip.  Period.
> 
> --
> Ciao,
> Enrico


From mah at inode.at  Mon Dec  4 13:32:17 2006
From: mah at inode.at (Michael Haberler)
Date: Mon, 04 Dec 2006 19:32:17 +0100
Subject: [p2p-sip] Where are we "getting wrapped around the DNS axle"
In-Reply-To: <003701c717c2$3700c4c0$029b0a0a@matthewdeskxps>
References: <003701c717c2$3700c4c0$029b0a0a@matthewdeskxps>
Message-ID: <457469B1.6050705@inode.at>

Matthew Kaufman schrieb:
> I would argue that doing a DNS-protocol lookup given the fixed IP addresses
> of the canonical root DNS servers (and, if you like, the fixed IP addresses
> of some non-traditional root DNS servers) is *exactly the same* as doing a
> DHT-protocol lookup given the fixed IP addresses of some seed nodes for the
> DHT. 
Amen.

I too fail to understand the conceptual and political difference between
"set of supernodes" and "set of root servers" IP addresses - both define
the entry point into a name resolution space.

It's shifting bottlenecks (if the DNS is viewed as such) - including the
downstream politics around the "single supernode set".


-Michael


From dbarrett at quinthar.com  Mon Dec  4 13:53:51 2006
From: dbarrett at quinthar.com (David Barrett)
Date: Mon, 4 Dec 2006 10:53:51 -0800
Subject: [p2p-sip] Where are we "getting wrapped around the DNS axle"
In-Reply-To: <24CCCC428EFEA2469BF046DB3C7A8D221413A5@namail5.corp.adobe.com>
Message-ID: <046e01c717d5$8b9e0390$0202a8c0@Quinthar>

Everything you describe can be done with DNS.  Yes, you could create a
replacement to DNS that accomplishes the same thing with different
technology, and then fundamentally tie P2P-SIP to that instead.  It just
seems like a waste of time.

DNS already offers the foundation for the brilliant future you want to
create:

- Multiple, technically and legally independent providers
- Massive redundancy
- Massive scalability
- No single point of failure
- No dependency on any single legal entity
- International legal framework for conflict resolution

Using DNS is absolutely nothing like being tied down to a mobile provider.
Unlike my mobile phone, I can take my IP address to any DNS provider
whenever I want.

The "powerful shift" you're talking about already happened; the "overlay
access service" you want to create was created decades ago: it's called DNS,
its fee is paid from every internet account, and it's used by nearly every
internet application.  Think of it as the original P2P service.

I understand that people see DNS as having problems.  What I don't
understand is why it's assumed those problems will magically disappear when
we replace DNS with a fundamentally equivalent but technically more
complicated and completely untested service (that, incidentally, is based
upon DHT technology that is *already* obsolete).

Anyway, I accept P2P-SIP isn't going to go this direction.  I just haven't
heard any good explanation as to why, other than it's not "P2P" enough
(though what that means is unclear). 

To review: DNS can support every use case a DHT can support (minus the
impossible security standards set out in 3.2, which nobody can meet, though
it might be able to meet them "good enough" once it's clear what that
means).  It works everywhere real users are.  It can update fast.  It works
today.  It exists.

Unless we can do away entirely with namespace management (keydrives, or
perhaps emailing around public keys), we need something almost exactly like
DNS.

-david

> -----Original Message-----
> From: Henry Sinnreich [mailto:hsinnrei at adobe.com]
> Subject: RE: [p2p-sip] Where are we "getting wrapped around the DNS axle"
> 
> >We have been figuring a system that doesn't depend on any provider
> since >the very beginning.  If some are available, it can make use of
> them (e.g. >for bootstrap), otherwise it must work anyway
> 
> This is a very powerful model shift indeed.
> Such a P2P SIP enabled devices could have multiple accounts, so that
> users could configure it for their own community networks ("providers"
> if you will), but also use providers of _their own choice_.
> 
> Maybe the overlay Internet, the openDHT can interconnect all P2P SIP
> devices for an access fee, similar to the access fee paid today for
> broadband IP. Think of Overlay Access Service (OAS) similar to the IP
> broadband access existing today.
> 
> If the above gets implemented, there will be a huge market for P2P SIP
> communication devices, based on a common global standard and on open
> communications under user control.
> 
> Such devices are the opposite of today's locked down mobile phones that
> work only for one single provider.
> 
> Don't expect though this new business model to have smooth sailing...
> 
> Thanks, Henry
> 
> -----Original Message-----
> From: p2p-sip-bounces at cs.columbia.edu
> [mailto:p2p-sip-bounces at cs.columbia.edu] On Behalf Of Enrico Marocco
> Sent: Monday, December 04, 2006 4:26 AM
> To: David Barrett
> Cc: 'P2P-SIP'; 'David A. Bryan'
> Subject: Re: [p2p-sip] Where are we "getting wrapped around the DNS
> axle"
> 
> David Barrett wrote:
> > Furthermore, yes, I'm assuming whoever you registered your DNS name
> with
> > would continue to maintain it, else you'd move it to a more reliable
> > provider.  This doesn't seem at all an unreasonable or unusual demand.
> With
> > few exceptions, DNS providers generally aren't capricious.
> 
> That's the whole point.  We have been figuring a system that doesn't
> depend on any provider since the very beginning.  If some are available,
> it can make use of them (e.g. for bootstrap), otherwise it must work
> anyway, even if with limited (but still required, at least in some
> "uncommon" scenarios) functionalities.
> 
> This was the very first assumption and unfortunately a solution which
> depends on a service "provided" by someone doesn't meet it.  Nobody's
> asserting that such a solution wouldn't be better than p2p-sip; it's
> just that it isn't p2p-sip.  Period.
> 
> --
> Ciao,
> Enrico
> _______________________________________________
> p2p-sip mailing list
> p2p-sip at cs.columbia.edu
> https://lists.cs.columbia.edu/cucslists/listinfo/p2p-sip


From ag at ag-projects.com  Mon Dec  4 13:55:18 2006
From: ag at ag-projects.com (Adrian Georgescu)
Date: Mon, 4 Dec 2006 19:55:18 +0100
Subject: [p2p-sip] Fwd: what's wrong with DNS?
In-Reply-To: <5D1A7985295922448D5550C94DE29180919334@DEEXC1U01.de.lucent.com>
References: <5D1A7985295922448D5550C94DE29180919334@DEEXC1U01.de.lucent.com>
Message-ID: <5CAF1287-0BCC-4311-BF0C-E9A73820D63E@ag-projects.com>

Yes Keith, plus

locate and use services related to a SIP AoR in a non-centralized way

Adrian


On Dec 4, 2006, at 7:25 PM, Drage, Keith ((Keith)) wrote:

> To the best of my knowledge, the draft charter encompasses anything  
> that
> SIP can do.
>
> It discusses nothing about the application of SIP, merely provides an
> alternative mechanisms for reaching the address of record in SIP.
>
> What SIP can do is of course in the purview of the SIP or SIPPING WG.
>
> Regards
>
> Keith
>
>> -----Original Message-----
>> From: p2p-sip-bounces at cs.columbia.edu
>> [mailto:p2p-sip-bounces at cs.columbia.edu] On Behalf Of David Barrett
>> Sent: 28 November 2006 22:15
>> To: 'Sean Olson'; 'Henry Sinnreich'; 'Brian Rosen'; 'Matthew
>> Kaufman'; slavitch at gmail.com
>> Cc: p2p-sip at cs.columbia.edu; 'Kundan Singh'
>> Subject: Re: [p2p-sip] Fwd: what's wrong with DNS?
>>
>> Ah, perhaps this is my mistake.  Isn't it limited to the
>> general class of audio/video/realtime communications?  For
>> example, is background swarming file-transfers out of scope?
>> I apologize; I clearly don't know what I'm talking about here
>> regarding scope, I'm sorry for the confusion.
>>
>> To take it from another angle, what is out of scope?  Are we
>> competing head to head with UPnP for generalized resource
>> advertisement and discovery?
>>
>> -david
>>
>>> -----Original Message-----
>>> From: Sean Olson [mailto:seanol at exchange.microsoft.com]
>>> Sent: Tuesday, November 28, 2006 1:57 PM
>>> To: David Barrett; 'Henry Sinnreich'; 'Brian Rosen'; 'Matthew
>>> Kaufman'; slavitch at gmail.com
>>> Cc: p2p-sip at cs.columbia.edu; 'Kundan Singh'
>>> Subject: RE: [p2p-sip] Fwd: what's wrong with DNS?
>>>
>>> I don't understand -- we didn't limit the charter of the SIP WG to
>>> just voice. Why would it be different for P2PSIP?  Why
>> would you get
>>> less service from the P2P network than a traditional trapezoid? IM,
>>> file transfer, video, etc. are all reasonable modalities
>> for P2P. Am I
>>> missing something that would make these harder than voice?
>>>
>>> For multicast, the issue is not technical support but admin
>> enablement.
>>> Some do; some don't. I've got nothing against using
>> multicast as one
>>> option. But if it is the only option, that would be a deployment
>>> blocker in many real-world customers.
>>>
>>>
>>>
>>> -----Original Message-----
>>> From: David Barrett [mailto:dbarrett at quinthar.com]
>>> Sent: Tuesday, November 28, 2006 1:53 PM
>>> To: Sean Olson; 'Henry Sinnreich'; 'Brian Rosen'; 'Matthew
>> Kaufman';
>>> slavitch at gmail.com
>>> Cc: p2p-sip at cs.columbia.edu; 'Kundan Singh'
>>> Subject: RE: [p2p-sip] Fwd: what's wrong with DNS?
>>>
>>> Incidentally, I believe ad hoc networks *do* support multicast.
>>> Indeed, multicast on the LAN is actually quite common (it's
>> multicast
>>> on the WAN that's generally not available).
>>>
>>> Furthermore, it's my impression that the scope of our charter *is*
>>> limited to voice.  If we don't draw the line there, then
>> where *do* we
>>> draw the line?
>>>
>>> -david
>>>
>>>> -----Original Message-----
>>>> From: Sean Olson [mailto:seanol at exchange.microsoft.com]
>>>> Sent: Tuesday, November 28, 2006 1:45 PM
>>>> To: David Barrett; 'Henry Sinnreich'; 'Brian Rosen'; 'Matthew
>>>> Kaufman'; slavitch at gmail.com
>>>> Cc: p2p-sip at cs.columbia.edu; 'Kundan Singh'
>>>> Subject: RE: [p2p-sip] Fwd: what's wrong with DNS?
>>>>
>>>> I'm not advocating limiting this to voice. mDNS is great
>> except when
>>>> multicast is not available. But my real question is not so much
>>>> about technology but the scenario.
>>>>
>>>> Is this a scenario that is being considered?
>>>>
>>>>
>>>> -----Original Message-----
>>>> From: David Barrett [mailto:dbarrett at quinthar.com]
>>>> Sent: Tuesday, November 28, 2006 1:43 PM
>>>> To: Sean Olson; 'Henry Sinnreich'; 'Brian Rosen';
>> 'Matthew Kaufman';
>>>> slavitch at gmail.com
>>>> Cc: p2p-sip at cs.columbia.edu; 'Kundan Singh'
>>>> Subject: RE: [p2p-sip] Fwd: what's wrong with DNS?
>>>>
>>>> If they're in the same room, why do they need VoIP?  The range of
>>>> wifi isn't much wider than the range of the spoken voice.
>>>>
>>>> Even if they did somehow have wifi but not voice range, mDNS (as
>>>> Matthew has repeatedly pointed out) is exactly suited for this
>>>> problem.  If they
>>> want
>>>> VoIP, they'll probably also want file transfers, IM, and all the
>>>> other accoutrements of internet life.  Why would we have one
>>>> solution for
>>> VoIP,
>>>> and use something totally different for everything else?
>>>>
>>>> Why not just make mDNS really easy to use, and then use that for
>>>> these fringe VoIP cases?
>>>>
>>>> -david
>>>>
>>>>> -----Original Message-----
>>>>> From: Sean Olson [mailto:seanol at exchange.microsoft.com]
>>>>> Sent: Tuesday, November 28, 2006 1:36 PM
>>>>> To: David Barrett; 'Henry Sinnreich'; 'Brian Rosen'; 'Matthew
>>> Kaufman';
>>>>> slavitch at gmail.com
>>>>> Cc: p2p-sip at cs.columbia.edu; 'Kundan Singh'
>>>>> Subject: RE: [p2p-sip] Fwd: what's wrong with DNS?
>>>>>
>>>>> It would be nice to support an ad-hoc WiFi network
>> among peers and
>>>>> use P2PSIP for voice, chat, etc. This is a scenario
>> where DNS does
>>>>> not
>>> exist
>>>>> necessarily and it is actually fairly common.  Take for
>> example, a
>>>>> set
>>>> of
>>>>> consultants working on site with a customer. They
>> probably don't
>>>>> have direct access to the customer's network and maybe
>> not to the
>>>>> broader Internet either. They will however likely have
>> WiFi and be
>>>>> able to
>>> setup
>>>> a
>>>>> small ad-hoc network amongst themselves.
>>>>>
>>>>> Is this scenario out-of-scope for this WG?
>>>>>
>>>>>
>>>>> -----Original Message-----
>>>>> From: p2p-sip-bounces at cs.columbia.edu [mailto:p2p-sip-
>>>>> bounces at cs.columbia.edu] On Behalf Of David Barrett
>>>>> Sent: Tuesday, November 28, 2006 1:30 PM
>>>>> To: 'Henry Sinnreich'; 'Brian Rosen'; 'Matthew Kaufman';
>>>>> slavitch at gmail.com
>>>>> Cc: p2p-sip at cs.columbia.edu; 'Kundan Singh'
>>>>> Subject: Re: [p2p-sip] Fwd: what's wrong with DNS?
>>>>>
>>>>>> -----Original Message-----
>>>>>> From: Henry Sinnreich [mailto:hsinnrei at adobe.com]
>>>>>> Subject: RE: [p2p-sip] Fwd: what's wrong with DNS?
>>>>>>
>>>>>> BTW: Existing P2P traffic is independent of the DNS and it
>>>>>> dominates
>>>> the
>>>>>> Internet traffic. Numbers between 60% and 80% have
>> been published.
>>>>>
>>>>> I apologize, but I think that's just simply false and
>> misleading.
>>>>>
>>>>> First, virtually all of that traffic is file-sharing traffic
>>>>> designed
>>> to
>>>>> preserve anonymity, which is the direct opposite of our problem.
>>> Using
>>>>> that
>>>>> as justification to keep our system "clean" of DNS is a
>> red herring.
>>>>>
>>>>> But even ignoring that, most of that traffic is
>> Bittorrent, which
>>> *does*
>>>>> use
>>>>> DNS not only for search (to find torrents on btjunkie.org and so
>>> forth),
>>>>> but
>>>>> to find peers (to contact the tracker).
>>>>>
>>>>> At the end of the day, DNS is a fundamental part of the
>> internet,
>>>>> used
>>>> by
>>>>> virtually very user not only ever day, but almost every
>> minute.
>>>>> If
>>> DNS
>>>>> went
>>>>> down globally, the vast majority of internet traffic would also
>>>>> slow
>>> to
>>>> a
>>>>> stop (web, email, bittorrent, etc).  There would be a few
>>>>> survivors -- Skype with its hard-coded IPs would
>> persevere, along
>>>>> with eDonkey, Kazaa and other anonymous or totally
>> closed/private
>>>>> services, but about everything
>>> else
>>>>> would grind to a halt.  Planning on how to keep working without
>>>>> DNS is like planning on how to work without power -- you just
>>>>> don't.  If DNS goes down, we have bigger problems to
>> worry about.
>>>>>
>>>>> We certainly *can* reinvent the wheel and figure out how to keep
>>> working
>>>>> in
>>>>> this apocalypse scenario, I just don't see why we should bother.
>>>>>
>>>>> -david
>>>>>
>>>>> _______________________________________________
>>>>> p2p-sip mailing list
>>>>> p2p-sip at cs.columbia.edu
>>>>> https://lists.cs.columbia.edu/cucslists/listinfo/p2p-sip
>>
>> _______________________________________________
>> p2p-sip mailing list
>> p2p-sip at cs.columbia.edu
>> https://lists.cs.columbia.edu/cucslists/listinfo/p2p-sip
>>
>
> _______________________________________________
> p2p-sip mailing list
> p2p-sip at cs.columbia.edu
> https://lists.cs.columbia.edu/cucslists/listinfo/p2p-sip


From slavitch at gmail.com  Mon Dec  4 14:04:49 2006
From: slavitch at gmail.com (Michael Slavitch)
Date: Mon, 4 Dec 2006 14:04:49 -0500
Subject: [p2p-sip] Where are we "getting wrapped around the DNS axle"
In-Reply-To: <457469B1.6050705@inode.at>
References: <003701c717c2$3700c4c0$029b0a0a@matthewdeskxps>
	<457469B1.6050705@inode.at>
Message-ID: <1fc32c8f0612041104i745461d7s7d0c5b27350a7ff0@mail.gmail.com>

>"Doesn't depend on any provider" seems impossible.  At the very least, you
need someone to "provide" network access and electricity."

Good comparison. Both are utilities. You need both.  Nothing else.  "But
don't we also need someone to "provide" a name space?"

Yes. But that can be my receptionist. Or an editor. Or a meeting
facilitator. Or a regional branch manager.   Or a parent-teacher group.
Or a political party, or a club with 100m members.   No provider is needed
when the namespace is declared by the application. Sorry.

The application decides when wrong the credential-provider is, not the
pipe.  Where is a network provider in this space?  Where could a network
provider be in this space?   DNS is not even wrong. It is not relevant,
because the network is a pipe but nothing else.


"In other words, do you agree:

       1) P2P-SIP must guarantee that each name resolves to a single user

       2) Multiple users want the same name
"

In each overlay, yes. See above three times.  That is an application-level
responsibility.  Not a network responsibility. The network is a pipe.

"If both of those are anwered yes, then how can we avoid a "namespace
provider"?  DNS already does this today on a massive scale, and is available
everywhere real users are.  Why are we so desperate to avoid it?"

Again:  DNS doesn't exist in application namespaces.  It  time to decouple
the namespace provider from the service.  A network service provider can't
keep up with the application space, and doesn't want to in the same way that
the telecom provider won't keep up with the telecommunications boom.

Look at your garden variety PBX and tell my how much different it is from a
VAX,  while our desktops are now supercomputers. What a waste!

In P2P the namespace comes from the application, not the network.  The
network is a pipe.  The provider does not exist apart from that pipe.

My electric utility sells me electrons, it doesn't decide on what they are
used for.  In P2PSIP the network is the network but nothing more.




On 12/4/06, Michael Haberler <mah at inode.at> wrote:
>
> Matthew Kaufman schrieb:
> > I would argue that doing a DNS-protocol lookup given the fixed IP
> addresses
> > of the canonical root DNS servers (and, if you like, the fixed IP
> addresses
> > of some non-traditional root DNS servers) is *exactly the same* as doing
> a
> > DHT-protocol lookup given the fixed IP addresses of some seed nodes for
> the
> > DHT.
> Amen.
>
> I too fail to understand the conceptual and political difference between
> "set of supernodes" and "set of root servers" IP addresses - both define
> the entry point into a name resolution space.
>
> It's shifting bottlenecks (if the DNS is viewed as such) - including the
> downstream politics around the "single supernode set".
>
>
> -Michael
>
> _______________________________________________
> p2p-sip mailing list
> p2p-sip at cs.columbia.edu
> https://lists.cs.columbia.edu/cucslists/listinfo/p2p-sip
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: https://lists.cs.columbia.edu/pipermail/p2p-sip/attachments/20061204/4174a616/attachment.html

From ag at ag-projects.com  Mon Dec  4 14:23:58 2006
From: ag at ag-projects.com (Adrian Georgescu)
Date: Mon, 4 Dec 2006 20:23:58 +0100
Subject: [p2p-sip] Where are we "getting wrapped around the DNS axle"
In-Reply-To: <046e01c717d5$8b9e0390$0202a8c0@Quinthar>
References: <046e01c717d5$8b9e0390$0202a8c0@Quinthar>
Message-ID: <89227BAE-3B62-4909-8EB6-3EAA5DD26101@ag-projects.com>

David,

How do you sugest to implement a replacement of bidirectional  
signalling, nat traversal and data storage associated with a SIP AoR  
with a simple address resolution mechanism provided by today's DNS?

Adrian

On Dec 4, 2006, at 7:53 PM, David Barrett wrote:

> Everything you describe can be done with DNS.  Yes, you could create a
> replacement to DNS that accomplishes the same thing with different
> technology, and then fundamentally tie P2P-SIP to that instead.  It  
> just
> seems like a waste of time.
>
> DNS already offers the foundation for the brilliant future you want to
> create:
>
> - Multiple, technically and legally independent providers
> - Massive redundancy
> - Massive scalability
> - No single point of failure
> - No dependency on any single legal entity
> - International legal framework for conflict resolution
>


From dbarrett at quinthar.com  Mon Dec  4 14:43:45 2006
From: dbarrett at quinthar.com (David Barrett)
Date: Mon, 4 Dec 2006 11:43:45 -0800
Subject: [p2p-sip] Where are we "getting wrapped around the DNS axle"
In-Reply-To: <89227BAE-3B62-4909-8EB6-3EAA5DD26101@ag-projects.com>
Message-ID: <048401c717dc$83236f50$0202a8c0@Quinthar>

Bidirectional signaling: I'm sorry, I don't know what that means.

NAT traversal: Just like before.  My UA looks up your proxy by resolving
your domain name, it sends an INVITE, you receive it, and we do
STUN/TURN/ICE like before.

Data storage: You mean storing arbitrary data for each SIP AoR?  If this is
necessary, I say the P2P-SIP client maintains the master copy of whatever
this data is, and then it uploads it to the SIP proxy at REGISTER time.
Thus to look up the data for a given address, just look up the proxy and ask
it.


I think we should use DNS is we need to resolve names to IPs.  Anything
other than that probably isn't a good fit for DNS (though there *are* other
DNS records than IP, so perhaps those might fit in somewhere).

-david 

> -----Original Message-----
> From: Adrian Georgescu [mailto:ag at ag-projects.com]
> Sent: Monday, December 04, 2006 11:24 AM
> To: David Barrett
> Cc: 'Henry Sinnreich'; 'Enrico Marocco'; 'P2P-SIP'; 'David A. Bryan'
> Subject: Re: [p2p-sip] Where are we "getting wrapped around the DNS axle"
> 
> David,
> 
> How do you sugest to implement a replacement of bidirectional
> signalling, nat traversal and data storage associated with a SIP AoR
> with a simple address resolution mechanism provided by today's DNS?
> 
> Adrian
> 
> On Dec 4, 2006, at 7:53 PM, David Barrett wrote:
> 
> > Everything you describe can be done with DNS.  Yes, you could create a
> > replacement to DNS that accomplishes the same thing with different
> > technology, and then fundamentally tie P2P-SIP to that instead.  It
> > just
> > seems like a waste of time.
> >
> > DNS already offers the foundation for the brilliant future you want to
> > create:
> >
> > - Multiple, technically and legally independent providers
> > - Massive redundancy
> > - Massive scalability
> > - No single point of failure
> > - No dependency on any single legal entity
> > - International legal framework for conflict resolution
> >


From huang-ming.pan at comcast.net  Mon Dec  4 15:11:36 2006
From: huang-ming.pan at comcast.net (Peter Pan)
Date: Mon, 4 Dec 2006 12:11:36 -0800
Subject: [p2p-sip] Where are we "getting wrapped around the DNS axle"
References: <8b2769930612031140y1be88b95n611187a17bf3d8aa@mail.gmail.com><035201c71721$46701a60$0202a8c0@Quinthar>
	<8b2769930612031921y6cc68183h802df07893d4fdee@mail.gmail.com>
Message-ID: <005d01c717e0$6b3afe90$030aa8c0@comcast.net>

Hi,

Me totally agree with David Bryan on this issue. Somebodies out there have
proved p2p concept without DNS and earned worldwide respects, so can p2p-sip
same way. Slow-propagating DNS should not be the first aid for node lookup.
Besides, has existing DNS infrastructure without significant
scability/stability upgrade been proven to be able to support millions of
churning nodes?

As for mobility of p2p nodes, I think it needs upgrade of existing routing
infrastructure to support ip mobility and it is not DNS related at all.

Thanks.

Peter

----- Original Message -----
From: "David A. Bryan" <bryan at ethernot.org>
To: "David Barrett" <dbarrett at quinthar.com>
Cc: "P2P-SIP" <p2p-sip at cs.columbia.edu>
Sent: Sunday, December 03, 2006 7:21 PM
Subject: Re: [p2p-sip] Where are we "getting wrapped around the DNS axle"


> I still fail to see how having volunteer proxies is P2P. Just because
> "a lot people" are participating, it isn't necessarily decentralized,
> you've simply moved the central people you are dependant on from being
> a company to be one or more volunteers. The individual users using the
> service aren't participating (at least not all of them) in providing
> the services, they are only using them. You are still relying on
> centralized devices that are maintained and provided by others, in
> this case, some group of volunteer proxies and some dynamic DNS
> servers that may or may not be provided on a volunteer basis. Such a
> system is thus susceptible to their whims and desires of those
> providing it to stop at any time.
>
> I simply don't see that it is a truly decentralized system, where one
> (or a few) entities deciding to stop volunteering or being unable to
> volunteer can mean the system fails. Nothing wrong with it of course,
> it could work, and certainly would be interesting work to go do (as
> others pointed out, the work could likely be submitted as a draft to
> SIPPING) I don't think it is really in line with the type of system
> that at least many of us who have been working to get this group off
> the ground really had in mind.
>
> There certainly are also scenarios (the impeded access, security
> conscious small organization, and anonymous communications -- all of
> 3.2 in draft-bryan-sipping-p2p-usecases seem to fall into that boat)
> where I can't see the approach you propose working, either.
>
> I'm sorry I was a bit unclear with my question about how to do
> mobility. It is more of a question of handling the update period in a
> cached DNS-like system with sufficient speed. 5 minutes is a
> reasonable time period to change a phone location, but I am not at all
> convinced that conventional dynamic DNS as it is today can push
> updates that fast, or it would scale if you tried. DNS does a heck of
> a lot of caching. Not sure, but I would think there would be stale
> caches causing all sorts of problems. It could work very well for a
> permanent or semi-permanent desk phone of course, just not practical
> in all the scenarios we have discussed, at least in my mind.
>
> David
>
> On 12/3/06, David Barrett <dbarrett at quinthar.com> wrote:
> > > -----Original Message-----
> > > From: David A. Bryan
> > > Subject: Re: [p2p-sip] Where are we "getting wrapped around the DNS
axle"
> > >
> > > The big question here is who is maintaining these proxies and dynamic
> > > dns servers?
> >
> > A lot of people, as would be expected from any highly distributed
system:
> >
> > - On one end of the spectrum, a domain owner (eg, @intel.com) can choose
to
> > take total responsibility for all calls into that domain and just
install a
> > central SIP server that works in a classic sense.
> >
> > - On the other end of the spectrum, a random user can trust the kindness
of
> > any number of benevolent "free" networks (each identified by its own
domain,
> > like @p2psip.org or @myfreebiesip.com, with DNS used to find the
> > bootstrapping servers), leveraging the donated resources of volunteer
nodes.
> >
> > As for maintenance of specific systems:
> >
> > - Standard SIP proxies would be maintained by either standard SIP
providers
> > or large corporations (eg, intel.com might offer a robust SIP deployment
for
> > its employees).
> >
> > - Dynamic DNS servers would be maintained by the dynamic DNS providers,
of
> > which there are dozens (http://www.technopagan.org/dynamic/).  A Dynamic
DNS
> > provider is nothing more than a standard DNS registrar that happens to
> > support an automated configuration interface and fast-propagating
bindings.
> > I'd expect that in the future, *every* DNS registrar would support
dynamic
> > DNS.
> >
> > - Distributed SIP proxies (eg P2P-SIP proxies on non-NAT'd boxes
registered
> > into some public cloud) would be maintained in the hardware sense by the
> > owners of the non-NAT'd boxes themselves.  However the "cloud" in a more
> > general sense is maintained by the bootstrapping services.  Some might
> > bootstrap into a DHT, others might just maintain a big central list.
Either
> > way, any kind soul with time and resources on his hand can register a
> > domain, create a new cloud (using protocols to be defined by P2P-SIP)
and
> > make sure its good.
> >
> > DNS is far from enslaving the world under central rule; it's the only
system
> > with *truly* decentralized ownership, and I want that benefit in SIP.
> >
> >
> > > This proposal isn't really any different than a
> > > conventional SIP service in terms of the centralization model. Someone
> > > still needs to provide the proxies and someone is providing the
> > > dynamic DNS system, which basically starts to look like a centralized
> > > registrar that someone has to own, you just moved it from speaking SIP
> > > to speaking DNS.
> >
> > For users who like conventional SIP, yes, it's precisely no different,
by
> > design.
> >
> > For users who don't like conventional SIP they:
> > 1) Get a domain from any of dozens of dynamic DNS providers
> > 2) Find an existing bootstrapping service, or create a new one and host
it
> > 3) Enter the following into their P2P-SIP client:
> >         . Dynamic-DNS credentials (the only password they have)
> >         . Bootstrapping server IP address or domain name
> >
> > After which their P2P-SIP client finds public resources using the chosen
> > bootstrapping service (either asking "give me a STUN server" directly or
> > "give me a DHT of STUN servers", depending on what the bootstrapping
service
> > provides), registers with the SIP proxy built-into the P2P-SIP proxy of
some
> > non-NAT'd client (or multiple, for redundancy), uses dynamic DNS to
update
> > his record to point to his current SIP proxy(ies), and everything else
just
> > works as normal.  To the outside world, a random, infrastructure-less
> > P2P-SIP client looks *exactly no different* than a totally client-server
SIP
> > user.  Both have a normal SIP address that is resolved to a SIP proxy
using
> > normal DNS, and whether that SIP proxy is a big server in the sky or a
> > random non-firewalled client, it makes no difference and nobody is the
> > wiser.
> >
> >
> > > I also still haven't heard a good response to handling mobility. I
> > > want to be able to go from home to my mobile, drive 5 minutes, and
> > > then go to the office, and want to be reachable at all points, which I
> > > don't see a dynamic DNS approach doing.
> >
> > Yes, it can do that too.  Let's assume you had some device with two
wireless
> > adapters that started connected to your home network (IPhome), then
> > connected to some kind of cellular network (IPcar) as you walked out the
> > door and dropped IPhome, and then picked up your work network (IPwork)
as
> > you walked into the office and dropped IPcar.
> >
> > DNS supports multiple IP addresses per domain.  To start, your P2P-SIP
> > client would register IPhome, and then it would for a brief period
register
> > both IPhome and IPcar, and then when it went out of home range it'd
register
> > only IPcar, and then as you got into the office it's register IPcar and
> > IPwork, and then when you went into the faraday cage (which we all have,
> > right?) it'd register just IPwork.  Regardless, every time you switch
> > networks, there are two options:
> >
> > 1) You have connectivity, and thus can update your DNS record
> > 2) You do not have connectivity, and thus cannot update your DNS record
> >
> > So long as you can connect to the internet, you can update the relevant
> > records to ensure people on the internet can call you.  And if you can't
> > connect to the internet, it's not a problem.
> >
> > Furthermore, we haven't discussed the mDNS case where even if you walk
into
> > a faraday cage and your company internet connection goes out, you can
still
> > talk with everyone else in the office because they'll do an mDNS lookup
for
> > your domain (using multicast over the LAN, which is very widely
supported)
> > and will find you just fine.
> >
> > Again, DNS, dynamic-DNS, and mDNS handle *all* these cases as good as
> > anything else, and likely much better.  I apologize that it's not sexy,
but
> > it's entirely adequate and indeed probably superior.
> >
> > There are still plenty of challenges to work out and details to
standardize,
> > and the P2P-SIP WG is ideally suited to do it.  Sure it's not inventing
any
> > new class of fun DHT technology (or standardizing an antiquated,
obsolete
> > DHT), but the IETF isn't supposed to invent: it's supposed to
standardize
> > what is already there.  DNS, dynamic DNS, mDNS -- they're here and here
to
> > stay.  P2P-SIP should be figuring out and standardizing best practices,
not
> > inventing replacements.
> >
> > -david
> >
> >
> _______________________________________________
> p2p-sip mailing list
> p2p-sip at cs.columbia.edu
> https://lists.cs.columbia.edu/cucslists/listinfo/p2p-sip


From warlock at media.mit.edu  Mon Dec  4 17:14:46 2006
From: warlock at media.mit.edu (Jacky Mallett)
Date: Mon, 4 Dec 2006 17:14:46 -0500
Subject: [p2p-sip] Where are we "getting wrapped around the DNS axle"
Message-ID: <B1991DFF-6DCF-4231-B5D7-0F78AD7C76FB@media.mit.edu>

>
> The application decides when wrong the credential-provider is, not  
> the pipe.  Where is a network provider in this space?  Where could  
> a network provider be in this space?   DNS is not even wrong. It is  
> not relevant, because the network is a pipe but nothing else.
>
The problem with this line of reasoning is that a network is not a  
pipe.  That's the piece of cable from the back of your computer. The  
network is a system. And in the P2PSIP case, the problem is that that  
system is also a network.

I think there's a misunderstanding underlying some of these exchanges  
as to exactly what a peer-to-peer system is, and what a network is,  
and why that matters. I understand that you're seeing a network just  
as the physical system that carries your packets, and that is correct  
at one level. However, once any distributed communicating system  
becomes large enough it takes on what can loosely be described as  
network properties, if we understand network in this sense of a  
distributed communicating system whose application is the  
communication of packets - such as the Internet, and not just the  
physical hardware.

If your device, whatever it is, has enough capacity to hold the  
addresses of every single other device in your application, then this  
doesn't matter. If there is a single server that is large enough to  
hold every other device's address then it doesn't matter (at least in  
terms of network considerations, single point of failure, phone  
company fascism problms etc. still apply, obviously.)

Any communicating system where that is no longer the case, be it the  
postal system, the phone book, DNS,  Skype, or whatever has to solve  
a routing and lookup problem within itself that is independent of the  
underlying physical network. There is no way round this, the  
underlying mathematics of these systems is rather firm on that point.


> In each overlay, yes. See above three times.  That is an  
> application-level responsibility.  Not a network responsibility.  
> The network is a pipe.

Yes, it is an application responsibility - but the application has  
just taken on important attributes of what is commonly called a  
network. A network is not a pipe. A network  is an emergent effect of  
a set of communication protocols between autonomously functioning nodes.

> Again:  DNS doesn't exist in application namespaces.  It  time to  
> decouple the namespace provider from the service.  A network  
> service provider can't keep up with the application space, and  
> doesn't want to in the same way that the telecom provider won't  
> keep up with the telecommunications boom.

I think it's arguable that DNS actually is decoupled from the  
service, if by service you mean the physical pipe.

Let me suggest that whether or not an application chooses to  
implement its own namespace, or uses one conveniently provided by a  
well established existing system is up to the application. But that  
for any system that is sufficiently large, the resulting solution  
will look a lot like DNS. Either you are below the group size limit,  
where all nodes can have direct knowledge of each other, (or through  
a central server), or you're above it where they can't. Once you're  
above it, DNS or something very like it, is the only solution - and  
again, this is dictated by the underlying topological and  
communication constraints on the system. So why reinvent the wheel here?


> In P2P the namespace comes from the application, not the network.   
> The network is a pipe.  The provider does not exist apart from that  
> pipe.

The piece of cable that leads from your desktop to the wall jack is a  
pipe. The network itself is rather more complex than that. Peer to  
Peer systems are themselves networks built on top of networks. The  
fact that one is physical, and the other virtual doesn't change the  
mathematics of the resulting system. It does increase the size, and  
limits at which certain phenomena become apparent, and it also causes  
those of us who understand the underlying administrative traffic  
implications to be alternately amused and appalled, but it doesn't  
change the fact that they exist.

> My electric utility sells me electrons, it doesn't decide on what  
> they are used for.  In P2PSIP the network is the network but  
> nothing more.

This discussion isn't really about the underlying system providing  
you with packets, it's about those aspects of P2PSIP implementation  
where it takes on network properties, and whether or not services  
that have been developed to allow the underlying network to function  
can help solve those problems efficiently or not. There's a large  
network of power stations and cables behind delivering your electrons  
too, and i rather suspect if i took the network out of that equation,  
you might well find yourself strangely deficient of power, even  
though nothing had actually changed with the supply of electrons.

...Jacky







On 12/4/06, Michael Haberler <mah at inode.at> wrote:
Matthew Kaufman schrieb:
 > I would argue that doing a DNS-protocol lookup given the fixed IP  
addresses
 > of the canonical root DNS servers (and, if you like, the fixed IP  
addresses
 > of some non-traditional root DNS servers) is *exactly the same* as  
doing a
 > DHT-protocol lookup given the fixed IP addresses of some seed  
nodes for the
 > DHT.
Amen.

I too fail to understand the conceptual and political difference between
"set of supernodes" and "set of root servers" IP addresses - both define
the entry point into a name resolution space.

It's shifting bottlenecks (if the DNS is viewed as such) - including the
downstream politics around the "single supernode set".


-Michael

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

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

-------------- next part --------------
An HTML attachment was scrubbed...
URL: https://lists.cs.columbia.edu/pipermail/p2p-sip/attachments/20061204/ced1d745/attachment-0001.html

From hsinnrei at adobe.com  Mon Dec  4 17:40:16 2006
From: hsinnrei at adobe.com (Henry Sinnreich)
Date: Mon, 4 Dec 2006 14:40:16 -0800
Subject: [p2p-sip] Where are we "getting wrapped around the DNS axle"
In-Reply-To: <B1991DFF-6DCF-4231-B5D7-0F78AD7C76FB@media.mit.edu>
Message-ID: <24CCCC428EFEA2469BF046DB3C7A8D221413AF@namail5.corp.adobe.com>

> The problem with this line of reasoning is that a network is not a
pipe.  

>>The application decides when wrong the credential-provider is, not the
pipe. 

 

For some the "stupid Internet" is a pipe. It knows only UDP and TCP over
IP. This is enough for any p2p application level overlays.

For other there is the "Intelligent network"; such as VoIP networks with
network resident applications.

 

For what it's worth, p2p traffic dominates the Internet though the
overlays have various owners that are not service providers using the
same business models that current providers have nor do they require
DNS. As Skype has shown, one can be the leading voice/video, IM and
presence provider without DNS.

 

These reasonings have proven to be irreconcilable. 

This may also explain the meaning of this discussion trail.

 

Thanks, Henry 

 

________________________________

From: p2p-sip-bounces at cs.columbia.edu
[mailto:p2p-sip-bounces at cs.columbia.edu] On Behalf Of Jacky Mallett
Sent: Monday, December 04, 2006 4:15 PM
To: P2P-SIP
Subject: Re: [p2p-sip] Where are we "getting wrapped around the DNS
axle"

 

	 

	The application decides when wrong the credential-provider is,
not the pipe.  Where is a network provider in this space?  Where could a
network provider be in this space?   DNS is not even wrong. It is not
relevant, because the network is a pipe but nothing else.

	 

The problem with this line of reasoning is that a network is not a pipe.
That's the piece of cable from the back of your computer. The network is
a system. And in the P2PSIP case, the problem is that that system is
also a network. 

 

I think there's a misunderstanding underlying some of these exchanges as
to exactly what a peer-to-peer system is, and what a network is, and why
that matters. I understand that you're seeing a network just as the
physical system that carries your packets, and that is correct at one
level. However, once any distributed communicating system becomes large
enough it takes on what can loosely be described as network properties,
if we understand network in this sense of a distributed communicating
system whose application is the communication of packets - such as the
Internet, and not just the physical hardware.

 

If your device, whatever it is, has enough capacity to hold the
addresses of every single other device in your application, then this
doesn't matter. If there is a single server that is large enough to hold
every other device's address then it doesn't matter (at least in terms
of network considerations, single point of failure, phone company
fascism problms etc. still apply, obviously.)

 

Any communicating system where that is no longer the case, be it the
postal system, the phone book, DNS,  Skype, or whatever has to solve a
routing and lookup problem within itself that is independent of the
underlying physical network. There is no way round this, the underlying
mathematics of these systems is rather firm on that point.  

 

 

	In each overlay, yes. See above three times.  That is an
application-level responsibility.  Not a network responsibility. The
network is a pipe.

 

Yes, it is an application responsibility - but the application has just
taken on important attributes of what is commonly called a network. A
network is not a pipe. A network  is an emergent effect of a set of
communication protocols between autonomously functioning nodes.

 

	Again:  DNS doesn't exist in application namespaces.  It  time
to decouple the namespace provider from the service.  A network service
provider can't keep up with the application space, and doesn't want to
in the same way that the telecom provider won't keep up with the
telecommunications boom.

 

I think it's arguable that DNS actually is decoupled from the service,
if by service you mean the physical pipe.

 

Let me suggest that whether or not an application chooses to implement
its own namespace, or uses one conveniently provided by a well
established existing system is up to the application. But that for any
system that is sufficiently large, the resulting solution will look a
lot like DNS. Either you are below the group size limit, where all nodes
can have direct knowledge of each other, (or through a central server),
or you're above it where they can't. Once you're above it, DNS or
something very like it, is the only solution - and again, this is
dictated by the underlying topological and communication constraints on
the system. So why reinvent the wheel here? 

 

 

	In P2P the namespace comes from the application, not the
network.  The network is a pipe.  The provider does not exist apart from
that pipe.

 

The piece of cable that leads from your desktop to the wall jack is a
pipe. The network itself is rather more complex than that. Peer to Peer
systems are themselves networks built on top of networks. The fact that
one is physical, and the other virtual doesn't change the mathematics of
the resulting system. It does increase the size, and limits at which
certain phenomena become apparent, and it also causes those of us who
understand the underlying administrative traffic implications to be
alternately amused and appalled, but it doesn't change the fact that
they exist.

 

	My electric utility sells me electrons, it doesn't decide on
what they are used for.  In P2PSIP the network is the network but
nothing more.

 

This discussion isn't really about the underlying system providing you
with packets, it's about those aspects of P2PSIP implementation where it
takes on network properties, and whether or not services that have been
developed to allow the underlying network to function can help solve
those problems efficiently or not. There's a large network of power
stations and cables behind delivering your electrons too, and i rather
suspect if i took the network out of that equation, you might well find
yourself strangely deficient of power, even though nothing had actually
changed with the supply of electrons. 

 

...Jacky

 

 

 

 

 

 

 

On 12/4/06, Michael Haberler <mah at inode.at <mailto:mah at inode.at> >
wrote:

Matthew Kaufman schrieb:

> I would argue that doing a DNS-protocol lookup given the fixed IP
addresses 

> of the canonical root DNS servers (and, if you like, the fixed IP
addresses

> of some non-traditional root DNS servers) is *exactly the same* as
doing a

> DHT-protocol lookup given the fixed IP addresses of some seed nodes
for the 

> DHT.

Amen.

 

I too fail to understand the conceptual and political difference between

"set of supernodes" and "set of root servers" IP addresses - both define

the entry point into a name resolution space. 

 

It's shifting bottlenecks (if the DNS is viewed as such) - including the

downstream politics around the "single supernode set".

 

 

-Michael

 

_______________________________________________ 

p2p-sip mailing list

p2p-sip at cs.columbia.edu

https://lists.cs.columbia.edu/cucslists/listinfo/p2p-sip
<https://lists.cs.columbia.edu/cucslists/listinfo/p2p-sip> 

 

_______________________________________________

p2p-sip mailing list

p2p-sip at cs.columbia.edu

https://lists.cs.columbia.edu/cucslists/listinfo/p2p-sip

 

-------------- next part --------------
An HTML attachment was scrubbed...
URL: https://lists.cs.columbia.edu/pipermail/p2p-sip/attachments/20061204/c9b16c59/attachment-0001.html

From mah at inode.at  Mon Dec  4 18:13:34 2006
From: mah at inode.at (Michael Haberler)
Date: Tue, 05 Dec 2006 00:13:34 +0100
Subject: [p2p-sip] Where are we "getting wrapped around the DNS axle"
In-Reply-To: <24CCCC428EFEA2469BF046DB3C7A8D221413AF@namail5.corp.adobe.com>
References: <24CCCC428EFEA2469BF046DB3C7A8D221413AF@namail5.corp.adobe.com>
Message-ID: <4574AB9E.9050909@inode.at>

Henry Sinnreich schrieb:
>
> As Skype has shown, one can be the leading voice/video, IM and
> presence provider without DNS.
>
yeah, right - at the minor cost of a set of *hardcoded* IP addresses in
the most obfuscated  code known to mankind (that's "proprietary" for the
rest of us)

you can't be seriously suggesting that as an engineering choice


-Michael


From dbarrett at quinthar.com  Mon Dec  4 18:41:45 2006
From: dbarrett at quinthar.com (David Barrett)
Date: Mon, 4 Dec 2006 15:41:45 -0800
Subject: [p2p-sip] Where are we "getting wrapped around the DNS axle"
In-Reply-To: <4574AB9E.9050909@inode.at>
Message-ID: <04b701c717fd$c3b5e130$0202a8c0@Quinthar>

Not to mention that said hard-coded IP address goes to an uber-centralized
directory service.

Likewise, all of the P2P systems that truly avoid DNS also truly avoid
identity altogether, and thus aren't valid models for VoIP.

-david 

> -----Original Message-----
> From: p2p-sip-bounces at cs.columbia.edu [mailto:p2p-sip-
> bounces at cs.columbia.edu] On Behalf Of Michael Haberler
> Sent: Monday, December 04, 2006 3:14 PM
> To: Henry Sinnreich
> Cc: P2P-SIP
> Subject: Re: [p2p-sip] Where are we "getting wrapped around the DNS axle"
> 
> Henry Sinnreich schrieb:
> >
> > As Skype has shown, one can be the leading voice/video, IM and
> > presence provider without DNS.
> >
> yeah, right - at the minor cost of a set of *hardcoded* IP addresses in
> the most obfuscated  code known to mankind (that's "proprietary" for the
> rest of us)
> 
> you can't be seriously suggesting that as an engineering choice
> 
> 
> -Michael
> 
> _______________________________________________
> p2p-sip mailing list
> p2p-sip at cs.columbia.edu
> https://lists.cs.columbia.edu/cucslists/listinfo/p2p-sip


From huang-ming.pan at comcast.net  Mon Dec  4 18:55:11 2006
From: huang-ming.pan at comcast.net (Peter Pan)
Date: Mon, 4 Dec 2006 15:55:11 -0800
Subject: [p2p-sip] Where are we "getting wrapped around the DNS axle"
References: <24CCCC428EFEA2469BF046DB3C7A8D221413AF@namail5.corp.adobe.com>
	<4574AB9E.9050909@inode.at>
Message-ID: <000801c717ff$a33de3b0$030aa8c0@comcast.net>

----- Original Message -----
From: "Michael Haberler" <mah at inode.at>
To: "Henry Sinnreich" <hsinnrei at adobe.com>
> >
> > As Skype has shown, one can be the leading voice/video, IM and
> > presence provider without DNS.
> >
> yeah, right - at the minor cost of a set of *hardcoded* IP addresses in
> the most obfuscated  code known to mankind (that's "proprietary" for the
> rest of us)
>
> you can't be seriously suggesting that as an engineering choice
>
>
> -Michael
>

How many engineers are not seriously suggesting their bosses to use
proprietary MS Windows and Oracle database in their companies?

What is wrong with proprietary stuff? When it is widely accepted, it becomes
the standard. Ask famous device vendors such as Netgear/Linksys/Dlink/Accton
why they recognize Skype as "a standard".

What is wrong with hardcoded ip addresses to avoid dependency on p2pless
stuff?

Peter



From huang-ming.pan at comcast.net  Mon Dec  4 19:04:16 2006
From: huang-ming.pan at comcast.net (Peter Pan)
Date: Mon, 4 Dec 2006 16:04:16 -0800
Subject: [p2p-sip] Where are we "getting wrapped around the DNS axle"
References: <04b701c717fd$c3b5e130$0202a8c0@Quinthar>
Message-ID: <001401c71800$e7cd8160$030aa8c0@comcast.net>

IMO, sticking to URL-formatted identities is instead an obstacle for p2p
VoIP network and traditional PSTN to converge, because that convergency
needs centralized ENUM system (DNS-based!!).

Words "centralized" and "p2p voip" here apparently are contridictory.

Peter
----- Original Message -----
From: "David Barrett" <dbarrett at quinthar.com>
To: "'Michael Haberler'" <mah at inode.at>; "'Henry Sinnreich'"
<hsinnrei at adobe.com>
Cc: "'P2P-SIP'" <p2p-sip at cs.columbia.edu>
Sent: Monday, December 04, 2006 3:41 PM
Subject: Re: [p2p-sip] Where are we "getting wrapped around the DNS axle"


> Not to mention that said hard-coded IP address goes to an uber-centralized
> directory service.
>
> Likewise, all of the P2P systems that truly avoid DNS also truly avoid
> identity altogether, and thus aren't valid models for VoIP.
>
> -david
>
> > -----Original Message-----
> > From: p2p-sip-bounces at cs.columbia.edu [mailto:p2p-sip-
> > bounces at cs.columbia.edu] On Behalf Of Michael Haberler
> > Sent: Monday, December 04, 2006 3:14 PM
> > To: Henry Sinnreich
> > Cc: P2P-SIP
> > Subject: Re: [p2p-sip] Where are we "getting wrapped around the DNS
axle"
> >
> > Henry Sinnreich schrieb:
> > >
> > > As Skype has shown, one can be the leading voice/video, IM and
> > > presence provider without DNS.
> > >
> > yeah, right - at the minor cost of a set of *hardcoded* IP addresses in
> > the most obfuscated  code known to mankind (that's "proprietary" for the
> > rest of us)
> >
> > you can't be seriously suggesting that as an engineering choice
> >
> >
> > -Michael
> >
> > _______________________________________________
> > p2p-sip mailing list
> > p2p-sip at cs.columbia.edu
> > https://lists.cs.columbia.edu/cucslists/listinfo/p2p-sip
>
> _______________________________________________
> p2p-sip mailing list
> p2p-sip at cs.columbia.edu
> https://lists.cs.columbia.edu/cucslists/listinfo/p2p-sip


From warlock at media.mit.edu  Mon Dec  4 19:10:33 2006
From: warlock at media.mit.edu (Jacky Mallett)
Date: Mon, 4 Dec 2006 19:10:33 -0500
Subject: [p2p-sip] Where are we "getting wrapped around the DNS axle"
In-Reply-To: <24CCCC428EFEA2469BF046DB3C7A8D221413AF@namail5.corp.adobe.com>
References: <24CCCC428EFEA2469BF046DB3C7A8D221413AF@namail5.corp.adobe.com>
Message-ID: <1554FC11-8EC8-436D-826F-6F3BD3F2261E@media.mit.edu>


On Dec 4, 2006, at 5:40 PM, Henry Sinnreich wrote:

> > The problem with this line of reasoning is that a network is not  
> a pipe.
>
> >>The application decides when wrong the credential-provider is,  
> not the pipe.
>
>
>
> For some the ?stupid Internet? is a pipe. It knows only UDP and TCP  
> over IP. This is enough for any p2p application level overlays.

As i said, it depends on what level of abstraction you are looking at  
the system, and this is way above the issue of what protocol you are  
using.  When i send my packets i don't need to worry about the design  
of the network, because its been designed, so that i more or less  
don't have to. When i build any large distributed application, i do  
have to worry about network design - within my application - because  
the system i am building now has network attributes.

> For other there is the ?Intelligent network?; such as VoIP networks  
> with network resident applications.
this is what i referred to as a distributed communicating systems.  
The Internet that provides you with your UDP/TCP pipe is one such,  
VoIP networks are another, DNS in and of itself is another, so for  
that matter is the postal system.
> For what it?s worth, p2p traffic dominates the Internet though the  
> overlays have various owners that are not service providers using  
> the same business models that current providers have nor do they  
> require DNS. As Skype has shown, one can be the leading voice/ 
> video, IM and presence provider without DNS.
My point wasn't that you absolutely need DNS, my point, and that of  
others here, was that you need something that looks awfully like it,  
so why not use it in the first place. Skype has its own internal  
version of DNS, which nobody else can use. I thought we were  
advocating open systems here?

A better analogy than references to skype  is that when i take my  
parcel to UPS, i don't have to use a special UPS only addressing  
scheme - i can use the same one the post office uses - UPS then  
provides a different delivery mechanism. In this sense, DNS is  
providing a shared way of sorting out the addressing scheme, what  
then gets delivered and how it gets there is the distributed  
applications problem.

...jacky







>
>
> From: p2p-sip-bounces at cs.columbia.edu [mailto:p2p-sip- 
> bounces at cs.columbia.edu] On Behalf Of Jacky Mallett
> Sent: Monday, December 04, 2006 4:15 PM
> To: P2P-SIP
> Subject: Re: [p2p-sip] Where are we "getting wrapped around the DNS  
> axle"
>
>
>
>
> The application decides when wrong the credential-provider is, not  
> the pipe.  Where is a network provider in this space?  Where could  
> a network provider be in this space?   DNS is not even wrong. It is  
> not relevant, because the network is a pipe but nothing else.
>
>
>
> The problem with this line of reasoning is that a network is not a  
> pipe.  That's the piece of cable from the back of your computer.  
> The network is a system. And in the P2PSIP case, the problem is  
> that that system is also a network.
>
>
>
> I think there's a misunderstanding underlying some of these  
> exchanges as to exactly what a peer-to-peer system is, and what a  
> network is, and why that matters. I understand that you're seeing a  
> network just as the physical system that carries your packets, and  
> that is correct at one level. However, once any distributed  
> communicating system becomes large enough it takes on what can  
> loosely be described as network properties, if we understand  
> network in this sense of a distributed communicating system whose  
> application is the communication of packets - such as the Internet,  
> and not just the physical hardware.
>
>
>
> If your device, whatever it is, has enough capacity to hold the  
> addresses of every single other device in your application, then  
> this doesn't matter. If there is a single server that is large  
> enough to hold every other device's address then it doesn't matter  
> (at least in terms of network considerations, single point of  
> failure, phone company fascism problms etc. still apply, obviously.)
>
>
>
> Any communicating system where that is no longer the case, be it  
> the postal system, the phone book, DNS,  Skype, or whatever has to  
> solve a routing and lookup problem within itself that is  
> independent of the underlying physical network. There is no way  
> round this, the underlying mathematics of these systems is rather  
> firm on that point.
>
>
>
>
>
>> In each overlay, yes. See above three times.  That is an  
>> application-level responsibility.  Not a network responsibility.  
>> The network is a pipe.
>
>
> Yes, it is an application responsibility - but the application has  
> just taken on important attributes of what is commonly called a  
> network. A network is not a pipe. A network  is an emergent effect  
> of a set of communication protocols between autonomously  
> functioning nodes.
>
>
>
>> Again:  DNS doesn't exist in application namespaces.  It  time to  
>> decouple the namespace provider from the service.  A network  
>> service provider can't keep up with the application space, and  
>> doesn't want to in the same way that the telecom provider won't  
>> keep up with the telecommunications boom.
>
>
> I think it's arguable that DNS actually is decoupled from the  
> service, if by service you mean the physical pipe.
>
>
>
> Let me suggest that whether or not an application chooses to  
> implement its own namespace, or uses one conveniently provided by a  
> well established existing system is up to the application. But that  
> for any system that is sufficiently large, the resulting solution  
> will look a lot like DNS. Either you are below the group size  
> limit, where all nodes can have direct knowledge of each other, (or  
> through a central server), or you're above it where they can't.  
> Once you're above it, DNS or something very like it, is the only  
> solution - and again, this is dictated by the underlying  
> topological and communication constraints on the system. So why  
> reinvent the wheel here?
>
>
>
>
>
>> In P2P the namespace comes from the application, not the network.   
>> The network is a pipe.  The provider does not exist apart from  
>> that pipe.
>
>
> The piece of cable that leads from your desktop to the wall jack is  
> a pipe. The network itself is rather more complex than that. Peer  
> to Peer systems are themselves networks built on top of networks.  
> The fact that one is physical, and the other virtual doesn't change  
> the mathematics of the resulting system. It does increase the size,  
> and limits at which certain phenomena become apparent, and it also  
> causes those of us who understand the underlying administrative  
> traffic implications to be alternately amused and appalled, but it  
> doesn't change the fact that they exist.
>
>
>
>> My electric utility sells me electrons, it doesn't decide on what  
>> they are used for.  In P2PSIP the network is the network but  
>> nothing more.
>
>
> This discussion isn't really about the underlying system providing  
> you with packets, it's about those aspects of P2PSIP implementation  
> where it takes on network properties, and whether or not services  
> that have been developed to allow the underlying network to  
> function can help solve those problems efficiently or not. There's  
> a large network of power stations and cables behind delivering your  
> electrons too, and i rather suspect if i took the network out of  
> that equation, you might well find yourself strangely deficient of  
> power, even though nothing had actually changed with the supply of  
> electrons.
>
>
>
> ...Jacky
>
>
>
>
>
>
>
>
>
>
>
>
> On 12/4/06, Michael Haberler <mah at inode.at> wrote:
>
> Matthew Kaufman schrieb:
>
> > I would argue that doing a DNS-protocol lookup given the fixed IP  
> addresses
>
> > of the canonical root DNS servers (and, if you like, the fixed IP  
> addresses
>
> > of some non-traditional root DNS servers) is *exactly the same*  
> as doing a
>
> > DHT-protocol lookup given the fixed IP addresses of some seed  
> nodes for the
>
> > DHT.
>
> Amen.
>
>
>
> I too fail to understand the conceptual and political difference  
> between
>
> "set of supernodes" and "set of root servers" IP addresses - both  
> define
>
> the entry point into a name resolution space.
>
>
>
> It's shifting bottlenecks (if the DNS is viewed as such) -  
> including the
>
> downstream politics around the "single supernode set".
>
>
>
>
>
> -Michael
>
>
>
> _______________________________________________
>
> p2p-sip mailing list
>
> p2p-sip at cs.columbia.edu
>
> https://lists.cs.columbia.edu/cucslists/listinfo/p2p-sip
>
>
>
> _______________________________________________
>
> p2p-sip mailing list
>
> p2p-sip at cs.columbia.edu
>
> https://lists.cs.columbia.edu/cucslists/listinfo/p2p-sip
>
>
>
>

-------------- next part --------------
An HTML attachment was scrubbed...
URL: https://lists.cs.columbia.edu/pipermail/p2p-sip/attachments/20061204/31c2be3a/attachment-0001.html

From enrico.marocco at telecomitalia.it  Tue Dec  5 06:20:05 2006
From: enrico.marocco at telecomitalia.it (Enrico Marocco)
Date: Tue, 05 Dec 2006 12:20:05 +0100
Subject: [p2p-sip] Where are we "getting wrapped around the DNS axle"
In-Reply-To: <046701c717d2$40c98c70$0202a8c0@Quinthar>
References: <046701c717d2$40c98c70$0202a8c0@Quinthar>
Message-ID: <457555E5.2050309@telecomitalia.it>

David Barrett wrote:
> I think this is the mental leap I haven't been able to make.  "Doesn't
> depend on any provider" seems impossible.  At the very least, you need
> someone to "provide" network access and electricity.  But don't we also need
> someone to "provide" a namespace?

You don't need anyone to provide network access in the ad-hoc case.  I
understand that currently it is a very niche, but since the IETF is so
interested to have two WGs focused on it, we cannot consider the ad-hoc
scenario out of scope.

> In other words, do you agree:
> 
> 	1) P2P-SIP must guarantee that each name resolves to a single user

Generally speaking, no.  The peer protocol must guarantee that the name
used by a user enrolled within an overlay could not be used by another
user in the _same_ overlay.  Then, if the overlay is also accessible
from the Internet, that name is also unique.  However, the latter case,
which is absolutely _optional_ and depends on the deployment, already
uses DNS as the "namespace provider", as specified in RFC 3263.

> 	2) Multiple users want the same name

Possibly, but if an enrollment mechanism is _deployed_ in the overlay
(please note that it may be standardized as for jabber, or not), only
the firstly enrolled user can use it.

-- 
Ciao,
Enrico

From enrico.marocco at telecomitalia.it  Tue Dec  5 06:21:24 2006
From: enrico.marocco at telecomitalia.it (Enrico Marocco)
Date: Tue, 05 Dec 2006 12:21:24 +0100
Subject: [p2p-sip] Where are we "getting wrapped around the DNS axle"
In-Reply-To: <003701c717c2$3700c4c0$029b0a0a@matthewdeskxps>
References: <003701c717c2$3700c4c0$029b0a0a@matthewdeskxps>
Message-ID: <45755634.9030904@telecomitalia.it>

Matthew Kaufman wrote:
> Enrico Marocco:
>> That's the whole point.  We have been figuring a system that doesn't
>> depend on any provider since the very beginning.  If some are available,
>> it can make use of them (e.g. for bootstrap), otherwise it must work
>> anyway, even if with limited (but still required, at least in some
>> "uncommon" scenarios) functionalities.
> 
> I would argue that doing a DNS-protocol lookup given the fixed IP addresses
> of the canonical root DNS servers (and, if you like, the fixed IP addresses
> of some non-traditional root DNS servers) is *exactly the same* as doing a
> DHT-protocol lookup given the fixed IP addresses of some seed nodes for the
> DHT. And I would argue that if one can find the node at which an answer
> would be stored, one can send updated information to that node as often as
> bandwidth permits.

Come on, everybody agrees that when DNS is available both for
provisioning and for lookup, it is better to use it instead of fixed IP
addresses.  It's usage has been proposed and welcomed in many
discussions about bootstrap and interwoking (let me mention
draft-marocco-p2psip-interwork).

The goal of the working group is to specify the minimal set of tools for
managing decentralized overlays capable of routing SIP messages,
preventing identity thefts _within_ an overlay and providing a resource
discovery mechanism for NAT traversal.  It seems that people writing on
this list consider DHT as the most promising technology for achieving
that goal;  if you disagree, just describe your alternative based on
DNS, mDNS or whatever, and try to gather consensus in the way people in
IETF are used to (i.e. writing I-Ds).  On the other hand, if you
disagree not only on the technical solution, but also on the goal, since
it is stated in the charter -- the _contract_ between the WG and the
IESG -- I think you are in the wrong place.

-- 
Ciao,
Enrico

From mah at inode.at  Tue Dec  5 06:19:29 2006
From: mah at inode.at (Michael Haberler)
Date: Tue, 05 Dec 2006 12:19:29 +0100
Subject: [p2p-sip] Where are we "getting wrapped around the DNS axle"
In-Reply-To: <000801c717ff$a33de3b0$030aa8c0@comcast.net>
References: <24CCCC428EFEA2469BF046DB3C7A8D221413AF@namail5.corp.adobe.com>
	<4574AB9E.9050909@inode.at>
	<000801c717ff$a33de3b0$030aa8c0@comcast.net>
Message-ID: <457555C1.2040501@inode.at>

Peter Pan schrieb:
> What is wrong with proprietary stuff? When it is widely accepted, it becomes
> the standard. Ask famous device vendors such as Netgear/Linksys/Dlink/Accton
> why they recognize Skype as "a standard".
>   
Nothing. But somehow I got the impression this list is about open
standards, which is why I fail to understand the cheerleading for
closed, obfuscated architectures with centralized authentication.

This looks to me like there's a strange mix of engineering and political
intents at work.
> What is wrong with hardcoded ip addresses to avoid dependency on p2pless
> stuff?
>   
I gave up patching binaries after my Commodore PET exposure and decided
this isnt going to work for me. YMMV.

-Michael


From eunsooshim at gmail.com  Tue Dec  5 10:05:37 2006
From: eunsooshim at gmail.com (Eunsoo Shim)
Date: Tue, 5 Dec 2006 10:05:37 -0500
Subject: [p2p-sip] DNS for inter-domain? Yes; intra-domain? No
Message-ID: <7b683f3f0612050705q64a08b6gfdadb62f1cd492c2@mail.gmail.com>

It is not easy for me to follow every details of the current hot debate
about DNS for P2P SIP.
Though, it seems to me that one thing is confusing, at least, some of us.
Some of us may remember the presentation on a two-layer P2P SIP architecture
back in the ad-hoc meeting in Dallas.
http://www.p2psip.org/ietf-65/IETF65_P2PSIP_BOF_P2P_SIP_Architecture_Shim.pdf

Slide 4 of the presentation shows how various P2P SIP domains can
interconnect.
Two choices were shown in the slide - Inter-domain P2P overlay and DNS.
Actually later the authors of the draft proposed only DNS for P2P SIP domain
interconnections.

Take a look at also the slides presented by Dean during the first BOF in
Dallas.
http://www.p2psip.org/ietf-65/IETF65_P2PSIP_BOF_Charter_Willis.pdf
See Assumption 4 and 5.

So, if someone wants to set up her/his own P2P SIP domain and interconnect
with other P2P SIP domains, she/he just needs to register the proxy for the
domain with DNS.
Peers of other domains will find the proxy via DNS.
If this is what the current debate is about, I think actually there is not
much difference among people.

However, for intra-domain lookup, some of us think we need to develop a DHT
(or DHT-like) protocol for P2P SIP.
Using DNS for intra-domain lookup is not different from using a central
location server.
The client/server SIP architecture already assumes such a central location
server for each domain.
So my understanding is that what P2PSIP group is trying to do is designing
mechanisms for P2P-style intra-domain location lookup where location records
are distributed among peers rather than in a central location server. A DNS
server is just another type of such a central server.

Am I missing any important aspect?

Thanks.

Eunsoo
-------------- next part --------------
An HTML attachment was scrubbed...
URL: https://lists.cs.columbia.edu/pipermail/p2p-sip/attachments/20061205/6b5230e0/attachment.html

From mlepinski at bbn.com  Tue Dec  5 10:22:54 2006
From: mlepinski at bbn.com (Matt Lepinski)
Date: Tue, 05 Dec 2006 10:22:54 -0500
Subject: [p2p-sip] Avoiding Obsolete DHT Technologies
In-Reply-To: <046e01c717d5$8b9e0390$0202a8c0@Quinthar>
References: <046e01c717d5$8b9e0390$0202a8c0@Quinthar>
Message-ID: <45758ECE.40900@bbn.com>

I've seen the term "Obsolete" come up several times in this thread. In 
particular, it seems that some participants in this discussion 
(including David Barett whose post I quote below) believe that a DHT 
that uses a chord-like ring topology is an obsolete technology. If I 
understand correctly, the claim is that such DHT technology is incapable 
of dealing with the churn inherent in modern P2P use cases.

My question to David (or anyone else on the list who is knowledge about 
DHT technology) is:
Is it the "ring-topology overlay" which is has become obsolete? If so, 
it seems like we should be fine as long as we require our eventual 
protocol architecture to support an arbitrary DHT (and hopefully any 
mandatory-to-implement DHT that we specify can avoid the pitfalls of a 
ring topology).

Alternatively, is it the notion of a DHT that has become obsolete? If 
so, is there something inherent in DHTs that make them unsuitable for 
modern use cases?  In such a case, what requirements should we adopt for 
a P2P solution to avoid the problems inherent in DHTs.

Thanks a lot,
- Matt Lepinski :->

David Barrett wrote:

>Everything you describe can be done with DNS.  Yes, you could create a
>replacement to DNS that accomplishes the same thing with different
>technology, and then fundamentally tie P2P-SIP to that instead.  It just
>seems like a waste of time.
>
>DNS already offers the foundation for the brilliant future you want to
>create:
>
>- Multiple, technically and legally independent providers
>- Massive redundancy
>- Massive scalability
>- No single point of failure
>- No dependency on any single legal entity
>- International legal framework for conflict resolution
>
>Using DNS is absolutely nothing like being tied down to a mobile provider.
>Unlike my mobile phone, I can take my IP address to any DNS provider
>whenever I want.
>
>The "powerful shift" you're talking about already happened; the "overlay
>access service" you want to create was created decades ago: it's called DNS,
>its fee is paid from every internet account, and it's used by nearly every
>internet application.  Think of it as the original P2P service.
>
>I understand that people see DNS as having problems.  What I don't
>understand is why it's assumed those problems will magically disappear when
>we replace DNS with a fundamentally equivalent but technically more
>complicated and completely untested service (that, incidentally, is based
>upon DHT technology that is *already* obsolete).
>
>Anyway, I accept P2P-SIP isn't going to go this direction.  I just haven't
>heard any good explanation as to why, other than it's not "P2P" enough
>(though what that means is unclear). 
>
>To review: DNS can support every use case a DHT can support (minus the
>impossible security standards set out in 3.2, which nobody can meet, though
>it might be able to meet them "good enough" once it's clear what that
>means).  It works everywhere real users are.  It can update fast.  It works
>today.  It exists.
>
>Unless we can do away entirely with namespace management (keydrives, or
>perhaps emailing around public keys), we need something almost exactly like
>DNS.
>
>-david
>
>  
>
>>-----Original Message-----
>>From: Henry Sinnreich [mailto:hsinnrei at adobe.com]
>>Subject: RE: [p2p-sip] Where are we "getting wrapped around the DNS axle"
>>
>>    
>>
>>>We have been figuring a system that doesn't depend on any provider
>>>      
>>>
>>since >the very beginning.  If some are available, it can make use of
>>them (e.g. >for bootstrap), otherwise it must work anyway
>>
>>This is a very powerful model shift indeed.
>>Such a P2P SIP enabled devices could have multiple accounts, so that
>>users could configure it for their own community networks ("providers"
>>if you will), but also use providers of _their own choice_.
>>
>>Maybe the overlay Internet, the openDHT can interconnect all P2P SIP
>>devices for an access fee, similar to the access fee paid today for
>>broadband IP. Think of Overlay Access Service (OAS) similar to the IP
>>broadband access existing today.
>>
>>If the above gets implemented, there will be a huge market for P2P SIP
>>communication devices, based on a common global standard and on open
>>communications under user control.
>>
>>Such devices are the opposite of today's locked down mobile phones that
>>work only for one single provider.
>>
>>Don't expect though this new business model to have smooth sailing...
>>
>>Thanks, Henry
>>
>>-----Original Message-----
>>From: p2p-sip-bounces at cs.columbia.edu
>>[mailto:p2p-sip-bounces at cs.columbia.edu] On Behalf Of Enrico Marocco
>>Sent: Monday, December 04, 2006 4:26 AM
>>To: David Barrett
>>Cc: 'P2P-SIP'; 'David A. Bryan'
>>Subject: Re: [p2p-sip] Where are we "getting wrapped around the DNS
>>axle"
>>
>>David Barrett wrote:
>>    
>>
>>>Furthermore, yes, I'm assuming whoever you registered your DNS name
>>>      
>>>
>>with
>>    
>>
>>>would continue to maintain it, else you'd move it to a more reliable
>>>provider.  This doesn't seem at all an unreasonable or unusual demand.
>>>      
>>>
>>With
>>    
>>
>>>few exceptions, DNS providers generally aren't capricious.
>>>      
>>>
>>That's the whole point.  We have been figuring a system that doesn't
>>depend on any provider since the very beginning.  If some are available,
>>it can make use of them (e.g. for bootstrap), otherwise it must work
>>anyway, even if with limited (but still required, at least in some
>>"uncommon" scenarios) functionalities.
>>
>>This was the very first assumption and unfortunately a solution which
>>depends on a service "provided" by someone doesn't meet it.  Nobody's
>>asserting that such a solution wouldn't be better than p2p-sip; it's
>>just that it isn't p2p-sip.  Period.
>>
>>--
>>Ciao,
>>Enrico
>>_______________________________________________
>>p2p-sip mailing list
>>p2p-sip at cs.columbia.edu
>>https://lists.cs.columbia.edu/cucslists/listinfo/p2p-sip
>>    
>>
>
>_______________________________________________
>p2p-sip mailing list
>p2p-sip at cs.columbia.edu
>https://lists.cs.columbia.edu/cucslists/listinfo/p2p-sip
>
>  
>



From br at brianrosen.net  Tue Dec  5 10:41:38 2006
From: br at brianrosen.net (Brian Rosen)
Date: Tue, 5 Dec 2006 10:41:38 -0500
Subject: [p2p-sip] Avoiding Obsolete DHT Technologies
In-Reply-To: <45758ECE.40900@bbn.com>
Message-ID: <024901c71883$dc6d4700$640fa8c0@cis.neustar.com>

I see the discussion somewhat differently.  I don't think David Barett is
making a judgment call on the obsolescence of any form of DHT.  He is
arguing that DNS, and specifically dynamic DNS, is a better choice to use
for intra-overlay routing of messages.  It IS, undeniably simpler.

The question to the group is, does that solve our requirements?  I don't
think it does.  It may solve a different set of requirements, and I have
suggested that David et. al. write the idea up as an Internet Draft and
submit it to sipping.  It may be an interesting enough concept that it is
taken up as a fairly straightforward BCP thing that uses existing mechanisms
to get a different form of plan jane sip.

I do see your concerns about ring topology as informing our choice of one
mandatory to implement DHT.

Brian


> -----Original Message-----
> From: p2p-sip-bounces at cs.columbia.edu [mailto:p2p-sip-
> bounces at cs.columbia.edu] On Behalf Of Matt Lepinski
> Sent: Tuesday, December 05, 2006 10:23 AM
> To: David Barrett
> Cc: 'P2P-SIP'
> Subject: Re: [p2p-sip] Avoiding Obsolete DHT Technologies
> 
> I've seen the term "Obsolete" come up several times in this thread. In
> particular, it seems that some participants in this discussion
> (including David Barett whose post I quote below) believe that a DHT
> that uses a chord-like ring topology is an obsolete technology. If I
> understand correctly, the claim is that such DHT technology is incapable
> of dealing with the churn inherent in modern P2P use cases.
> 
> My question to David (or anyone else on the list who is knowledge about
> DHT technology) is:
> Is it the "ring-topology overlay" which is has become obsolete? If so,
> it seems like we should be fine as long as we require our eventual
> protocol architecture to support an arbitrary DHT (and hopefully any
> mandatory-to-implement DHT that we specify can avoid the pitfalls of a
> ring topology).
> 
> Alternatively, is it the notion of a DHT that has become obsolete? If
> so, is there something inherent in DHTs that make them unsuitable for
> modern use cases?  In such a case, what requirements should we adopt for
> a P2P solution to avoid the problems inherent in DHTs.
> 
> Thanks a lot,
> - Matt Lepinski :->
> 
> David Barrett wrote:
> 
> >Everything you describe can be done with DNS.  Yes, you could create a
> >replacement to DNS that accomplishes the same thing with different
> >technology, and then fundamentally tie P2P-SIP to that instead.  It just
> >seems like a waste of time.
> >
> >DNS already offers the foundation for the brilliant future you want to
> >create:
> >
> >- Multiple, technically and legally independent providers
> >- Massive redundancy
> >- Massive scalability
> >- No single point of failure
> >- No dependency on any single legal entity
> >- International legal framework for conflict resolution
> >
> >Using DNS is absolutely nothing like being tied down to a mobile
> provider.
> >Unlike my mobile phone, I can take my IP address to any DNS provider
> >whenever I want.
> >
> >The "powerful shift" you're talking about already happened; the "overlay
> >access service" you want to create was created decades ago: it's called
> DNS,
> >its fee is paid from every internet account, and it's used by nearly
> every
> >internet application.  Think of it as the original P2P service.
> >
> >I understand that people see DNS as having problems.  What I don't
> >understand is why it's assumed those problems will magically disappear
> when
> >we replace DNS with a fundamentally equivalent but technically more
> >complicated and completely untested service (that, incidentally, is based
> >upon DHT technology that is *already* obsolete).
> >
> >Anyway, I accept P2P-SIP isn't going to go this direction.  I just
> haven't
> >heard any good explanation as to why, other than it's not "P2P" enough
> >(though what that means is unclear).
> >
> >To review: DNS can support every use case a DHT can support (minus the
> >impossible security standards set out in 3.2, which nobody can meet,
> though
> >it might be able to meet them "good enough" once it's clear what that
> >means).  It works everywhere real users are.  It can update fast.  It
> works
> >today.  It exists.
> >
> >Unless we can do away entirely with namespace management (keydrives, or
> >perhaps emailing around public keys), we need something almost exactly
> like
> >DNS.
> >
> >-david
> >
> >
> >
> >>-----Original Message-----
> >>From: Henry Sinnreich [mailto:hsinnrei at adobe.com]
> >>Subject: RE: [p2p-sip] Where are we "getting wrapped around the DNS
> axle"
> >>
> >>
> >>
> >>>We have been figuring a system that doesn't depend on any provider
> >>>
> >>>
> >>since >the very beginning.  If some are available, it can make use of
> >>them (e.g. >for bootstrap), otherwise it must work anyway
> >>
> >>This is a very powerful model shift indeed.
> >>Such a P2P SIP enabled devices could have multiple accounts, so that
> >>users could configure it for their own community networks ("providers"
> >>if you will), but also use providers of _their own choice_.
> >>
> >>Maybe the overlay Internet, the openDHT can interconnect all P2P SIP
> >>devices for an access fee, similar to the access fee paid today for
> >>broadband IP. Think of Overlay Access Service (OAS) similar to the IP
> >>broadband access existing today.
> >>
> >>If the above gets implemented, there will be a huge market for P2P SIP
> >>communication devices, based on a common global standard and on open
> >>communications under user control.
> >>
> >>Such devices are the opposite of today's locked down mobile phones that
> >>work only for one single provider.
> >>
> >>Don't expect though this new business model to have smooth sailing...
> >>
> >>Thanks, Henry
> >>
> >>-----Original Message-----
> >>From: p2p-sip-bounces at cs.columbia.edu
> >>[mailto:p2p-sip-bounces at cs.columbia.edu] On Behalf Of Enrico Marocco
> >>Sent: Monday, December 04, 2006 4:26 AM
> >>To: David Barrett
> >>Cc: 'P2P-SIP'; 'David A. Bryan'
> >>Subject: Re: [p2p-sip] Where are we "getting wrapped around the DNS
> >>axle"
> >>
> >>David Barrett wrote:
> >>
> >>
> >>>Furthermore, yes, I'm assuming whoever you registered your DNS name
> >>>
> >>>
> >>with
> >>
> >>
> >>>would continue to maintain it, else you'd move it to a more reliable
> >>>provider.  This doesn't seem at all an unreasonable or unusual demand.
> >>>
> >>>
> >>With
> >>
> >>
> >>>few exceptions, DNS providers generally aren't capricious.
> >>>
> >>>
> >>That's the whole point.  We have been figuring a system that doesn't
> >>depend on any provider since the very beginning.  If some are available,
> >>it can make use of them (e.g. for bootstrap), otherwise it must work
> >>anyway, even if with limited (but still required, at least in some
> >>"uncommon" scenarios) functionalities.
> >>
> >>This was the very first assumption and unfortunately a solution which
> >>depends on a service "provided" by someone doesn't meet it.  Nobody's
> >>asserting that such a solution wouldn't be better than p2p-sip; it's
> >>just that it isn't p2p-sip.  Period.
> >>
> >>--
> >>Ciao,
> >>Enrico
> >>_______________________________________________
> >>p2p-sip mailing list
> >>p2p-sip at cs.columbia.edu
> >>https://lists.cs.columbia.edu/cucslists/listinfo/p2p-sip
> >>
> >>
> >
> >_______________________________________________
> >p2p-sip mailing list
> >p2p-sip at cs.columbia.edu
> >https://lists.cs.columbia.edu/cucslists/listinfo/p2p-sip
> >
> >
> >
> 
> 
> _______________________________________________
> p2p-sip mailing list
> p2p-sip at cs.columbia.edu
> https://lists.cs.columbia.edu/cucslists/listinfo/p2p-sip


From hsinnrei at adobe.com  Tue Dec  5 11:19:54 2006
From: hsinnrei at adobe.com (Henry Sinnreich)
Date: Tue, 5 Dec 2006 08:19:54 -0800
Subject: [p2p-sip] Where are we "getting wrapped around the DNS axle"
In-Reply-To: <457555C1.2040501@inode.at>
Message-ID: <24CCCC428EFEA2469BF046DB3C7A8D221413B9@namail5.corp.adobe.com>

Michael,

This WG is about P2P SIP _standards_ as you well know :-)

We and you just seem to differ in our approach to use or not use DNS
inside the p2p system. 

Linking different p2p systems may however use DNS or may use another
overlay. 

There is work on hierarchical is done in several countries:

http://www.ietf.org/internet-drafts/draft-shi-p2psip-hier-arch-00.txt 
and also France Telecom+Polytechnic (I need to find it) on hierarchical
overlays that scale easily to full projected Internet size. My apology
for missing some other references.

Thanks, Henry

-----Original Message-----
From: Michael Haberler [mailto:mah at inode.at] 
Sent: Tuesday, December 05, 2006 5:19 AM
To: Peter Pan
Cc: Henry Sinnreich; P2P-SIP
Subject: Re: [p2p-sip] Where are we "getting wrapped around the DNS
axle"

Peter Pan schrieb:
> What is wrong with proprietary stuff? When it is widely accepted, it
becomes
> the standard. Ask famous device vendors such as
Netgear/Linksys/Dlink/Accton
> why they recognize Skype as "a standard".
>   
Nothing. But somehow I got the impression this list is about open
standards, which is why I fail to understand the cheerleading for
closed, obfuscated architectures with centralized authentication.

This looks to me like there's a strange mix of engineering and political
intents at work.
> What is wrong with hardcoded ip addresses to avoid dependency on
p2pless
> stuff?
>   
I gave up patching binaries after my Commodore PET exposure and decided
this isnt going to work for me. YMMV.

-Michael



From dean.willis at softarmor.com  Tue Dec  5 12:30:09 2006
From: dean.willis at softarmor.com (Dean Willis)
Date: Tue, 5 Dec 2006 11:30:09 -0600
Subject: [p2p-sip] Where are we "getting wrapped around the DNS axle"
In-Reply-To: <003701c717c2$3700c4c0$029b0a0a@matthewdeskxps>
References: <003701c717c2$3700c4c0$029b0a0a@matthewdeskxps>
Message-ID: <CA76E486-93FA-4A84-A730-9A430A3F64C7@softarmor.com>


On Dec 4, 2006, at 10:35 AM, Matthew Kaufman wrote:

> Enrico Marocco:
>>
>> That's the whole point.  We have been figuring a system that doesn't
>> depend on any provider since the very beginning.  If some are  
>> available,
>> it can make use of them (e.g. for bootstrap), otherwise it must work
>> anyway, even if with limited (but still required, at least in some
>> "uncommon" scenarios) functionalities.
>
> I would argue that doing a DNS-protocol lookup given the fixed IP  
> addresses
> of the canonical root DNS servers (and, if you like, the fixed IP  
> addresses
> of some non-traditional root DNS servers) is *exactly the same* as  
> doing a
> DHT-protocol lookup given the fixed IP addresses of some seed nodes  
> for the
> DHT. And I would argue that if one can find the node at which an  
> answer
> would be stored, one can send updated information to that node as  
> often as
> bandwidth permits.

I think that's the point I was trying to make. Where DNS is used to  
find a bootstrap node, either manual configuration or dynamic DNS  
could be used to insert the resource record for that bootstrap node  
into the DNS.

There may be other valid ways to find a bootstrap node than using  
DNS. For example, one might be configured statically, or one might  
use some sort of multicast service like zeroconf, UPNP, or some sort  
of undetermined layer-2 mechanism.

But this is a DIFFERENT PROBLEM than the central problem of P2PSIP.  
That problem, as I see it, is finding a SIP-level route to a node  
when that node may be both highly mobile and "on the other side" of  
multiple NAT and/or firewall boundaries from the node that is trying  
to contact it.

This requires more than a name-to-address lookup operation. It  
requires message routing along ephemeral "channels" that dynamically  
adapt to changing network topology. For those of you who are old  
enough to remember Token Ring, it's something like source-route  
bridging. But it's a lot harder than the problem scope of SRB,  
because the number of nodes is larger and the network topology is  
both more complicated and less stable.

--
Dean




From slavitch at gmail.com  Tue Dec  5 13:54:24 2006
From: slavitch at gmail.com (Michael Slavitch)
Date: Tue, 5 Dec 2006 13:54:24 -0500
Subject: [p2p-sip] Where are we "getting wrapped around the DNS axle"
In-Reply-To: <0E8B6DDE-20B4-4D0B-9ABD-089153A2AAB1@media.mit.edu>
References: <003701c717c2$3700c4c0$029b0a0a@matthewdeskxps>
	<457469B1.6050705@inode.at>
	<1fc32c8f0612041104i745461d7s7d0c5b27350a7ff0@mail.gmail.com>
	<0E8B6DDE-20B4-4D0B-9ABD-089153A2AAB1@media.mit.edu>
Message-ID: <1fc32c8f0612051054l44396cb2y2a24bd5cc6432951@mail.gmail.com>

On 12/4/06, Jacky Mallett <warlock at media.mit.edu> wrote:

> The problem with this line of reasoning is that a network is not a pipe.
> That's the piece of cable from the back of your computer. The network is a
> system. And in the P2PSIP case, the problem is that the system is also a
> network.
>
> I think there's a misunderstanding underlying some of these exchanges as to
> exactly what a peer-to-peer system is, and what a network is, and why that
> matters. I understand that you're seeing a network just as the physical
> system that carries your packets, and that is correct at one level. However,
> once any distributed communicating system becomes large enough it takes on
> what can loosely be described as network properties, if we understand
> network in this sense of a distributed communicating system whose
> application is the communication of packets - such as the Internet, and not
> just the physical hardware.
>

I've been designing communications systems for over 20 years.  For
over 95% of business and consumers the  perspective  is of a
best-effort pipe. From an application perspective it is a pipe.  Once
you get behind a NAT the Internet is a pipe.  The stuff in the NAT? A
cheap cloud, a commodity. The pipe is not well-ordered, it is not
deterministic, performance is best effort and fractal.  The Internet
is what it is, and it is messy and chaotic.  It is largely impossible
for 'the network' to provide services end to end when the network is a
chaotic collage of disjoint backbones and meshes, except for a rare
few researchers governments and very large businesses.  For the rest
of us the Internet looks like 3 AM in the Bronx.

> If your device, whatever it is, has enough capacity to hold the addresses of
> every single other device in your application, then this doesn't matter. If
> there is a single server that is large enough to hold every other device's
> address then it doesn't matter (at least in terms of network considerations,
> single point of failure, phone company fascism problms etc. still apply,
> obviously.)

A sub-$1000 Dell notebook can hold the entire global PSTN dial plan,
the entire US portion can be stored in a cheap flash card.  Cheap
computers are now supercomputers while our providers are steampunk
relics that will never scale and keep up.

One billion dollars spent by one million end user P2P systems,  one
million Christmas shoppers or 0.3% of the US population, will go a lot
farther than one billion dollars spent on core infrastructure.  That
power of scale gives a 100x advantage to the end-nodes, and the
CPU/resource power of the end-nodes is growing very rapidly over time.

The aggregate CPU power of low-cost video-game systems being sold
_just this Christmas alone_ dwarfs the capability of _the entire
deployed infrastructure_ of all service providers wordwide,
_combined_.

It is obvious that the network can't keep up with applications. It is
even more striking when the entire infrastructure of the multi billion
dollar telecommunications empires can't keep up with video game
vendors, which can deploy and install an entire generation of new
equipment between Thanksgiving and New Year's.

The biggest threat to traditional network vendors isn't Huawei.

It is Nintendo.

>
> Any communicating system where that is no longer the case, be it the postal
> system, the phone book, DNS,  Skype, or whatever has to solve a routing and
> lookup problem within itself that is independent of the underlying physical
> network. There is no way round this, the underlying mathematics of these
> systems is rather firm on that point.
>

I fully agree with you there.  It is an essential core problem of P2PSIP.

> Yes, it is an application responsibility - but the application has just
> taken on important attributes of what is commonly called a network. A
> network is not a pipe. A network  is an emergent effect of a set of
> communication protocols between autonomously functioning nodes.

Indeed it is taking over the attributes of the network.  It must
because there is no one network, there is a hodgepodge of networks
over which traffic happens, and that hodge-podge has multidimensional
characteristics that change in terms of location, physical connection
and time.  The only solution to such situations are 'overlays' that
are self-organizing and that re-organize based on the given conditions
at the time.

In the absense of a single end to  network providing all those
attributes there is no option but to put them into the application.
The real protocols that matter at the low level, be they QoS, MPLS,
identity, or even basic routing and  the like, cannot be provisioned
or ordained by the user or the application at any cost, let alone with
automation and simplicity.  They are unavailable from the network in
any real sense.  They are reluctantly provided, if every, in any
commercial space.  However, they are needed. Hence they will come from
the application, and the application will be forced to guess and learn
where it is, what it can do, what it can talk to and how.   The
Internet is not deterministic. It never has been. That was the whole
point in the first place. The reality is despite delusions it never
will be, because that's not how things are organized in reality.

> Again:  DNS doesn't exist in application namespaces.  It  time to decouple
> the namespace provider from the service.  A network service provider can't
> keep up with the application space, and doesn't want to in the same way that
> the telecom provider won't keep up with the telecommunications boom.
>
> I think it's arguable that DNS actually is decoupled from the service, if by
> service you mean the physical pipe.

Yes and no.  They are leashed together.  Full DNS service including
reverse lookup always requires co-operation from the people supplying
the end-node IP and the access to pipe.

> Let me suggest that whether or not an application chooses to implement its
> own namespace, or uses one conveniently provided by a well established
> existing system is up to the application. But that for any system that is
> sufficiently large, the resulting solution will look a lot like DNS. Either

Yes, but with insertion, deletion, content, refresh rate and scope
that is application oriented not low-level IP resolution oriented.
DNS is great for mapping static names to static IP addresses. It
should, because that is what it was designed for.  For the application
it is more than obsolete.

> you are below the group size limit, where all nodes can have direct
> knowledge of each other, (or through a central server), or you're above it
> where they can't. Once you're above it, DNS or something very like it, is
> the only solution - and again, this is dictated by the underlying
> topological and communication constraints on the system. So why reinvent the
> wheel here?

Because the application layer demands the ability to steer the wheel.

>
> This discussion isn't really about the underlying system providing you with
> packets, it's about those aspects of P2PSIP implementation where it takes on
> network properties, and whether or not services that have been developed to
> allow the underlying network to function can help solve those problems
> efficiently or not. There's a large network of power stations and cables
> behind delivering your electrons too, and i rather suspect if i took the
> network out of that equation, you might well find yourself strangely
> deficient of power, even though nothing had actually changed with the supply
> of electrons.

It's all about who controls what.

Power companies are filled with people that come from the command and
control world.  There is a big debate right now about the current
provider-to-customer grid vs. distributed network meshes where the
endpoints have much more control over where they get their power and
when.

P2PSIP is in the same fight.

Right now I don't have control over what generator made those
electrons, nor their price, nor where they come from.  However I can
put panels on my roof, use a wind-turbine, and store excess power in a
redox battery.  How I organize application of electricity should be my
choice.

Television and movie networks are in the same obsolete boat. They are
broadcasting when millions of people are playing Half Life right now,
and deciding what the game is by themselves.

The same flexibility should become available to applications using P2PSIP.

From RADHIKA.R.ROY at saic.com  Tue Dec  5 14:29:45 2006
From: RADHIKA.R.ROY at saic.com (Roy, Radhika R.)
Date: Tue, 5 Dec 2006 14:29:45 -0500 
Subject: [p2p-sip] Where are we "getting wrapped around the DNS axle"
Message-ID: <62D92A9A02BCC845B202323D49A48426230F08@0591-ITS-EXMP02.us.saic.com>

Folks:

I am enclosing an URL for the paper, "Incentive for Large Peer-to-Peer
System," IEEE SAC, May 2006. 

http://ieeexplore.ieee.org/iel5/49/34142/01626429.pdf

Hope that we can take some rest after so much hot discussions on this
topic.

Radhika

From mlepinski at bbn.com  Tue Dec  5 14:51:30 2006
From: mlepinski at bbn.com (Matt Lepinski)
Date: Tue, 05 Dec 2006 14:51:30 -0500
Subject: [p2p-sip] Where are we "getting wrapped around the DNS axle"
In-Reply-To: <62D92A9A02BCC845B202323D49A48426230F08@0591-ITS-EXMP02.us.saic.com>
References: <62D92A9A02BCC845B202323D49A48426230F08@0591-ITS-EXMP02.us.saic.com>
Message-ID: <4575CDC2.7050005@bbn.com>

Radhika,

I've only read the introduction, but the paper seems to be an economic 
analysis of pricing/payment models for P2P systems. In particular, it 
appears that the main result of the paper seems to be that (for 
reasonably large P2P systems) the social welfare obtained from optimal 
incentive payments is not much better than the social welfare obtained 
from a very simple class of pricing model.

Personally, I find the analysis in the paper to be quite interesting 
but, unless I'm missing something, it doesn't seem to be related to the 
current discussion.

- Matt Lepinski :->

Roy, Radhika R. wrote:

>Folks:
>
>I am enclosing an URL for the paper, "Incentive for Large Peer-to-Peer
>System," IEEE SAC, May 2006. 
>
>http://ieeexplore.ieee.org/iel5/49/34142/01626429.pdf
>
>Hope that we can take some rest after so much hot discussions on this
>topic.
>
>Radhika
>_______________________________________________
>p2p-sip mailing list
>p2p-sip at cs.columbia.edu
>https://lists.cs.columbia.edu/cucslists/listinfo/p2p-sip
>
>  
>



From RADHIKA.R.ROY at saic.com  Tue Dec  5 15:44:24 2006
From: RADHIKA.R.ROY at saic.com (Roy, Radhika R.)
Date: Tue, 5 Dec 2006 15:44:24 -0500
Subject: [p2p-sip] Where are we "getting wrapped around the DNS axle"
In-Reply-To: <4575CDC2.7050005@bbn.com>
Message-ID: <62D92A9A02BCC845B202323D49A48426230F0A@0591-ITS-EXMP02.us.saic.com>

Hi, Matt:

You are right. The paper speaks for itself.

The key is this: The P2P system is a different paradigm and we have to
look into this with a little different way with different assumptions
and perspectives.

Otherwise, we will miss a lot of things as people did during the
pre-Internet period.

Best regards,
Radhika

-----Original Message-----
From: Matt Lepinski [mailto:mlepinski at bbn.com] 
Sent: Tuesday, December 05, 2006 2:52 PM
To: Roy, Radhika R.
Cc: P2P-SIP
Subject: Re: [p2p-sip] Where are we "getting wrapped around the DNS
axle"

Radhika,

I've only read the introduction, but the paper seems to be an economic 
analysis of pricing/payment models for P2P systems. In particular, it 
appears that the main result of the paper seems to be that (for 
reasonably large P2P systems) the social welfare obtained from optimal 
incentive payments is not much better than the social welfare obtained 
from a very simple class of pricing model.

Personally, I find the analysis in the paper to be quite interesting 
but, unless I'm missing something, it doesn't seem to be related to the 
current discussion.

- Matt Lepinski :->

Roy, Radhika R. wrote:

>Folks:
>
>I am enclosing an URL for the paper, "Incentive for Large Peer-to-Peer
>System," IEEE SAC, May 2006. 
>
>http://ieeexplore.ieee.org/iel5/49/34142/01626429.pdf
>
>Hope that we can take some rest after so much hot discussions on this
>topic.
>
>Radhika


From slavitch at gmail.com  Tue Dec  5 15:52:10 2006
From: slavitch at gmail.com (Michael Slavitch)
Date: Tue, 5 Dec 2006 15:52:10 -0500
Subject: [p2p-sip] Where are we "getting wrapped around the DNS axle"
In-Reply-To: <62D92A9A02BCC845B202323D49A48426230F0A@0591-ITS-EXMP02.us.saic.com>
References: <4575CDC2.7050005@bbn.com>
	<62D92A9A02BCC845B202323D49A48426230F0A@0591-ITS-EXMP02.us.saic.com>
Message-ID: <1fc32c8f0612051252u35c79473g8e2bbf4cbe89fbac@mail.gmail.com>

"The P2P system is a different paradigm and we have to
look into this with a little different way with different assumptions
and perspectives.

Otherwise, we will miss a lot of things as people did during the
pre-Internet period."

Boiled down to the essentials.  Thank you.


The fact that it describes revenue models should not go missed on this
or any audience.

Regards
Michael

From dbarrett at quinthar.com  Tue Dec  5 16:35:41 2006
From: dbarrett at quinthar.com (David Barrett)
Date: Tue, 5 Dec 2006 13:35:41 -0800
Subject: [p2p-sip] Where are we "getting wrapped around the DNS axle"
In-Reply-To: <457555E5.2050309@telecomitalia.it>
Message-ID: <05ef01c718b5$506df880$0202a8c0@Quinthar>

> -----Original Message-----
> From: Enrico Marocco [mailto:enrico.marocco at telecomitalia.it]
> Subject: Re: [p2p-sip] Where are we "getting wrapped around the DNS axle"
> 
> > In other words, do you agree:
> >
> > 	1) P2P-SIP must guarantee that each name resolves to a single user
> 
> Generally speaking, no.  The peer protocol must guarantee that the name
> used by a user enrolled within an overlay could not be used by another
> user in the _same_ overlay.

I'll grant it can be useful to have a totally unauthenticated, insecure
namespace where you can be Enrico for a while and then the moment you log
off I can log on as Enrico and be indistinguishable to everyone else.  I'd
wager this most often works in an ad hoc or "trade show" environment, where
mDNS is totally adequate.

However, I'm assuming a more valuable use case is an authenticated world
that has long-lived names verified in a secure manner.  And in this world,
DNS is totally adequate, for networks as small as a few fixed nodes in the
jungle, or as big as the entire internet.

What I'm waiting to hear about is a critical set of users who:

	- Don't need authentication
	- Can't use mDNS

Or, a critical set of users who:

	- Do need authentication
	- Can't use DNS

I'll totally grant we can construct a network to fit either of these.  For
the first, set up a network in the jungle and explicitly disable multicast.
For the second, set up a private network with a thousand nodes, and refuse
to install a DNS server (basically, disable all other network activity
except Gnutella and eDonkey).

But these don't seem like realistic use cases.  In the first, just tell
these jungle warriors "enable multicast" and they'll be fine.  In the
second, they're going to need a DNS server anyway, so just tell them to use
that.


> > 	2) Multiple users want the same name
> 
> Possibly, but if an enrollment mechanism is _deployed_ in the overlay
> (please note that it may be standardized as for jabber, or not), only
> the firstly enrolled user can use it.

Until he logs off or temporarily loses network connectivity, in which anyone
can steal his name.  (Though I guess we could have him generate a private
key and the first to register a public key with a name "wins" and if you
lose that private key you're basically screwed because there's no "password
change" feature.)

-david


From dbarrett at quinthar.com  Tue Dec  5 16:41:18 2006
From: dbarrett at quinthar.com (David Barrett)
Date: Tue, 5 Dec 2006 13:41:18 -0800
Subject: [p2p-sip] DNS for inter-domain? Yes; intra-domain? No
In-Reply-To: <7b683f3f0612050705q64a08b6gfdadb62f1cd492c2@mail.gmail.com>
Message-ID: <060b01c718b6$1b94a770$0202a8c0@Quinthar>

 

  _____  

From: Eunsoo Shim
Subject: [p2p-sip] DNS for inter-domain? Yes; intra-domain? No

 

So my understanding is that what P2PSIP group is trying to do is designing
mechanisms for P2P-style intra-domain location lookup where location records
are distributed among peers rather than in a central location server.

.

Am I missing any important aspect?



I agree, this is what P2P-SIP is trying to do.  I'm just not certain why,
because mDNS can do the same thing in a way that is already standardized,
already deployed, and already works.  Basically, for every real-world case,
it seems mDNS is adequate for ad-hoc unauthenticated namespaces, and DNS is
adequate for global, authenticated namespaces.  There is already a seamless
interface to graduate from mDNS to DNS depending upon what is available; it
already covers the whole range of use cases (minus the extreme security
cases, which I believe remain unsolved problems by any means).

 

I don't dispute you could make a DHT bootstrapped with UPnP or whatever.
You can build a lot of things.  I just don't see why we should bother when
there are easier, adequate technologies available.

 

-david

-------------- next part --------------
An HTML attachment was scrubbed...
URL: https://lists.cs.columbia.edu/pipermail/p2p-sip/attachments/20061205/ef131e88/attachment.html

From dbarrett at quinthar.com  Tue Dec  5 16:44:26 2006
From: dbarrett at quinthar.com (David Barrett)
Date: Tue, 5 Dec 2006 13:44:26 -0800
Subject: [p2p-sip] Avoiding Obsolete DHT Technologies
In-Reply-To: <024901c71883$dc6d4700$640fa8c0@cis.neustar.com>
Message-ID: <061001c718b6$89416fb0$0202a8c0@Quinthar>

> -----Original Message-----
> From: Brian Rosen [mailto:br at brianrosen.net]
> Subject: RE: [p2p-sip] Avoiding Obsolete DHT Technologies

> The question to the group is, does that solve our requirements?  I don't
> think it does.  It may solve a different set of requirements, and I have
> suggested that David et. al. write the idea up as an Internet Draft and
> submit it to sipping.  It may be an interesting enough concept that it is
> taken up as a fairly straightforward BCP thing that uses existing
> mechanisms
> to get a different form of plan jane sip.

I agree it's not a wide departure from how SIP already works, and I think
that's a strength.  But I still don't see which use cases (other than the
extreme security ones) it fails to meet.

-david


From dbarrett at quinthar.com  Tue Dec  5 16:51:09 2006
From: dbarrett at quinthar.com (David Barrett)
Date: Tue, 5 Dec 2006 13:51:09 -0800
Subject: [p2p-sip] Avoiding Obsolete DHT Technologies
In-Reply-To: <45758ECE.40900@bbn.com>
Message-ID: <061701c718b7$79e76c80$0202a8c0@Quinthar>

> -----Original Message-----
> From: Matt Lepinski [mailto:mlepinski at bbn.com]
> Cc: 'P2P-SIP'
> Subject: Re: [p2p-sip] Avoiding Obsolete DHT Technologies
> 
> I've seen the term "Obsolete" come up several times in this thread. In
> particular, it seems that some participants in this discussion
> (including David Barett whose post I quote below) believe that a DHT
> that uses a chord-like ring topology is an obsolete technology. If I
> understand correctly, the claim is that such DHT technology is incapable
> of dealing with the churn inherent in modern P2P use cases.
> 
> My question to David (or anyone else on the list who is knowledge about
> DHT technology) is:
> Is it the "ring-topology overlay" which is has become obsolete?

I won't pretend to be a DHT expert, but I don't think Chord has been proven
in the real world at any large scale, and has largely been superseded even
in research circles.  The only one that has been proven to work is Petar's
Kademlia (eDonkey, eMule, Overnet, etc), so if we're to pick any, that seems
like the place to start.

But my real point is that we shouldn't use a DHT at all if there are easier
options available that are already standardized, such as mDNS/DNS.

-david


From br at brianrosen.net  Tue Dec  5 17:41:27 2006
From: br at brianrosen.net (Brian Rosen)
Date: Tue, 5 Dec 2006 17:41:27 -0500
Subject: [p2p-sip] Where are we "getting wrapped around the DNS axle"
In-Reply-To: <05ef01c718b5$506df880$0202a8c0@Quinthar>
Message-ID: <03a801c718be$825a7e50$640fa8c0@cis.neustar.com>

I think the elephant in the room you keep ignoring is the desire to control
who can call me.  The point of an overlay network for many people is to
restrict who can get to me.  The security mechanisms inside a domain are a
whole lot different from the Internet as a whole.  By forcing use of DNS,
you are requiring that I set up some completely separate system to control
access to me.  Within a DHT, only people inside the domain (who have
credentials for the domain) can find out about me, find out how to contact
me, and actually make contact.  Now, admittedly, some of that is security by
obscurity that I don't subscribe to, but most if it is genuine security by
restricting access to people granted credentials in the domain.

Brian

> -----Original Message-----
> From: p2p-sip-bounces at cs.columbia.edu [mailto:p2p-sip-
> bounces at cs.columbia.edu] On Behalf Of David Barrett
> Sent: Tuesday, December 05, 2006 4:36 PM
> To: 'Enrico Marocco'
> Cc: 'P2P-SIP'; 'David A. Bryan'
> Subject: Re: [p2p-sip] Where are we "getting wrapped around the DNS axle"
> 
> > -----Original Message-----
> > From: Enrico Marocco [mailto:enrico.marocco at telecomitalia.it]
> > Subject: Re: [p2p-sip] Where are we "getting wrapped around the DNS
> axle"
> >
> > > In other words, do you agree:
> > >
> > > 	1) P2P-SIP must guarantee that each name resolves to a single user
> >
> > Generally speaking, no.  The peer protocol must guarantee that the name
> > used by a user enrolled within an overlay could not be used by another
> > user in the _same_ overlay.
> 
> I'll grant it can be useful to have a totally unauthenticated, insecure
> namespace where you can be Enrico for a while and then the moment you log
> off I can log on as Enrico and be indistinguishable to everyone else.  I'd
> wager this most often works in an ad hoc or "trade show" environment,
> where
> mDNS is totally adequate.
> 
> However, I'm assuming a more valuable use case is an authenticated world
> that has long-lived names verified in a secure manner.  And in this world,
> DNS is totally adequate, for networks as small as a few fixed nodes in the
> jungle, or as big as the entire internet.
> 
> What I'm waiting to hear about is a critical set of users who:
> 
> 	- Don't need authentication
> 	- Can't use mDNS
> 
> Or, a critical set of users who:
> 
> 	- Do need authentication
> 	- Can't use DNS
> 
> I'll totally grant we can construct a network to fit either of these.  For
> the first, set up a network in the jungle and explicitly disable
> multicast.
> For the second, set up a private network with a thousand nodes, and refuse
> to install a DNS server (basically, disable all other network activity
> except Gnutella and eDonkey).
> 
> But these don't seem like realistic use cases.  In the first, just tell
> these jungle warriors "enable multicast" and they'll be fine.  In the
> second, they're going to need a DNS server anyway, so just tell them to
> use
> that.
> 
> 
> > > 	2) Multiple users want the same name
> >
> > Possibly, but if an enrollment mechanism is _deployed_ in the overlay
> > (please note that it may be standardized as for jabber, or not), only
> > the firstly enrolled user can use it.
> 
> Until he logs off or temporarily loses network connectivity, in which
> anyone
> can steal his name.  (Though I guess we could have him generate a private
> key and the first to register a public key with a name "wins" and if you
> lose that private key you're basically screwed because there's no
> "password
> change" feature.)
> 
> -david
> 
> _______________________________________________
> p2p-sip mailing list
> p2p-sip at cs.columbia.edu
> https://lists.cs.columbia.edu/cucslists/listinfo/p2p-sip


From eunsooshim at gmail.com  Tue Dec  5 18:04:27 2006
From: eunsooshim at gmail.com (Eunsoo Shim)
Date: Tue, 5 Dec 2006 18:04:27 -0500
Subject: [p2p-sip] DNS for inter-domain? Yes; intra-domain? No
In-Reply-To: <060b01c718b6$1b94a770$0202a8c0@Quinthar>
References: <7b683f3f0612050705q64a08b6gfdadb62f1cd492c2@mail.gmail.com>
	<060b01c718b6$1b94a770$0202a8c0@Quinthar>
Message-ID: <7b683f3f0612051504wedf437ci789da9d601660d7c@mail.gmail.com>

I understand that mDNS can be used in ad-hoc networks where multicast works.
Some of us, including me, have discussed about using mDNS and noticed that
there is little to do to use mDNS or any other multicast-based discovery
protocol.
Those people decided not to focus on the approach because assuming multicast
availability was not general and they did not want to limit P2PSIP into such
network environments.
They said such an approach (using mDNS or similar ones) could be discussed
maybe in SIPPING.

The proposed charter is targeting to address intra-domain lookup that works
even if there is no multicast support.

Using DNS for intra-domain lookup is not different from using a central
server, which is not interesting to many of us, I believe. So transition
from mDNS to DNS is not appealing, either.

What do you say?
Thanks.

Eunsoo


On 12/5/06, David Barrett <dbarrett at quinthar.com> wrote:
>
>
>   ------------------------------
>
> *From:* Eunsoo Shim
> *Subject:* [p2p-sip] DNS for inter-domain? Yes; intra-domain? No
>
>
>
> So my understanding is that what P2PSIP group is trying to do is designing
> mechanisms for P2P-style intra-domain location lookup where location records
> are distributed among peers rather than in a central location server.
>
> ?
>
> Am I missing any important aspect?
>
>  I agree, this is what P2P-SIP is trying to do.  I'm just not certain why,
> because mDNS can do the same thing in a way that is already standardized,
> already deployed, and already works.  Basically, for every real-world case,
> it seems mDNS is adequate for ad-hoc unauthenticated namespaces, and DNS is
> adequate for global, authenticated namespaces.  There is already a seamless
> interface to graduate from mDNS to DNS depending upon what is available; it
> already covers the whole range of use cases (minus the extreme security
> cases, which I believe remain unsolved problems by any means).
>
>
>
> I don't dispute you could make a DHT bootstrapped with UPnP or whatever.
> You can build a lot of things.  I just don't see why we should bother when
> there are easier, adequate technologies available.
>
>
>
> -david
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: https://lists.cs.columbia.edu/pipermail/p2p-sip/attachments/20061205/540d24ec/attachment-0001.html

From dbarrett at quinthar.com  Tue Dec  5 18:51:23 2006
From: dbarrett at quinthar.com (David Barrett)
Date: Tue, 5 Dec 2006 15:51:23 -0800
Subject: [p2p-sip] Where are we "getting wrapped around the DNS axle"
In-Reply-To: <03a801c718be$825a7e50$640fa8c0@cis.neustar.com>
Message-ID: <062101c718c8$45495c70$0202a8c0@Quinthar>

I'm sorry, I don't follow.   The whole notion that someone else controls who
I can talk to (and can talk to me) is precisely *opposite* our goals.  If
your UA doesn't want to take calls from anyone outside of *@group.com, then
it has every ability to define, manage, and enforce that access control
policy at the edge.

Furthermore, if you for some reason don't want to control at the edge, it's
vastly *easier* to do it with normal directories than with certificates and
DHTs.

Perhaps you can give a real-world use case of what you're describing, and
explain how you'd intend to solve it with a DHT?  Then I can present how I'd
solve it with DNS/mDNS, and we can contrast which is more straightforward --
to implement, to administer, and to use.

DNS and DHT are equivalent services.  Both take a name, and spit out
information (generally IPs, though it can be a variety of records).  One is
a drop-in replacement for the other.  The only difference is one exists on a
massive, standard, secure, universal, global scale today, and the other
doesn't.

I think whatever DHT-based solution you present, if you do a search/replace
of DHT to DNS, that'll be basically what I'm proposing.

-david

> -----Original Message-----
> From: Brian Rosen [mailto:br at brianrosen.net]
> Sent: Tuesday, December 05, 2006 2:41 PM
> To: 'David Barrett'; 'Enrico Marocco'
> Cc: 'P2P-SIP'; 'David A. Bryan'
> Subject: RE: [p2p-sip] Where are we "getting wrapped around the DNS axle"
> 
> I think the elephant in the room you keep ignoring is the desire to
> control
> who can call me.  The point of an overlay network for many people is to
> restrict who can get to me.  The security mechanisms inside a domain are a
> whole lot different from the Internet as a whole.  By forcing use of DNS,
> you are requiring that I set up some completely separate system to control
> access to me.  Within a DHT, only people inside the domain (who have
> credentials for the domain) can find out about me, find out how to contact
> me, and actually make contact.  Now, admittedly, some of that is security
> by
> obscurity that I don't subscribe to, but most if it is genuine security by
> restricting access to people granted credentials in the domain.
> 
> Brian
> 
> > -----Original Message-----
> > From: p2p-sip-bounces at cs.columbia.edu [mailto:p2p-sip-
> > bounces at cs.columbia.edu] On Behalf Of David Barrett
> > Sent: Tuesday, December 05, 2006 4:36 PM
> > To: 'Enrico Marocco'
> > Cc: 'P2P-SIP'; 'David A. Bryan'
> > Subject: Re: [p2p-sip] Where are we "getting wrapped around the DNS
> axle"
> >
> > > -----Original Message-----
> > > From: Enrico Marocco [mailto:enrico.marocco at telecomitalia.it]
> > > Subject: Re: [p2p-sip] Where are we "getting wrapped around the DNS
> > axle"
> > >
> > > > In other words, do you agree:
> > > >
> > > > 	1) P2P-SIP must guarantee that each name resolves to a
single
> user
> > >
> > > Generally speaking, no.  The peer protocol must guarantee that the
> name
> > > used by a user enrolled within an overlay could not be used by another
> > > user in the _same_ overlay.
> >
> > I'll grant it can be useful to have a totally unauthenticated, insecure
> > namespace where you can be Enrico for a while and then the moment you
> log
> > off I can log on as Enrico and be indistinguishable to everyone else.
> I'd
> > wager this most often works in an ad hoc or "trade show" environment,
> > where
> > mDNS is totally adequate.
> >
> > However, I'm assuming a more valuable use case is an authenticated world
> > that has long-lived names verified in a secure manner.  And in this
> world,
> > DNS is totally adequate, for networks as small as a few fixed nodes in
> the
> > jungle, or as big as the entire internet.
> >
> > What I'm waiting to hear about is a critical set of users who:
> >
> > 	- Don't need authentication
> > 	- Can't use mDNS
> >
> > Or, a critical set of users who:
> >
> > 	- Do need authentication
> > 	- Can't use DNS
> >
> > I'll totally grant we can construct a network to fit either of these.
> For
> > the first, set up a network in the jungle and explicitly disable
> > multicast.
> > For the second, set up a private network with a thousand nodes, and
> refuse
> > to install a DNS server (basically, disable all other network activity
> > except Gnutella and eDonkey).
> >
> > But these don't seem like realistic use cases.  In the first, just tell
> > these jungle warriors "enable multicast" and they'll be fine.  In the
> > second, they're going to need a DNS server anyway, so just tell them to
> > use
> > that.
> >
> >
> > > > 	2) Multiple users want the same name
> > >
> > > Possibly, but if an enrollment mechanism is _deployed_ in the overlay
> > > (please note that it may be standardized as for jabber, or not), only
> > > the firstly enrolled user can use it.
> >
> > Until he logs off or temporarily loses network connectivity, in which
> > anyone
> > can steal his name.  (Though I guess we could have him generate a
> private
> > key and the first to register a public key with a name "wins" and if you
> > lose that private key you're basically screwed because there's no
> > "password
> > change" feature.)
> >
> > -david
> >
> > _______________________________________________
> > p2p-sip mailing list
> > p2p-sip at cs.columbia.edu
> > https://lists.cs.columbia.edu/cucslists/listinfo/p2p-sip


From dbarrett at quinthar.com  Tue Dec  5 19:02:02 2006
From: dbarrett at quinthar.com (David Barrett)
Date: Tue, 5 Dec 2006 16:02:02 -0800
Subject: [p2p-sip] DNS for inter-domain? Yes; intra-domain? No
In-Reply-To: <7b683f3f0612051504wedf437ci789da9d601660d7c@mail.gmail.com>
Message-ID: <063a01c718c9$c64285d0$0202a8c0@Quinthar>

Does the actual, real-world availability of multicast factor into the
decision as to whether or not to use it?

 

I mean, we *could* assume that every odd IP address won't work, and then
design around that.  But odd IP addresses *do* work.

 

Likewise, multicast *is* (or could easily be made) available for all of the
real-world use cases that demand it.

 

It just seems like there are much more interesting *actual* problems to
solve than those that don't affect real users.  But if you're excited about
solving the non-multicast ad-hoc P2P-SIP problem on the assumption that
someday, somebody will have reason to use it (even though we can't really
brainstorm anybody who would use it today), then I guess that's a cool
research project.

 

But it's a *research* project, not an IETF standardization exercise.  DNS
and mDNS work everywhere the *actual* users are, and thus I think the IETF
should focus on them, first.

 

-david

 

  _____  

From: Eunsoo Shim [mailto:eunsooshim at gmail.com] 
Sent: Tuesday, December 05, 2006 3:04 PM
To: David Barrett
Cc: P2P-SIP
Subject: Re: [p2p-sip] DNS for inter-domain? Yes; intra-domain? No

 

I understand that mDNS can be used in ad-hoc networks where multicast works.
Some of us, including me, have discussed about using mDNS and noticed that
there is little to do to use mDNS or any other multicast-based discovery
protocol. 
Those people decided not to focus on the approach because assuming multicast
availability was not general and they did not want to limit P2PSIP into such
network environments.
They said such an approach (using mDNS or similar ones) could be discussed
maybe in SIPPING. 

The proposed charter is targeting to address intra-domain lookup that works
even if there is no multicast support.

Using DNS for intra-domain lookup is not different from using a central
server, which is not interesting to many of us, I believe. So transition
from mDNS to DNS is not appealing, either. 

What do you say?
Thanks.

Eunsoo

On 12/5/06, David Barrett <dbarrett at quinthar.com> wrote: 

 

  _____  

From: Eunsoo Shim 
Subject: [p2p-sip] DNS for inter-domain? Yes; intra-domain? No

 

So my understanding is that what P2PSIP group is trying to do is designing
mechanisms for P2P-style intra-domain location lookup where location records
are distributed among peers rather than in a central location server.

.

Am I missing any important aspect?

I agree, this is what P2P-SIP is trying to do.  I'm just not certain why,
because mDNS can do the same thing in a way that is already standardized,
already deployed, and already works.  Basically, for every real-world case,
it seems mDNS is adequate for ad-hoc unauthenticated namespaces, and DNS is
adequate for global, authenticated namespaces.  There is already a seamless
interface to graduate from mDNS to DNS depending upon what is available; it
already covers the whole range of use cases (minus the extreme security
cases, which I believe remain unsolved problems by any means).

 

I don't dispute you could make a DHT bootstrapped with UPnP or whatever.
You can build a lot of things.  I just don't see why we should bother when
there are easier, adequate technologies available.

 

-david

 

-------------- next part --------------
An HTML attachment was scrubbed...
URL: https://lists.cs.columbia.edu/pipermail/p2p-sip/attachments/20061205/52c1335a/attachment.html

From hgs at cs.columbia.edu  Tue Dec  5 19:20:15 2006
From: hgs at cs.columbia.edu (Henning Schulzrinne)
Date: Tue, 5 Dec 2006 19:20:15 -0500
Subject: [p2p-sip] DNS for inter-domain? Yes; intra-domain? No
In-Reply-To: <063a01c718c9$c64285d0$0202a8c0@Quinthar>
References: <063a01c718c9$c64285d0$0202a8c0@Quinthar>
Message-ID: <811C3452-A1FF-4415-90B1-B54AEE44F091@cs.columbia.edu>

I have to admit that I'm confused by this whole discussion. I thought  
we had agreement that

(1) specifying mDNS + SIP operation is useful (and hopefully simple),  
but easily handled within the framework of SIPPING;

(2) specifying dynamic DNS + SIP might be useful, subject to working  
out issues such as proxy discovery, authentication and the like and  
that interested parties should work out, at least to the extent of  
identifying the pieces that might need work (as a random example: can  
I expect my SIP Digest secret to work as a DNS update secret?). If,  
as claimed, this is easy to do, SIPPING is again the natural home for  
such a BCP effort, with the DNS-related WGs as the home for any DNS  
extensions that might prove necessary.

Maybe those individuals interested in solving (2) should form an  
informal design team and get beyond the hand-waving stage and  
identify the problems, if any, that would have to be addressed in  
either protocol or operational standards. I have unfortunately seen  
little such effort so far, although I suspect that such level of  
detail isn't terribly well suited to be hashed out in front of  
several hundred mailing list members.


We also have had 2 BOFs and numerous ad-hoc meetings on this topic,  
with a rough consensus on the current charter. Are you trying to keep  
people from working on a specific topic you consider inferior?

We seem to be rehashing the same arguments again and again.

Henning



On Dec 5, 2006, at 7:02 PM, David Barrett wrote:

> Does the actual, real-world availability of multicast factor into  
> the decision as to whether or not to use it?
>
>
>
> I mean, we *could* assume that every odd IP address won?t work, and  
> then design around that.  But odd IP addresses *do* work.
>
>
>
> Likewise, multicast *is* (or could easily be made) available for  
> all of the real-world use cases that demand it.
>
>
>
> It just seems like there are much more interesting *actual*  
> problems to solve than those that don?t affect real users.  But if  
> you?re excited about solving the non-multicast ad-hoc P2P-SIP  
> problem on the assumption that someday, somebody will have reason  
> to use it (even though we can?t really brainstorm anybody who would  
> use it today), then I guess that?s a cool research project.
>
>
>
> But it?s a *research* project, not an IETF standardization  
> exercise.  DNS and mDNS work everywhere the *actual* users are, and  
> thus I think the IETF should focus on them, first.
>
>
>
> -david
>
>
>
> From: Eunsoo Shim [mailto:eunsooshim at gmail.com]
> Sent: Tuesday, December 05, 2006 3:04 PM
> To: David Barrett
> Cc: P2P-SIP
> Subject: Re: [p2p-sip] DNS for inter-domain? Yes; intra-domain? No
>
>
>
> I understand that mDNS can be used in ad-hoc networks where  
> multicast works.
> Some of us, including me, have discussed about using mDNS and  
> noticed that there is little to do to use mDNS or any other  
> multicast-based discovery protocol.
> Those people decided not to focus on the approach because assuming  
> multicast availability was not general and they did not want to  
> limit P2PSIP into such network environments.
> They said such an approach (using mDNS or similar ones) could be  
> discussed maybe in SIPPING.
>
> The proposed charter is targeting to address intra-domain lookup  
> that works even if there is no multicast support.
>
> Using DNS for intra-domain lookup is not different from using a  
> central server, which is not interesting to many of us, I believe.  
> So transition from mDNS to DNS is not appealing, either.
>
> What do you say?
> Thanks.
>
> Eunsoo
>
> On 12/5/06, David Barrett <dbarrett at quinthar.com> wrote:
>
>
>
> From: Eunsoo Shim
> Subject: [p2p-sip] DNS for inter-domain? Yes; intra-domain? No
>
>
>
> So my understanding is that what P2PSIP group is trying to do is  
> designing mechanisms for P2P-style intra-domain location lookup  
> where location records are distributed among peers rather than in a  
> central location server.
>
> ?
>
> Am I missing any important aspect?
>
> I agree, this is what P2P-SIP is trying to do.  I'm just not  
> certain why, because mDNS can do the same thing in a way that is  
> already standardized, already deployed, and already works.   
> Basically, for every real-world case, it seems mDNS is adequate for  
> ad-hoc unauthenticated namespaces, and DNS is adequate for global,  
> authenticated namespaces.  There is already a seamless interface to  
> graduate from mDNS to DNS depending upon what is available; it  
> already covers the whole range of use cases (minus the extreme  
> security cases, which I believe remain unsolved problems by any  
> means).
>
>
>
> I don't dispute you could make a DHT bootstrapped with UPnP or  
> whatever.  You can build a lot of things.  I just don't see why we  
> should bother when there are easier, adequate technologies available.
>
>
>
> -david
>
>
>
> _______________________________________________
> p2p-sip mailing list
> p2p-sip at cs.columbia.edu
> https://lists.cs.columbia.edu/cucslists/listinfo/p2p-sip

-------------- next part --------------
An HTML attachment was scrubbed...
URL: https://lists.cs.columbia.edu/pipermail/p2p-sip/attachments/20061205/7ffc85ac/attachment-0001.html

From br at brianrosen.net  Tue Dec  5 20:02:15 2006
From: br at brianrosen.net (Brian Rosen)
Date: Tue, 5 Dec 2006 20:02:15 -0500
Subject: [p2p-sip] Where are we "getting wrapped around the DNS axle"
In-Reply-To: <062101c718c8$45495c70$0202a8c0@Quinthar>
Message-ID: <03f101c718d2$30c8e720$640fa8c0@cis.neustar.com>


Okay, let's walk through this:

Someone, doesn't matter who, starts up a domain called p2ps.com.  They put
up a website, and offer to give out credentials for their p2psip domain.
They have a membership, code of conduct, and policies you like, and a
business model you can deal with, so you sign up.

p2ps gives you a credential.  That credential will entitle you to join the
DHT.  If you don't have one, you can't join the DHT, and you can't find out
who else is in the domain, or how to contact them.  All the messages within
the DHT are encrypted, and the only way you can get at any of them is to
present credentials.  

When you place a call, to anyone in the p2ps domain, you present your
credentials issued by the domain with your INVITE.  You don't accept a call
from anyone without the domain credentials.  In fact, NO signaling will
enter any element within the domain without the proper credentials, and you,
as a peer, can validate the credentials of any DHT message or any SIP
message with those credentials.

Once you get the credential, the only centralized service needed is whatever
it is that allows the domain to revoke credentials, if any, and that can be
cached by the peers.  The name to address mechanism is impervious to a lot
of damage because nothing is authoritative save the credential; it's all
distributed.  Any isolated set of peers can communicate with each other, and
as soon as connections to another peer or set of peers is restored
communication with them becomes possible.

Brian


> -----Original Message-----
> From: David Barrett [mailto:dbarrett at quinthar.com]
> Sent: Tuesday, December 05, 2006 6:51 PM
> To: 'Brian Rosen'; 'Enrico Marocco'
> Cc: 'P2P-SIP'; 'David A. Bryan'
> Subject: RE: [p2p-sip] Where are we "getting wrapped around the DNS axle"
> 
> I'm sorry, I don't follow.   The whole notion that someone else controls
> who
> I can talk to (and can talk to me) is precisely *opposite* our goals.  If
> your UA doesn't want to take calls from anyone outside of *@group.com,
> then
> it has every ability to define, manage, and enforce that access control
> policy at the edge.
> 
> Furthermore, if you for some reason don't want to control at the edge,
> it's
> vastly *easier* to do it with normal directories than with certificates
> and
> DHTs.
> 
> Perhaps you can give a real-world use case of what you're describing, and
> explain how you'd intend to solve it with a DHT?  Then I can present how
> I'd
> solve it with DNS/mDNS, and we can contrast which is more straightforward
> --
> to implement, to administer, and to use.
> 
> DNS and DHT are equivalent services.  Both take a name, and spit out
> information (generally IPs, though it can be a variety of records).  One
> is
> a drop-in replacement for the other.  The only difference is one exists on
> a
> massive, standard, secure, universal, global scale today, and the other
> doesn't.
> 
> I think whatever DHT-based solution you present, if you do a
> search/replace
> of DHT to DNS, that'll be basically what I'm proposing.
> 
> -david
> 
> > -----Original Message-----
> > From: Brian Rosen [mailto:br at brianrosen.net]
> > Sent: Tuesday, December 05, 2006 2:41 PM
> > To: 'David Barrett'; 'Enrico Marocco'
> > Cc: 'P2P-SIP'; 'David A. Bryan'
> > Subject: RE: [p2p-sip] Where are we "getting wrapped around the DNS
> axle"
> >
> > I think the elephant in the room you keep ignoring is the desire to
> > control
> > who can call me.  The point of an overlay network for many people is to
> > restrict who can get to me.  The security mechanisms inside a domain are
> a
> > whole lot different from the Internet as a whole.  By forcing use of
> DNS,
> > you are requiring that I set up some completely separate system to
> control
> > access to me.  Within a DHT, only people inside the domain (who have
> > credentials for the domain) can find out about me, find out how to
> contact
> > me, and actually make contact.  Now, admittedly, some of that is
> security
> > by
> > obscurity that I don't subscribe to, but most if it is genuine security
> by
> > restricting access to people granted credentials in the domain.
> >
> > Brian
> >
> > > -----Original Message-----
> > > From: p2p-sip-bounces at cs.columbia.edu [mailto:p2p-sip-
> > > bounces at cs.columbia.edu] On Behalf Of David Barrett
> > > Sent: Tuesday, December 05, 2006 4:36 PM
> > > To: 'Enrico Marocco'
> > > Cc: 'P2P-SIP'; 'David A. Bryan'
> > > Subject: Re: [p2p-sip] Where are we "getting wrapped around the DNS
> > axle"
> > >
> > > > -----Original Message-----
> > > > From: Enrico Marocco [mailto:enrico.marocco at telecomitalia.it]
> > > > Subject: Re: [p2p-sip] Where are we "getting wrapped around the DNS
> > > axle"
> > > >
> > > > > In other words, do you agree:
> > > > >
> > > > > 	1) P2P-SIP must guarantee that each name resolves to a
> single
> > user
> > > >
> > > > Generally speaking, no.  The peer protocol must guarantee that the
> > name
> > > > used by a user enrolled within an overlay could not be used by
> another
> > > > user in the _same_ overlay.
> > >
> > > I'll grant it can be useful to have a totally unauthenticated,
> insecure
> > > namespace where you can be Enrico for a while and then the moment you
> > log
> > > off I can log on as Enrico and be indistinguishable to everyone else.
> > I'd
> > > wager this most often works in an ad hoc or "trade show" environment,
> > > where
> > > mDNS is totally adequate.
> > >
> > > However, I'm assuming a more valuable use case is an authenticated
> world
> > > that has long-lived names verified in a secure manner.  And in this
> > world,
> > > DNS is totally adequate, for networks as small as a few fixed nodes in
> > the
> > > jungle, or as big as the entire internet.
> > >
> > > What I'm waiting to hear about is a critical set of users who:
> > >
> > > 	- Don't need authentication
> > > 	- Can't use mDNS
> > >
> > > Or, a critical set of users who:
> > >
> > > 	- Do need authentication
> > > 	- Can't use DNS
> > >
> > > I'll totally grant we can construct a network to fit either of these.
> > For
> > > the first, set up a network in the jungle and explicitly disable
> > > multicast.
> > > For the second, set up a private network with a thousand nodes, and
> > refuse
> > > to install a DNS server (basically, disable all other network activity
> > > except Gnutella and eDonkey).
> > >
> > > But these don't seem like realistic use cases.  In the first, just
> tell
> > > these jungle warriors "enable multicast" and they'll be fine.  In the
> > > second, they're going to need a DNS server anyway, so just tell them
> to
> > > use
> > > that.
> > >
> > >
> > > > > 	2) Multiple users want the same name
> > > >
> > > > Possibly, but if an enrollment mechanism is _deployed_ in the
> overlay
> > > > (please note that it may be standardized as for jabber, or not),
> only
> > > > the firstly enrolled user can use it.
> > >
> > > Until he logs off or temporarily loses network connectivity, in which
> > > anyone
> > > can steal his name.  (Though I guess we could have him generate a
> > private
> > > key and the first to register a public key with a name "wins" and if
> you
> > > lose that private key you're basically screwed because there's no
> > > "password
> > > change" feature.)
> > >
> > > -david
> > >
> > > _______________________________________________
> > > p2p-sip mailing list
> > > p2p-sip at cs.columbia.edu
> > > https://lists.cs.columbia.edu/cucslists/listinfo/p2p-sip


From br at brianrosen.net  Tue Dec  5 20:08:08 2006
From: br at brianrosen.net (Brian Rosen)
Date: Tue, 5 Dec 2006 20:08:08 -0500
Subject: [p2p-sip] DNS for inter-domain? Yes; intra-domain? No
In-Reply-To: <063a01c718c9$c64285d0$0202a8c0@Quinthar>
Message-ID: <03f401c718d2$ffbcf1c0$640fa8c0@cis.neustar.com>

I?ve spent a good deal of time dealing with systems that depend on multicast
and I conclude that multicast is a very wonderful thing, every network can,
and should support it, and almost none actually do.

Multicast DOES NOT work in 99.9% of networks beyond a subnet.  It?s a lot
more than a research project to fix that.  IPTV may cause some networks to
allow multicast in some very restricted circumstances.  It's a step, but you
won't be able to use mDNS within those networks as I understand deployment
plans.

I think in some really interesting ad-hoc networks we may be able to deploy
multicast.  I would not be willing to place bets on that.  They certainly
don't support it now.

Brian

________________________________________
From: p2p-sip-bounces at cs.columbia.edu
[mailto:p2p-sip-bounces at cs.columbia.edu] On Behalf Of David Barrett
Sent: Tuesday, December 05, 2006 7:02 PM
To: 'Eunsoo Shim'
Cc: 'P2P-SIP'
Subject: Re: [p2p-sip] DNS for inter-domain? Yes; intra-domain? No

Does the actual, real-world availability of multicast factor into the
decision as to whether or not to use it?

I mean, we *could* assume that every odd IP address won?t work, and then
design around that.? But odd IP addresses *do* work.

Likewise, multicast *is* (or could easily be made) available for all of the
real-world use cases that demand it.

It just seems like there are much more interesting *actual* problems to
solve than those that don?t affect real users.? But if you?re excited about
solving the non-multicast ad-hoc P2P-SIP problem on the assumption that
someday, somebody will have reason to use it (even though we can?t really
brainstorm anybody who would use it today), then I guess that?s a cool
research project.

But it?s a *research* project, not an IETF standardization exercise.? DNS
and mDNS work everywhere the *actual* users are, and thus I think the IETF
should focus on them, first.

-david

________________________________________
From: Eunsoo Shim [mailto:eunsooshim at gmail.com] 
Sent: Tuesday, December 05, 2006 3:04 PM
To: David Barrett
Cc: P2P-SIP
Subject: Re: [p2p-sip] DNS for inter-domain? Yes; intra-domain? No

I understand that mDNS can be used in ad-hoc networks where multicast works.
Some of us, including me, have discussed about using mDNS and noticed that
there is little to do to use mDNS or any other multicast-based discovery
protocol. 
Those people decided not to focus on the approach because assuming multicast
availability was not general and they did not want to limit P2PSIP into such
network environments.
They said such an approach (using mDNS or similar ones) could be discussed
maybe in SIPPING. 

The proposed charter is targeting to address intra-domain lookup that works
even if there is no multicast support.

Using DNS for intra-domain lookup is not different from using a central
server, which is not interesting to many of us, I believe. So transition
from mDNS to DNS is not appealing, either. 

What do you say?
Thanks.

Eunsoo
On 12/5/06, David Barrett <dbarrett at quinthar.com> wrote: 
?
________________________________________
From: Eunsoo Shim 
Subject: [p2p-sip] DNS for inter-domain? Yes; intra-domain? No
?
So my understanding is that what P2PSIP group is trying to do is designing
mechanisms for P2P-style intra-domain location lookup where location records
are distributed among peers rather than in a central location server.


Am I missing any important aspect?
I agree, this is what P2P-SIP is trying to do.? I'm just not certain why,
because mDNS can do the same thing in a way that is already standardized,
already deployed, and already works.? Basically, for every real-world case,
it seems mDNS is adequate for ad-hoc unauthenticated namespaces, and DNS is
adequate for global, authenticated namespaces.? There is already a seamless
interface to graduate from mDNS to DNS depending upon what is available; it
already covers the whole range of use cases (minus the extreme security
cases, which I believe remain unsolved problems by any means).
?
I don't dispute you could make a DHT bootstrapped with UPnP or whatever.?
You can build a lot of things.? I just don't see why we should bother when
there are easier, adequate technologies available.
?
-david




From slavitch at gmail.com  Tue Dec  5 20:25:34 2006
From: slavitch at gmail.com (Michael Slavitch)
Date: Tue, 5 Dec 2006 20:25:34 -0500
Subject: [p2p-sip] DNS for inter-domain? Yes; intra-domain? No
In-Reply-To: <03f401c718d2$ffbcf1c0$640fa8c0@cis.neustar.com>
References: <063a01c718c9$c64285d0$0202a8c0@Quinthar>
	<03f401c718d2$ffbcf1c0$640fa8c0@cis.neustar.com>
Message-ID: <1fc32c8f0612051725u4c05e51p88e2aadad7554409@mail.gmail.com>

This discussion ended a long time ago and it is now just a game of
loop-de-loop trolling.

From matthew at matthew.at  Tue Dec  5 21:02:11 2006
From: matthew at matthew.at (Matthew Kaufman)
Date: Tue, 5 Dec 2006 18:02:11 -0800
Subject: [p2p-sip] DNS for inter-domain? Yes; intra-domain? No
In-Reply-To: <03f401c718d2$ffbcf1c0$640fa8c0@cis.neustar.com>
Message-ID: <005401c718da$8b203770$029b0a0a@matthewdeskxps>


Brian RosenL
>
> Multicast DOES NOT work in 99.9% of networks beyond a subnet.  

Agreed. Absolutely.

> I think in some really interesting ad-hoc networks we may be able to
> deploy multicast.  I would not be willing to place bets on that.
> They certainly don't support it now.

All of the ad-hoc network scenarios that have been proposed are confined to
a single subnet, on which multicast will work. Therefore, on those networks
and in those use cases, zeroconf mDNS (not to be confused with network-wide
multicast DNS, about which I am not talking) will work.

Any ad-hoc network that is built which is larger will need to be built with
DHCP servers on each subnet which can provide the gateway address *and* DNS
server addresses pointing at local DNS servers which will also need to be
set up... and I say *need* to be set up because without that level of
infrastructure, a routed ad-hoc network can't run any of the *other*
applications it would have been deployed to support (web intranet access,
email, etc.) If those DNS servers can accept dynamic updates or are manually
configured appropriately, that covers the other half of the problem.

While it is clear that the mailing list has morphed from "discussion of
peer-to-peer issues with SIP" into "the official mailing list of a working
group which is set on using a DHT because that's what 'P2P' means to them",
the first topic could still be interesting. Maybe one of the subsets of
people should find a new mailing list.

Matthew Kaufman
matthew at matthew.at


From dbarrett at quinthar.com  Wed Dec  6 03:07:39 2006
From: dbarrett at quinthar.com (David Barrett)
Date: Wed, 6 Dec 2006 00:07:39 -0800
Subject: [p2p-sip] DNS for inter-domain? Yes; intra-domain? No
In-Reply-To: <811C3452-A1FF-4415-90B1-B54AEE44F091@cs.columbia.edu>
Message-ID: <066801c7190d$996faad0$0202a8c0@Quinthar>

I think the value of this discussion is in identifying and clarifying
misimpressions about how DNS and mDNS work (and do not work) so we can make
a more informed choice on whether or not to use them.  Already we've
clarified:

 

-          How can DNS records be automatically updated? (Through dynamic
DNS protocols)

-          How fast can DNS propagate?  (Extremely fast with short TTLs)

-          What if I have multiple SIP proxies? (Put multiple IPs into the
DNS record)

-          Is DNS only for IP addresses? (No, can store other records)

-          How does DNS work in on ad-hoc networks? (Using mDNS via
multicast)

-          Where is multicast supported? (Only everywhere it's needed - on
subnets)

-          Is mDNS an IETF standard? (It's on the standards track.)

-          Is mDNS actually in use? (It's the foundation of zeroconf, which
is used extensively by Apple)

-          Are there any use-cases that DHT supports that DNS/mDNS cannot?
(Apparently not.)

 

Lack of clarity on these points could call for the unnecessary dismissal of
these technologies on false premises, which I think nobody wants.

 

-david

 

  _____  

From: Henning Schulzrinne [mailto:hgs at cs.columbia.edu] 
Sent: Tuesday, December 05, 2006 4:20 PM
To: David Barrett
Cc: 'Eunsoo Shim'; 'P2P-SIP'
Subject: Re: [p2p-sip] DNS for inter-domain? Yes; intra-domain? No

 

I have to admit that I'm confused by this whole discussion. I thought we had
agreement that

 

(1) specifying mDNS + SIP operation is useful (and hopefully simple), but
easily handled within the framework of SIPPING;

 

(2) specifying dynamic DNS + SIP might be useful, subject to working out
issues such as proxy discovery, authentication and the like and that
interested parties should work out, at least to the extent of identifying
the pieces that might need work (as a random example: can I expect my SIP
Digest secret to work as a DNS update secret?). If, as claimed, this is easy
to do, SIPPING is again the natural home for such a BCP effort, with the
DNS-related WGs as the home for any DNS extensions that might prove
necessary.

 

Maybe those individuals interested in solving (2) should form an informal
design team and get beyond the hand-waving stage and identify the problems,
if any, that would have to be addressed in either protocol or operational
standards. I have unfortunately seen little such effort so far, although I
suspect that such level of detail isn't terribly well suited to be hashed
out in front of several hundred mailing list members.

 

 

We also have had 2 BOFs and numerous ad-hoc meetings on this topic, with a
rough consensus on the current charter. Are you trying to keep people from
working on a specific topic you consider inferior?

 

We seem to be rehashing the same arguments again and again. 

 

Henning

 

 

 

On Dec 5, 2006, at 7:02 PM, David Barrett wrote:

 

Does the actual, real-world availability of multicast factor into the
decision as to whether or not to use it?

 

I mean, we *could* assume that every odd IP address won't work, and then
design around that.  But odd IP addresses *do* work.

 

Likewise, multicast *is* (or could easily be made) available for all of the
real-world use cases that demand it.

 

It just seems like there are much more interesting *actual* problems to
solve than those that don't affect real users.  But if you're excited about
solving the non-multicast ad-hoc P2P-SIP problem on the assumption that
someday, somebody will have reason to use it (even though we can't really
brainstorm anybody who would use it today), then I guess that's a cool
research project.

 

But it's a *research* project, not an IETF standardization exercise.  DNS
and mDNS work everywhere the *actual* users are, and thus I think the IETF
should focus on them, first.

 

-david

 

  _____  

From: Eunsoo Shim [mailto:eunsooshim at gmail.com] 
Sent: Tuesday, December 05, 2006 3:04 PM
To: David Barrett
Cc: P2P-SIP
Subject: Re: [p2p-sip] DNS for inter-domain? Yes; intra-domain? No

 

I understand that mDNS can be used in ad-hoc networks where multicast works.
Some of us, including me, have discussed about using mDNS and noticed that
there is little to do to use mDNS or any other multicast-based discovery
protocol. 
Those people decided not to focus on the approach because assuming multicast
availability was not general and they did not want to limit P2PSIP into such
network environments.
They said such an approach (using mDNS or similar ones) could be discussed
maybe in SIPPING. 

The proposed charter is targeting to address intra-domain lookup that works
even if there is no multicast support.

Using DNS for intra-domain lookup is not different from using a central
server, which is not interesting to many of us, I believe. So transition
from mDNS to DNS is not appealing, either. 

What do you say?
Thanks.

Eunsoo

On 12/5/06, David Barrett <dbarrett at quinthar.com> wrote: 

 

  _____  

From: Eunsoo Shim 
Subject: [p2p-sip] DNS for inter-domain? Yes; intra-domain? No

 

So my understanding is that what P2PSIP group is trying to do is designing
mechanisms for P2P-style intra-domain location lookup where location records
are distributed among peers rather than in a central location server.

.

Am I missing any important aspect?

I agree, this is what P2P-SIP is trying to do.  I'm just not certain why,
because mDNS can do the same thing in a way that is already standardized,
already deployed, and already works.  Basically, for every real-world case,
it seems mDNS is adequate for ad-hoc unauthenticated namespaces, and DNS is
adequate for global, authenticated namespaces.  There is already a seamless
interface to graduate from mDNS to DNS depending upon what is available; it
already covers the whole range of use cases (minus the extreme security
cases, which I believe remain unsolved problems by any means).

 

I don't dispute you could make a DHT bootstrapped with UPnP or whatever.
You can build a lot of things.  I just don't see why we should bother when
there are easier, adequate technologies available.

 

-david

 

_______________________________________________

p2p-sip mailing list

p2p-sip at cs.columbia.edu

https://lists.cs.columbia.edu/cucslists/listinfo/p2p-sip

 

-------------- next part --------------
An HTML attachment was scrubbed...
URL: https://lists.cs.columbia.edu/pipermail/p2p-sip/attachments/20061206/aa60b5ae/attachment-0001.html

From dbarrett at quinthar.com  Wed Dec  6 03:07:39 2006
From: dbarrett at quinthar.com (David Barrett)
Date: Wed, 6 Dec 2006 00:07:39 -0800
Subject: [p2p-sip] Where are we "getting wrapped around the DNS axle"
In-Reply-To: <03f101c718d2$30c8e720$640fa8c0@cis.neustar.com>
Message-ID: <067901c7190d$9a3c70b0$0202a8c0@Quinthar>

Thank you for the detailed use case.  This is extremely helpful.  Yes, I can
see how that would work entirely.

However, that is the exact scenario that classic SIP is designed to excel
in.  I totally accept that you could provide a functionally equivalent
solution to DNS using a DHT.  What I don't understand is why you would put
in the extra work to do so when a perfectly adequate solution already
exists.

Can you help me understand why classic SIP is inadequate in this scenario?

I can't imagine scalability being your concern.  Today, in the real world, I
have 600MHz, single-CPU, 512MB boxes so ancient you can buy anymore, and
they can handle over 16,000 simultaneous TCP connections on each.  For
$100/mo (800Hosting.com) you can rent a dedicated server far more powerful,
hosted in a very reliable datacenter.  (These aren't SIP servers, but their
performance characteristics are very comparable.)  Assuming 2-3x the
simultaneous users are offline at any point in time, it seems sensible to
expect that whatever business model in place can earn at least $100/mo from
the 100K+ users that a single server can easily support.

I can't imagine performance being the problem, because any decent central
server will do call lookups and authentication checks orders of magnitudes
faster than a DHT (in the most basic sense, O(1) versus O(log(N)) for any
operation). 

And I can't imagine security being the problem, because a central directory
is much easier to maintain and faster to correct than any decentralized
system.  It is both simpler and more secure.

My impression was that P2P-SIP was going to take SIP to places it can't go
already.  If we're merely retreading the same ground a different way, I
think there must be a very compelling problem with the existing solution to
demand it, and I don't see it.

Can you give a more challenging use case that better leverages the strengths
of P2P?  Perhaps something involving users who sometimes are on the internet
and sometimes on a private network, or perhaps users who have a domain that
points to a hosted blog provider and can't run a SIP server?  Because the
use case as given is precisely what DNS and classic SIP excels at most.
  
-david

> -----Original Message-----
> From: Brian Rosen [mailto:br at brianrosen.net]
> Sent: Tuesday, December 05, 2006 5:02 PM
> To: 'David Barrett'; 'Enrico Marocco'
> Cc: 'P2P-SIP'; 'David A. Bryan'
> Subject: RE: [p2p-sip] Where are we "getting wrapped around the DNS axle"
> 
> 
> Okay, let's walk through this:
> 
> Someone, doesn't matter who, starts up a domain called p2ps.com.  They put
> up a website, and offer to give out credentials for their p2psip domain.
> They have a membership, code of conduct, and policies you like, and a
> business model you can deal with, so you sign up.
> 
> p2ps gives you a credential.  That credential will entitle you to join the
> DHT.  If you don't have one, you can't join the DHT, and you can't find
> out
> who else is in the domain, or how to contact them.  All the messages
> within
> the DHT are encrypted, and the only way you can get at any of them is to
> present credentials.
> 
> When you place a call, to anyone in the p2ps domain, you present your
> credentials issued by the domain with your INVITE.  You don't accept a
> call
> from anyone without the domain credentials.  In fact, NO signaling will
> enter any element within the domain without the proper credentials, and
> you,
> as a peer, can validate the credentials of any DHT message or any SIP
> message with those credentials.
> 
> Once you get the credential, the only centralized service needed is
> whatever
> it is that allows the domain to revoke credentials, if any, and that can
> be
> cached by the peers.  The name to address mechanism is impervious to a lot
> of damage because nothing is authoritative save the credential; it's all
> distributed.  Any isolated set of peers can communicate with each other,
> and
> as soon as connections to another peer or set of peers is restored
> communication with them becomes possible.
> 
> Brian
> 
> 
> > -----Original Message-----
> > From: David Barrett [mailto:dbarrett at quinthar.com]
> > Sent: Tuesday, December 05, 2006 6:51 PM
> > To: 'Brian Rosen'; 'Enrico Marocco'
> > Cc: 'P2P-SIP'; 'David A. Bryan'
> > Subject: RE: [p2p-sip] Where are we "getting wrapped around the DNS
> axle"
> >
> > I'm sorry, I don't follow.   The whole notion that someone else controls
> > who
> > I can talk to (and can talk to me) is precisely *opposite* our goals.
> If
> > your UA doesn't want to take calls from anyone outside of *@group.com,
> > then
> > it has every ability to define, manage, and enforce that access control
> > policy at the edge.
> >
> > Furthermore, if you for some reason don't want to control at the edge,
> > it's
> > vastly *easier* to do it with normal directories than with certificates
> > and
> > DHTs.
> >
> > Perhaps you can give a real-world use case of what you're describing,
> and
> > explain how you'd intend to solve it with a DHT?  Then I can present how
> > I'd
> > solve it with DNS/mDNS, and we can contrast which is more
> straightforward
> > --
> > to implement, to administer, and to use.
> >
> > DNS and DHT are equivalent services.  Both take a name, and spit out
> > information (generally IPs, though it can be a variety of records).  One
> > is
> > a drop-in replacement for the other.  The only difference is one exists
> on
> > a
> > massive, standard, secure, universal, global scale today, and the other
> > doesn't.
> >
> > I think whatever DHT-based solution you present, if you do a
> > search/replace
> > of DHT to DNS, that'll be basically what I'm proposing.
> >
> > -david
> >
> > > -----Original Message-----
> > > From: Brian Rosen [mailto:br at brianrosen.net]
> > > Sent: Tuesday, December 05, 2006 2:41 PM
> > > To: 'David Barrett'; 'Enrico Marocco'
> > > Cc: 'P2P-SIP'; 'David A. Bryan'
> > > Subject: RE: [p2p-sip] Where are we "getting wrapped around the DNS
> > axle"
> > >
> > > I think the elephant in the room you keep ignoring is the desire to
> > > control
> > > who can call me.  The point of an overlay network for many people is
> to
> > > restrict who can get to me.  The security mechanisms inside a domain
> are
> > a
> > > whole lot different from the Internet as a whole.  By forcing use of
> > DNS,
> > > you are requiring that I set up some completely separate system to
> > control
> > > access to me.  Within a DHT, only people inside the domain (who have
> > > credentials for the domain) can find out about me, find out how to
> > contact
> > > me, and actually make contact.  Now, admittedly, some of that is
> > security
> > > by
> > > obscurity that I don't subscribe to, but most if it is genuine
> security
> > by
> > > restricting access to people granted credentials in the domain.
> > >
> > > Brian
> > >
> > > > -----Original Message-----
> > > > From: p2p-sip-bounces at cs.columbia.edu [mailto:p2p-sip-
> > > > bounces at cs.columbia.edu] On Behalf Of David Barrett
> > > > Sent: Tuesday, December 05, 2006 4:36 PM
> > > > To: 'Enrico Marocco'
> > > > Cc: 'P2P-SIP'; 'David A. Bryan'
> > > > Subject: Re: [p2p-sip] Where are we "getting wrapped around the DNS
> > > axle"
> > > >
> > > > > -----Original Message-----
> > > > > From: Enrico Marocco [mailto:enrico.marocco at telecomitalia.it]
> > > > > Subject: Re: [p2p-sip] Where are we "getting wrapped around the
> DNS
> > > > axle"
> > > > >
> > > > > > In other words, do you agree:
> > > > > >
> > > > > > 	1) P2P-SIP must guarantee that each name resolves to a
> > single
> > > user
> > > > >
> > > > > Generally speaking, no.  The peer protocol must guarantee that the
> > > name
> > > > > used by a user enrolled within an overlay could not be used by
> > another
> > > > > user in the _same_ overlay.
> > > >
> > > > I'll grant it can be useful to have a totally unauthenticated,
> > insecure
> > > > namespace where you can be Enrico for a while and then the moment
> you
> > > log
> > > > off I can log on as Enrico and be indistinguishable to everyone
> else.
> > > I'd
> > > > wager this most often works in an ad hoc or "trade show"
> environment,
> > > > where
> > > > mDNS is totally adequate.
> > > >
> > > > However, I'm assuming a more valuable use case is an authenticated
> > world
> > > > that has long-lived names verified in a secure manner.  And in this
> > > world,
> > > > DNS is totally adequate, for networks as small as a few fixed nodes
> in
> > > the
> > > > jungle, or as big as the entire internet.
> > > >
> > > > What I'm waiting to hear about is a critical set of users who:
> > > >
> > > > 	- Don't need authentication
> > > > 	- Can't use mDNS
> > > >
> > > > Or, a critical set of users who:
> > > >
> > > > 	- Do need authentication
> > > > 	- Can't use DNS
> > > >
> > > > I'll totally grant we can construct a network to fit either of
> these.
> > > For
> > > > the first, set up a network in the jungle and explicitly disable
> > > > multicast.
> > > > For the second, set up a private network with a thousand nodes, and
> > > refuse
> > > > to install a DNS server (basically, disable all other network
> activity
> > > > except Gnutella and eDonkey).
> > > >
> > > > But these don't seem like realistic use cases.  In the first, just
> > tell
> > > > these jungle warriors "enable multicast" and they'll be fine.  In
> the
> > > > second, they're going to need a DNS server anyway, so just tell them
> > to
> > > > use
> > > > that.
> > > >
> > > >
> > > > > > 	2) Multiple users want the same name
> > > > >
> > > > > Possibly, but if an enrollment mechanism is _deployed_ in the
> > overlay
> > > > > (please note that it may be standardized as for jabber, or not),
> > only
> > > > > the firstly enrolled user can use it.
> > > >
> > > > Until he logs off or temporarily loses network connectivity, in
> which
> > > > anyone
> > > > can steal his name.  (Though I guess we could have him generate a
> > > private
> > > > key and the first to register a public key with a name "wins" and if
> > you
> > > > lose that private key you're basically screwed because there's no
> > > > "password
> > > > change" feature.)
> > > >
> > > > -david
> > > >
> > > > _______________________________________________
> > > > p2p-sip mailing list
> > > > p2p-sip at cs.columbia.edu
> > > > https://lists.cs.columbia.edu/cucslists/listinfo/p2p-sip


From tme at multicasttech.com  Wed Dec  6 11:34:56 2006
From: tme at multicasttech.com (Marshall Eubanks)
Date: Wed, 6 Dec 2006 11:34:56 -0500
Subject: [p2p-sip] DNS for inter-domain? Yes; intra-domain? No
In-Reply-To: <03f401c718d2$ffbcf1c0$640fa8c0@cis.neustar.com>
References: <03f401c718d2$ffbcf1c0$640fa8c0@cis.neustar.com>
Message-ID: <A1D8A6A5-B64F-455D-90DA-9B2513C4FB31@multicasttech.com>

Hello;

On Dec 5, 2006, at 8:08 PM, Brian Rosen wrote:

> I?ve spent a good deal of time dealing with systems that depend on  
> multicast
> and I conclude that multicast is a very wonderful thing, every  
> network can,
> and should support it, and almost none actually do.
>
> Multicast DOES NOT work in 99.9% of networks beyond a subnet.  It?s  
> a lot

I actually monitor this, and it's more like 3 % (to a factor of 2 or  
3), see

http://www.multicasttech.com/status

The only place where support is decent is in the NREN/academic  
network world.

There is also the AMT automatic tunneling proposal, which comes close  
to P2P :
http://www.ietf.org/internet-drafts/draft-ietf-mboned-auto- 
multicast-07.txt

> more than a research project to fix that.  IPTV may cause some  
> networks to
> allow multicast in some very restricted circumstances.  It's a  
> step, but you
> won't be able to use mDNS within those networks as I understand  
> deployment
> plans.
>

Not generally; these are typically NBMA networks.

> I think in some really interesting ad-hoc networks we may be able  
> to deploy
> multicast.  I would not be willing to place bets on that.  They  
> certainly
> don't support it now.

There is a lot of talk about it...

>
> Brian

Regards
Marshall

>
> ________________________________________
> From: p2p-sip-bounces at cs.columbia.edu
> [mailto:p2p-sip-bounces at cs.columbia.edu] On Behalf Of David Barrett
> Sent: Tuesday, December 05, 2006 7:02 PM
> To: 'Eunsoo Shim'
> Cc: 'P2P-SIP'
> Subject: Re: [p2p-sip] DNS for inter-domain? Yes; intra-domain? No
>
> Does the actual, real-world availability of multicast factor into the
> decision as to whether or not to use it?
>
> I mean, we *could* assume that every odd IP address won?t work, and  
> then
> design around that.  But odd IP addresses *do* work.
>
> Likewise, multicast *is* (or could easily be made) available for  
> all of the
> real-world use cases that demand it.
>
> It just seems like there are much more interesting *actual*  
> problems to
> solve than those that don?t affect real users.  But if you?re  
> excited about
> solving the non-multicast ad-hoc P2P-SIP problem on the assumption  
> that
> someday, somebody will have reason to use it (even though we can?t  
> really
> brainstorm anybody who would use it today), then I guess that?s a cool
> research project.
>
> But it?s a *research* project, not an IETF standardization  
> exercise.  DNS
> and mDNS work everywhere the *actual* users are, and thus I think  
> the IETF
> should focus on them, first.
>
> -david
>
> ________________________________________
> From: Eunsoo Shim [mailto:eunsooshim at gmail.com]
> Sent: Tuesday, December 05, 2006 3:04 PM
> To: David Barrett
> Cc: P2P-SIP
> Subject: Re: [p2p-sip] DNS for inter-domain? Yes; intra-domain? No
>
> I understand that mDNS can be used in ad-hoc networks where  
> multicast works.
> Some of us, including me, have discussed about using mDNS and  
> noticed that
> there is little to do to use mDNS or any other multicast-based  
> discovery
> protocol.
> Those people decided not to focus on the approach because assuming  
> multicast
> availability was not general and they did not want to limit P2PSIP  
> into such
> network environments.
> They said such an approach (using mDNS or similar ones) could be  
> discussed
> maybe in SIPPING.
>
> The proposed charter is targeting to address intra-domain lookup  
> that works
> even if there is no multicast support.
>
> Using DNS for intra-domain lookup is not different from using a  
> central
> server, which is not interesting to many of us, I believe. So  
> transition
> from mDNS to DNS is not appealing, either.
>
> What do you say?
> Thanks.
>
> Eunsoo
> On 12/5/06, David Barrett <dbarrett at quinthar.com> wrote:
>
> ________________________________________
> From: Eunsoo Shim
> Subject: [p2p-sip] DNS for inter-domain? Yes; intra-domain? No
>
> So my understanding is that what P2PSIP group is trying to do is  
> designing
> mechanisms for P2P-style intra-domain location lookup where  
> location records
> are distributed among peers rather than in a central location server.
> ?
> Am I missing any important aspect?
> I agree, this is what P2P-SIP is trying to do.  I'm just not  
> certain why,
> because mDNS can do the same thing in a way that is already  
> standardized,
> already deployed, and already works.  Basically, for every real- 
> world case,
> it seems mDNS is adequate for ad-hoc unauthenticated namespaces,  
> and DNS is
> adequate for global, authenticated namespaces.  There is already a  
> seamless
> interface to graduate from mDNS to DNS depending upon what is  
> available; it
> already covers the whole range of use cases (minus the extreme  
> security
> cases, which I believe remain unsolved problems by any means).
>
> I don't dispute you could make a DHT bootstrapped with UPnP or  
> whatever.
> You can build a lot of things.  I just don't see why we should  
> bother when
> there are easier, adequate technologies available.
>
> -david
>
>
>
> _______________________________________________
> p2p-sip mailing list
> p2p-sip at cs.columbia.edu
> https://lists.cs.columbia.edu/cucslists/listinfo/p2p-sip



From hgs at cs.columbia.edu  Wed Dec  6 12:38:06 2006
From: hgs at cs.columbia.edu (Henning Schulzrinne)
Date: Wed, 6 Dec 2006 12:38:06 -0500
Subject: [p2p-sip] DNS for inter-domain? Yes; intra-domain? No
In-Reply-To: <066801c7190d$996faad0$0202a8c0@Quinthar>
References: <066801c7190d$996faad0$0202a8c0@Quinthar>
Message-ID: <B49031A1-7AF1-45CE-8226-2FF8346A752F@cs.columbia.edu>


> -          Is mDNS an IETF standard? (It?s on the standards track.)
>
> -
As far as I know, this is not true. (The draft says "standards  
track", but that's a self-declaration.)

https://datatracker.ietf.org/public/idindex.cgi? 
command=id_detail&id=9784

-------------- next part --------------
An HTML attachment was scrubbed...
URL: https://lists.cs.columbia.edu/pipermail/p2p-sip/attachments/20061206/76276f63/attachment.html

From tme at multicasttech.com  Wed Dec  6 14:04:42 2006
From: tme at multicasttech.com (Marshall Eubanks)
Date: Wed, 6 Dec 2006 14:04:42 -0500
Subject: [p2p-sip] DNS for inter-domain? Yes; intra-domain? No
In-Reply-To: <B49031A1-7AF1-45CE-8226-2FF8346A752F@cs.columbia.edu>
References: <066801c7190d$996faad0$0202a8c0@Quinthar>
	<B49031A1-7AF1-45CE-8226-2FF8346A752F@cs.columbia.edu>
Message-ID: <B584BD11-F158-473A-8957-A81CE70452F4@multicasttech.com>

There was a long discussion about this on the IETF list last year. I  
am attaching the
letter from Stuart Cheshire that started it; I think it explains the  
relationship between
various protocols and proposals pretty well, except that LLMNR is  
part of Vista

http://www.microsoft.com/technet/community/columns/cableguy/cg1106.mspx

Regards
Marshall

On Dec 6, 2006, at 12:38 PM, Henning Schulzrinne wrote:

>
>> -          Is mDNS an IETF standard? (It?s on the standards track.)
>>
>> -
> As far as I know, this is not true. (The draft says "standards  
> track", but that's a self-declaration.)
>
> https://datatracker.ietf.org/public/idindex.cgi? 
> command=id_detail&id=9784
>
> _______________________________________________
> p2p-sip mailing list
> p2p-sip at cs.columbia.edu
> https://lists.cs.columbia.edu/cucslists/listinfo/p2p-sip


	From: 	  cheshire at apple.com
	Subject: 	Re: Last Call: 'Linklocal Multicast Name Resolution  
(LLMNR)' to Proposed Standard
	Date: 	August 25, 2005 1:37:42 AM EDT
	To: 	  iesg at ietf.org, ietf at ietf.org


> The IESG has received a request from the DNS Extensions WG to consider
> the following document:
>
> - 'Linklocal Multicast Name Resolution (LLMNR) '
>   <draft-ietf-dnsext-mdns-42.txt> as a Proposed Standard
>
> The IESG plans to make a decision in the next few weeks, and solicits
> final comments on this action.  Please send any comments to the
> iesg at ietf.org or ietf at ietf.org mailing lists by 2005-08-24.
>

My question about this document is one I've asked before in private;
it's time now that it was asked publicly.

What is this document for? No one has implemented this LLMNR  
protocol. No
one has any plans to implement this protocol. No company plans to ship
products using this protocol. Even Microsoft has not even hinted at  
plans
to use LLMNR in Longhorn/Vista. All of Microsoft's hype in this area
centers on a naming protocol they call Peer Name Resolution Protocol
(PNRP). Noah Horton, Program Manager in charge of PNRP, says, "PNRP  
is in
Windows Vista Beta 1. It is there and it is beautiful."

<http://blogs.msdn.com/noahh/archive/2005/08/11/450535.aspx>

I see no similar gossip surrounding LLMNR in Vista. In fact, a Google
search for "LLMNR Windows Vista" finds just twelve hits. A typical burst
of line noise finds more Google hits than that. Eleven of those hits are
accidental -- copies of the IANA port number list. I found only one hit
that looked like a plausible relevant match, but when I followed the  
link
it was an article that begins:


> Say "howdy" to Bonjour
> Apple's zero-config scheme leaves Microsoft and UPnP in the dust
>

<http://www.infoworld.com/article/05/06/03/23TCtigerSB_1.html>

So, what's LLMNR for, if no one actually wants to implement it?

I have my guess, but it's best explained by a brief bit of history:

Back in 1997 I attended a presentation at Stanford about SLP. I was
stunned by all the really obvious flaws it had. I wrote up a critique of
SLP and suggested that Apple should just implement the tried-and-true
AppleTalk Name Binding Protocol (NBP) running over IP Multicast. Partly
as a result of that and other similar things, Apple later hired me.

At an IETF social I discussed this idea of NBP/IP with Bill Woodcock. He
said that the IETF would reject any attempt to mix an AppleTalk protocol
with IP, but... if I could work out how to do the same thing using DNS
packets, then that would be much more acceptable to the IETF community.
It took me a little while, but I worked out how to encode the necessary
semantics in terms of standard DNS records and queries.

I proposed this idea, and was told I needed a working group. So, the
Zeroconf working group was created.

I proposed the idea again, and was told that it wasn't the charter of  
the
Zeroconf working group; I'd have to take it to DNSEXT.

I proposed the idea to DNSEXT and was told (and I quote):

     Multicast DNS... stupid idea.
     Zero Configuration... stupid idea.
     You all stupid people. You go away now please.

(To be fair to the speaker, he was not a native English speaker. My
attempt to say the same in Japanese would have been much less eloquent.
And at least he was honest about his position on the subject.)

So, the DNSEXT working group had made itself clear. They thought
Multicast DNS and DNS Service Discovery were stupid ideas, and would  
play
no part in doing anything that would help or encourage or facilitate
their creation.

So, I did go away. I designed the protocol, documented it in Internet
Drafts, solicited feedback, implemented the protocol, and shipped it  
in a
wide range of products. It's in Mac OS X 10.2, 10.3, and 10.4. It's in
Bonjour for Windows. It's in Apple's AirPort base stations, and it's  
what
lets them effortlessly share printers and stream music to your  
stereo. It
runs on Linux, Solaris, FreeBSD, and pretty much every other Unix
platform. Some Linux distributions already ship with it included. In
addition to Apple's own Open Source implementation, there are several
other competing Open Source implementations under various different
licenses. There are implementations in Java and Python and C#. It's in
your TiVos, your Axis network cameras, and Roku SoundBridges. It's in
pretty much any recent network printer made by HP, Canon, Brother,  
Xerox,
Lexmark, Epson (and probably other vendors I'm forgetting right now). It
was in the printers being used in the IETF terminal room at IETF 63 in
Paris, and if you work in any reasonably large organization that's  
bought
a network printer in the last year or two then it's probably running on
your network right now. If you only have Windows machines, then you're
not out of luck -- you can discover those printers using Bonjour for
Windows:

<http://www.apple.com/bonjour/>.

It is, though I say it myself, a huge success.

What is DNSEXT's response to this? Well, it was clear that the original
plan, ignoring mDNS and hoping it would die on its own, had failed. What
they needed was an IETF antibody. A decoy. A honey pot. Something to go
out and compete against mDNS in the marketplace of ideas. Something that
would have the appearance of being the "official" successor to mDNS, a
siren to lure potential implementers onto the rocks.

Now, I can already hear a certain minority saying, "You've convinced me.
LLMNR should be published as an IETF Standard. Apple's heresy must be
stopped at any cost."

The question for the wider IETF community is whether that's what the  
IETF
wants to do. Is the role of the IETF to foster interoperability, or to
sabotage it? Are Internet Standard RFCs supposed to be useful things to
be implemented, or are they supposed to be traps to ensnare the
uninformed?

It reminds me of OSI. We had a perfectly good networking protocol --
TCP/IP -- so ISO decided they needed to make their own
similar-but-not-quite-the-same protocol to compete with it, just for the
sake of not wanting to adopt something invented elsewhere. ISO thought
that because they were an "official standards body", whereas TCP/IP was
just something that a bunch of guys had made that happened to work, that
ISO could, by legislative fiat, oust the successful incumbent protocol.
How many man-years of work were wasted by that abortive effort? Did OSI
emerge from that debacle looking good, or looking bad?

In fact, this comparison to OSI is overly generous. OSI was implemented
and actually worked. It was just unnecessary. The problem with LLMNR is
that it is deliberately limited to prevent it from being used for  
service
discovery, which, you may recall, was the whole motivation for beginning
this work in the first place. LLMNR is presented as being the "official"
sanctioned successor to mDNS, as if it were somehow equivalent in
functionality but better designed, while in fact it is so
self-contradictory and nonsensical that it actually does nothing useful
at all.

Time and time again LLMNR has come up for last call. Each time I read  
it,
and point out the fatal flaws and contradictions (that would have been
painfully obvious to anyone had they actually tried to implement it). A
few changes get made, and the document comes back around again. Piece by
piece I'm designing a protocol for my competitors, so they can use it to
compete with my own!

If the proponents of LLMNR are sincerely supporting it because they
believe it offers useful functionality (and I like to believe that's the
case), then lets see some actual implementations.

Whatever happened to "rough consensus and RUNNING CODE?"

Once we have some actual implementations to try out, then we can compare
them with mDNS and see what might be necessary (on both sides) to make
them interoperate usefully. I'm quite happy for LLMNR to be a compatible
subset of mDNS. What's silly is for LLMNR to be gratuitously  
incompatible
for the sake of being incompatible. Bernard Aboba has an FAQ Web site
where he writes:

<http://www.drizzle.com/~aboba/DNSEXT/llmnrfaq.html>


> Apple's mDNS protocol differs from LLMNR (and DNS) in more than
> half a dozen ways.
>

He then goes on to list a bunch of similarities like "Apple's mDNS does
not share the DNS cache", and finishes with:


> Apple mDNS and LLMNR use different ports, as well as different
> multicast addresses, and because of the many protocol
> differences, do not interoperate.
>

Isn't that circular logic? They use different ports and multicast
addresses because they're incompatible, but the main reason they're
incompatible is because they use different ports and multicast  
addresses?
Surely that particular incompatibility could be remedied easily, merely
by NOT choosing to use a different port and multicast address?

Stuart Cheshire <cheshire at apple.com>
  * Wizard Without Portfolio, Apple Computer, Inc.
  * www.stuartcheshire.org


_______________________________________________
Ietf mailing list
Ietf at ietf.org
https://www1.ietf.org/mailman/listinfo/ietf



From huang-ming.pan at comcast.net  Wed Dec  6 15:00:36 2006
From: huang-ming.pan at comcast.net (Peter Pan)
Date: Wed, 6 Dec 2006 12:00:36 -0800
Subject: [p2p-sip] Where are we "getting wrapped around the DNS axle"
References: <067901c7190d$9a3c70b0$0202a8c0@Quinthar>
Message-ID: <001a01c71971$32289df0$030aa8c0@comcast.net>

> Can you give a more challenging use case that better leverages the
strengths
> of P2P?  Perhaps something involving users who sometimes are on the
internet
> and sometimes on a private network, or perhaps users who have a domain
that
> points to a hosted blog provider and can't run a SIP server?  Because the
> use case as given is precisely what DNS and classic SIP excels at most.
>
> -david
>

For 100K+ simultaneous online users, may I give you this as a case? Will you
need more than $100/month to offer network bandwidth for centralized
voicemail service?

Another is rtp media relay when clients are behind NATs that forbid direct
media flow between the clients.

I'm not sure if the cost and server complexity can grow linearly or even
exponentially when user number hits millions.



From huang-ming.pan at comcast.net  Wed Dec  6 15:08:42 2006
From: huang-ming.pan at comcast.net (Peter Pan)
Date: Wed, 6 Dec 2006 12:08:42 -0800
Subject: [p2p-sip] DNS for inter-domain? Yes; intra-domain? No
References: <066801c7190d$996faad0$0202a8c0@Quinthar>
Message-ID: <003601c71972$605d93a0$030aa8c0@comcast.net>

Great summary. Please see my feedback inline.

Peter
  ----- Original Message ----- 
  From: David Barrett 
  To: 'Henning Schulzrinne' 
  Cc: 'P2P-SIP' 
  Sent: Wednesday, December 06, 2006 12:07 AM
  Subject: Re: [p2p-sip] DNS for inter-domain? Yes; intra-domain? No


  I think the value of this discussion is in identifying and clarifying misimpressions about how DNS and mDNS work (and do not work) so we can make a more informed choice on whether or not to use them.  Already we've clarified:

   

  -          How can DNS records be automatically updated? (Through dynamic DNS protocols)

  [pp]: what is the cost and feasibility to upgrade existing DNS infrastructure to support dynamic DNS globally? is it similiar to the deployment of ipv6 and ip mobility?

  -          How fast can DNS propagate?  (Extremely fast with short TTLs)

  -          What if I have multiple SIP proxies? (Put multiple IPs into the DNS record)

  -          Is DNS only for IP addresses? (No, can store other records)

  [pp]: can/will these DNS server also store data like voicemail or other p2p application data?

  -          How does DNS work in on ad-hoc networks? (Using mDNS via multicast)

  -          Where is multicast supported? (Only everywhere it's needed - on subnets)

  -          Is mDNS an IETF standard? (It's on the standards track.)

  -          Is mDNS actually in use? (It's the foundation of zeroconf, which is used extensively by Apple)

  -          Are there any use-cases that DHT supports that DNS/mDNS cannot? (Apparently not.)

   

  Lack of clarity on these points could call for the unnecessary dismissal of these technologies on false premises, which I think nobody wants.

   

  -david

   


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

  From: Henning Schulzrinne [mailto:hgs at cs.columbia.edu] 
  Sent: Tuesday, December 05, 2006 4:20 PM
  To: David Barrett
  Cc: 'Eunsoo Shim'; 'P2P-SIP'
  Subject: Re: [p2p-sip] DNS for inter-domain? Yes; intra-domain? No

   

  I have to admit that I'm confused by this whole discussion. I thought we had agreement that

   

  (1) specifying mDNS + SIP operation is useful (and hopefully simple), but easily handled within the framework of SIPPING;

   

  (2) specifying dynamic DNS + SIP might be useful, subject to working out issues such as proxy discovery, authentication and the like and that interested parties should work out, at least to the extent of identifying the pieces that might need work (as a random example: can I expect my SIP Digest secret to work as a DNS update secret?). If, as claimed, this is easy to do, SIPPING is again the natural home for such a BCP effort, with the DNS-related WGs as the home for any DNS extensions that might prove necessary.

   

  Maybe those individuals interested in solving (2) should form an informal design team and get beyond the hand-waving stage and identify the problems, if any, that would have to be addressed in either protocol or operational standards. I have unfortunately seen little such effort so far, although I suspect that such level of detail isn't terribly well suited to be hashed out in front of several hundred mailing list members.

   

   

  We also have had 2 BOFs and numerous ad-hoc meetings on this topic, with a rough consensus on the current charter. Are you trying to keep people from working on a specific topic you consider inferior?

   

  We seem to be rehashing the same arguments again and again. 

   

  Henning

   

   

   

  On Dec 5, 2006, at 7:02 PM, David Barrett wrote:

   

  Does the actual, real-world availability of multicast factor into the decision as to whether or not to use it?

   

  I mean, we *could* assume that every odd IP address won't work, and then design around that.  But odd IP addresses *do* work.

   

  Likewise, multicast *is* (or could easily be made) available for all of the real-world use cases that demand it.

   

  It just seems like there are much more interesting *actual* problems to solve than those that don't affect real users.  But if you're excited about solving the non-multicast ad-hoc P2P-SIP problem on the assumption that someday, somebody will have reason to use it (even though we can't really brainstorm anybody who would use it today), then I guess that's a cool research project.

   

  But it's a *research* project, not an IETF standardization exercise.  DNS and mDNS work everywhere the *actual* users are, and thus I think the IETF should focus on them, first.

   

  -david

   


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

  From: Eunsoo Shim [mailto:eunsooshim at gmail.com] 
  Sent: Tuesday, December 05, 2006 3:04 PM
  To: David Barrett
  Cc: P2P-SIP
  Subject: Re: [p2p-sip] DNS for inter-domain? Yes; intra-domain? No

   

  I understand that mDNS can be used in ad-hoc networks where multicast works.
  Some of us, including me, have discussed about using mDNS and noticed that there is little to do to use mDNS or any other multicast-based discovery protocol. 
  Those people decided not to focus on the approach because assuming multicast availability was not general and they did not want to limit P2PSIP into such network environments.
  They said such an approach (using mDNS or similar ones) could be discussed maybe in SIPPING. 

  The proposed charter is targeting to address intra-domain lookup that works even if there is no multicast support.

  Using DNS for intra-domain lookup is not different from using a central server, which is not interesting to many of us, I believe. So transition from mDNS to DNS is not appealing, either. 

  What do you say?
  Thanks.

  Eunsoo

  On 12/5/06, David Barrett <dbarrett at quinthar.com> wrote: 

   


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

  From: Eunsoo Shim 
  Subject: [p2p-sip] DNS for inter-domain? Yes; intra-domain? No

   

  So my understanding is that what P2PSIP group is trying to do is designing mechanisms for P2P-style intra-domain location lookup where location records are distributed among peers rather than in a central location server.

  .

  Am I missing any important aspect?

  I agree, this is what P2P-SIP is trying to do.  I'm just not certain why, because mDNS can do the same thing in a way that is already standardized, already deployed, and already works.  Basically, for every real-world case, it seems mDNS is adequate for ad-hoc unauthenticated namespaces, and DNS is adequate for global, authenticated namespaces.  There is already a seamless interface to graduate from mDNS to DNS depending upon what is available; it already covers the whole range of use cases (minus the extreme security cases, which I believe remain unsolved problems by any means).

   

  I don't dispute you could make a DHT bootstrapped with UPnP or whatever.  You can build a lot of things.  I just don't see why we should bother when there are easier, adequate technologies available.

   

  -david

   

  _______________________________________________

  p2p-sip mailing list

  p2p-sip at cs.columbia.edu

  https://lists.cs.columbia.edu/cucslists/listinfo/p2p-sip

   



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


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

-------------- next part --------------
An HTML attachment was scrubbed...
URL: https://lists.cs.columbia.edu/pipermail/p2p-sip/attachments/20061206/ee96c34f/attachment-0001.html

From dbarrett at quinthar.com  Wed Dec  6 20:44:12 2006
From: dbarrett at quinthar.com (David Barrett)
Date: Wed, 6 Dec 2006 17:44:12 -0800
Subject: [p2p-sip] DNS for inter-domain? Yes; intra-domain? No
In-Reply-To: <003601c71972$605d93a0$030aa8c0@comcast.net>
Message-ID: <081d01c719a1$3663fce0$0202a8c0@Quinthar>

I'm not sure the DNS infrastructure needs much updating at all.  Perhaps
there will be more demand for DNS registrars to offer automated (aka
"dynamic") record updating, but that's really just standardizing what they
already do.

 

As for voicemail, no, I wouldn't expect to put that into DNS; that's
something best handled by some real storage provider.  You *could* do it in
a DHT, but it'd just be slower and less reliable, and the engineering cost
of building that would pay for central storage many, many times over.

 

Again, in a vacuum without taking anything else into consideration, be it
engineering cost, performance, reliability, existing systems, standards
support and so on, yes, a DHT could be used.  But when everything else is
factored in, it just makes no sense at all.

 

-david

 

  _____  

From: Peter Pan [mailto:huang-ming.pan at comcast.net] 
Sent: Wednesday, December 06, 2006 12:09 PM
To: David Barrett; 'Henning Schulzrinne'
Cc: 'P2P-SIP'
Subject: Re: [p2p-sip] DNS for inter-domain? Yes; intra-domain? No

 

Great summary. Please see my feedback inline.

 

Peter

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

From: David Barrett <mailto:dbarrett at quinthar.com>  

To: 'Henning <mailto:hgs at cs.columbia.edu>  Schulzrinne' 

Cc: 'P2P-SIP' <mailto:p2p-sip at cs.columbia.edu>  

Sent: Wednesday, December 06, 2006 12:07 AM

Subject: Re: [p2p-sip] DNS for inter-domain? Yes; intra-domain? No

 

I think the value of this discussion is in identifying and clarifying
misimpressions about how DNS and mDNS work (and do not work) so we can make
a more informed choice on whether or not to use them.  Already we've
clarified:

 

How can DNS records be automatically updated? (Through dynamic DNS
protocols)

-[pp]: what is the cost and feasibility to upgrade existing DNS
infrastructure to support dynamic DNS globally? is it similiar to the
deployment of ipv6 and ip mobility?

- How fast can DNS propagate?  (Extremely fast with short TTLs)

- What if I have multiple SIP proxies? (Put multiple IPs into the DNS
record)

Is DNS only for IP addresses? (No, can store other records)

-[pp]: can/will these DNS server also store data like voicemail or other p2p
application data?

- How does DNS work in on ad-hoc networks? (Using mDNS via multicast)

- Where is multicast supported? (Only everywhere it's needed - on subnets)

- Is mDNS an IETF standard? (It's on the standards track.)

- Is mDNS actually in use? (It's the foundation of zeroconf, which is used
extensively by Apple)

- Are there any use-cases that DHT supports that DNS/mDNS cannot?
(Apparently not.)

 

Lack of clarity on these points could call for the unnecessary dismissal of
these technologies on false premises, which I think nobody wants.

 

-david

 


  _____  


From: Henning Schulzrinne [mailto:hgs at cs.columbia.edu] 
Sent: Tuesday, December 05, 2006 4:20 PM
To: David Barrett
Cc: 'Eunsoo Shim'; 'P2P-SIP'
Subject: Re: [p2p-sip] DNS for inter-domain? Yes; intra-domain? No

 

I have to admit that I'm confused by this whole discussion. I thought we had
agreement that

 

(1) specifying mDNS + SIP operation is useful (and hopefully simple), but
easily handled within the framework of SIPPING;

 

(2) specifying dynamic DNS + SIP might be useful, subject to working out
issues such as proxy discovery, authentication and the like and that
interested parties should work out, at least to the extent of identifying
the pieces that might need work (as a random example: can I expect my SIP
Digest secret to work as a DNS update secret?). If, as claimed, this is easy
to do, SIPPING is again the natural home for such a BCP effort, with the
DNS-related WGs as the home for any DNS extensions that might prove
necessary.

 

Maybe those individuals interested in solving (2) should form an informal
design team and get beyond the hand-waving stage and identify the problems,
if any, that would have to be addressed in either protocol or operational
standards. I have unfortunately seen little such effort so far, although I
suspect that such level of detail isn't terribly well suited to be hashed
out in front of several hundred mailing list members.

 

 

We also have had 2 BOFs and numerous ad-hoc meetings on this topic, with a
rough consensus on the current charter. Are you trying to keep people from
working on a specific topic you consider inferior?

 

We seem to be rehashing the same arguments again and again. 

 

Henning

 

 

 

On Dec 5, 2006, at 7:02 PM, David Barrett wrote:

 

Does the actual, real-world availability of multicast factor into the
decision as to whether or not to use it?

 

I mean, we *could* assume that every odd IP address won't work, and then
design around that.  But odd IP addresses *do* work.

 

Likewise, multicast *is* (or could easily be made) available for all of the
real-world use cases that demand it.

 

It just seems like there are much more interesting *actual* problems to
solve than those that don't affect real users.  But if you're excited about
solving the non-multicast ad-hoc P2P-SIP problem on the assumption that
someday, somebody will have reason to use it (even though we can't really
brainstorm anybody who would use it today), then I guess that's a cool
research project.

 

But it's a *research* project, not an IETF standardization exercise.  DNS
and mDNS work everywhere the *actual* users are, and thus I think the IETF
should focus on them, first.

 

-david

 


  _____  


From: Eunsoo Shim [mailto:eunsooshim at gmail.com] 
Sent: Tuesday, December 05, 2006 3:04 PM
To: David Barrett
Cc: P2P-SIP
Subject: Re: [p2p-sip] DNS for inter-domain? Yes; intra-domain? No

 

I understand that mDNS can be used in ad-hoc networks where multicast works.
Some of us, including me, have discussed about using mDNS and noticed that
there is little to do to use mDNS or any other multicast-based discovery
protocol. 
Those people decided not to focus on the approach because assuming multicast
availability was not general and they did not want to limit P2PSIP into such
network environments.
They said such an approach (using mDNS or similar ones) could be discussed
maybe in SIPPING. 

The proposed charter is targeting to address intra-domain lookup that works
even if there is no multicast support.

Using DNS for intra-domain lookup is not different from using a central
server, which is not interesting to many of us, I believe. So transition
from mDNS to DNS is not appealing, either. 

What do you say?
Thanks.

Eunsoo

On 12/5/06, David Barrett <dbarrett at quinthar.com> wrote: 

 


  _____  


From: Eunsoo Shim 
Subject: [p2p-sip] DNS for inter-domain? Yes; intra-domain? No

 

So my understanding is that what P2PSIP group is trying to do is designing
mechanisms for P2P-style intra-domain location lookup where location records
are distributed among peers rather than in a central location server.

.

Am I missing any important aspect?

I agree, this is what P2P-SIP is trying to do.  I'm just not certain why,
because mDNS can do the same thing in a way that is already standardized,
already deployed, and already works.  Basically, for every real-world case,
it seems mDNS is adequate for ad-hoc unauthenticated namespaces, and DNS is
adequate for global, authenticated namespaces.  There is already a seamless
interface to graduate from mDNS to DNS depending upon what is available; it
already covers the whole range of use cases (minus the extreme security
cases, which I believe remain unsolved problems by any means).

 

I don't dispute you could make a DHT bootstrapped with UPnP or whatever.
You can build a lot of things.  I just don't see why we should bother when
there are easier, adequate technologies available.

 

-david

 

_______________________________________________

p2p-sip mailing list

p2p-sip at cs.columbia.edu

https://lists.cs.columbia.edu/cucslists/listinfo/p2p-sip

 


  _____  


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

-------------- next part --------------
An HTML attachment was scrubbed...
URL: https://lists.cs.columbia.edu/pipermail/p2p-sip/attachments/20061206/e48f4374/attachment-0001.html

From eunsooshim at gmail.com  Thu Dec  7 07:12:12 2006
From: eunsooshim at gmail.com (Eunsoo Shim)
Date: Thu, 7 Dec 2006 07:12:12 -0500
Subject: [p2p-sip] DNS for inter-domain? Yes; intra-domain? No
In-Reply-To: <081d01c719a1$3663fce0$0202a8c0@Quinthar>
References: <003601c71972$605d93a0$030aa8c0@comcast.net>
	<081d01c719a1$3663fce0$0202a8c0@Quinthar>
Message-ID: <7b683f3f0612070412h51f5d30er4cc6726e07139ee4@mail.gmail.com>

mDNS works only in a subnet. If you limit P2PSIP applications only to that
environment, mDNS should work. You can go to SIPPING to talk about it. Many
of us understand it.
If you suggest mDNS even for other environments, please have the Internet
changed to support multicast almost everywhere first. Then we can talk about
using mDNS for P2PSIP in a general sense.

dynDNS is just another DNS system, i.e. relying on DNS SERVERs. Using dynDNS
is same as retrieving a redirect server information from DNS and getting the
location of a particular user from the redirect server. That means, even we
don't have to update the DNS infrastructure! Just use existing DNS and SIP
and you can get what you propose.

You continuously try to imply that P2PSIP does not use DNS at all. I pointed
out that DNS was assumed for inter-domain connections for P2PSIP (and
actually SIP). All left from your argument is that we got to use DNS servers
even for intra-domain lookups, which is equivalent to using a central
location server.

Now it sounds to me that you are saying "p2p does not work. Stick to central
servers!"
I have no interest in arguing whether C/S architecture or P2P architecture
is better. Neither are many people in this list, I believe. We understand
there are tradeoffs. Maybe you can enjoy that kind of discussion with people
from Skype or eDonkey.

I think we passed already time for that kind of discussion here.

Already the current SIP systems use central servers for location
registration and voicemail storage.
If you are interested in another way/protocol of using central servers for
SIP, this is not a right place to propose it.
If you have a problem with DHT, please propose a different 'P2P' mechanism
for 'P2PSIP'.


Thanks.

Eunsoo




On 12/6/06, David Barrett <dbarrett at quinthar.com> wrote:
>
>  I'm not sure the DNS infrastructure needs much updating at all.  Perhaps
> there will be more demand for DNS registrars to offer automated (aka
> "dynamic") record updating, but that's really just standardizing what they
> already do.
>
>
>
> As for voicemail, no, I wouldn't expect to put that into DNS; that's
> something best handled by some real storage provider.  You **could** do it
> in a DHT, but it'd just be slower and less reliable, and the engineering
> cost of building that would pay for central storage many, many times over.
>
>
>
> Again, in a vacuum without taking anything else into consideration, be it
> engineering cost, performance, reliability, existing systems, standards
> support and so on, yes, a DHT could be used.  But when everything else is
> factored in, it just makes no sense at all.
>
>
>
> -david
>
>
>   ------------------------------
>
> *From:* Peter Pan [mailto:huang-ming.pan at comcast.net]
> *Sent:* Wednesday, December 06, 2006 12:09 PM
> *To:* David Barrett; 'Henning Schulzrinne'
> *Cc:* 'P2P-SIP'
> *Subject:* Re: [p2p-sip] DNS for inter-domain? Yes; intra-domain? No
>
>
>
> Great summary. Please see my feedback inline.
>
>
>
> Peter
>
>  ----- Original Message -----
>
> *From:* David Barrett <dbarrett at quinthar.com>
>
> *To:* 'Henning Schulzrinne' <hgs at cs.columbia.edu>
>
> *Cc:* 'P2P-SIP' <p2p-sip at cs.columbia.edu>
>
> *Sent:* Wednesday, December 06, 2006 12:07 AM
>
> *Subject:* Re: [p2p-sip] DNS for inter-domain? Yes; intra-domain? No
>
>
>
> I think the value of this discussion is in identifying and clarifying
> misimpressions about how DNS and mDNS work (and do not work) so we can make
> a more informed choice on whether or not to use them.  Already we've
> clarified:
>
>
>
> How can DNS records be automatically updated? *(Through dynamic DNS
> protocols)*
>
> -*[pp]: what is the cost and feasibility to upgrade existing DNS
> infrastructure to support dynamic DNS globally? is it similiar to the
> deployment of ipv6 and ip mobility?*
>
> - How fast can DNS propagate?  *(Extremely fast with short TTLs)*
>
> - What if I have multiple SIP proxies? *(Put multiple IPs into the DNS
> record)*
>
> Is DNS only for IP addresses? *(No, can store other records)*
>
> -*[pp]: can/will these DNS server also store data like voicemail or other
> p2p application data?*
>
> - How does DNS work in on ad-hoc networks? *(Using mDNS via multicast)*
>
> - Where is multicast supported? *(Only everywhere it's needed ? on
> subnets)*
>
> - Is mDNS an IETF standard? *(It's on the standards track.)*
>
> - Is mDNS actually in use? *(It's the foundation of zeroconf, which is
> used extensively by Apple)*
>
> - Are there any use-cases that DHT supports that DNS/mDNS cannot? *(Apparently
> not.)*
>
>
>
> Lack of clarity on these points could call for the unnecessary dismissal
> of these technologies on false premises, which I think nobody wants.
>
>
>
> -david
>
>
>   ------------------------------
>
> *From:* Henning Schulzrinne [mailto:hgs at cs.columbia.edu]
> *Sent:* Tuesday, December 05, 2006 4:20 PM
> *To:* David Barrett
> *Cc:* 'Eunsoo Shim'; 'P2P-SIP'
> *Subject:* Re: [p2p-sip] DNS for inter-domain? Yes; intra-domain? No
>
>
>
> I have to admit that I'm confused by this whole discussion. I thought we
> had agreement that
>
>
>
> (1) specifying mDNS + SIP operation is useful (and hopefully simple), but
> easily handled within the framework of SIPPING;
>
>
>
> (2) specifying dynamic DNS + SIP might be useful, subject to working out
> issues such as proxy discovery, authentication and the like and that
> interested parties should work out, at least to the extent of identifying
> the pieces that might need work (as a random example: can I expect my SIP
> Digest secret to work as a DNS update secret?). If, as claimed, this is easy
> to do, SIPPING is again the natural home for such a BCP effort, with the
> DNS-related WGs as the home for any DNS extensions that might prove
> necessary.
>
>
>
> Maybe those individuals interested in solving (2) should form an informal
> design team and get beyond the hand-waving stage and identify the problems,
> if any, that would have to be addressed in either protocol or operational
> standards. I have unfortunately seen little such effort so far, although I
> suspect that such level of detail isn't terribly well suited to be hashed
> out in front of several hundred mailing list members.
>
>
>
>
>
> We also have had 2 BOFs and numerous ad-hoc meetings on this topic, with a
> rough consensus on the current charter. Are you trying to keep people from
> working on a specific topic you consider inferior?
>
>
>
> We seem to be rehashing the same arguments again and again.
>
>
>
> Henning
>
>
>
>
>
>
>
> On Dec 5, 2006, at 7:02 PM, David Barrett wrote:
>
>
>
> Does the actual, real-world availability of multicast factor into the
> decision as to whether or not to use it?
>
>
>
> I mean, we **could** assume that every odd IP address won't work, and then
> design around that.  But odd IP addresses **do** work.
>
>
>
> Likewise, multicast **is** (or could easily be made) available for all of
> the real-world use cases that demand it.
>
>
>
> It just seems like there are much more interesting **actual** problems to
> solve than those that don't affect real users.  But if you're excited about
> solving the non-multicast ad-hoc P2P-SIP problem on the assumption that
> someday, somebody will have reason to use it (even though we can't really
> brainstorm anybody who would use it today), then I guess that's a cool
> research project.
>
>
>
> But it's a **research** project, not an IETF standardization exercise.
> DNS and mDNS work everywhere the **actual** users are, and thus I think
> the IETF should focus on them, first.
>
>
>
> -david
>
>
>   ------------------------------
>
> *From:* Eunsoo Shim [mailto:eunsooshim at gmail.com <eunsooshim at gmail.com>]
> *Sent:* Tuesday, December 05, 2006 3:04 PM
> *To:* David Barrett
> *Cc:* P2P-SIP
> *Subject:* Re: [p2p-sip] DNS for inter-domain? Yes; intra-domain? No
>
>
>
> I understand that mDNS can be used in ad-hoc networks where multicast
> works.
> Some of us, including me, have discussed about using mDNS and noticed that
> there is little to do to use mDNS or any other multicast-based discovery
> protocol.
> Those people decided not to focus on the approach because assuming
> multicast availability was not general and they did not want to limit P2PSIP
> into such network environments.
> They said such an approach (using mDNS or similar ones) could be discussed
> maybe in SIPPING.
>
> The proposed charter is targeting to address intra-domain lookup that
> works even if there is no multicast support.
>
> Using DNS for intra-domain lookup is not different from using a central
> server, which is not interesting to many of us, I believe. So transition
> from mDNS to DNS is not appealing, either.
>
> What do you say?
> Thanks.
>
> Eunsoo
>
> On 12/5/06, *David Barrett* <dbarrett at quinthar.com> wrote:
>
>
>   ------------------------------
>
> *From:* Eunsoo Shim
> *Subject:* [p2p-sip] DNS for inter-domain? Yes; intra-domain? No
>
>
>
> So my understanding is that what P2PSIP group is trying to do is designing
> mechanisms for P2P-style intra-domain location lookup where location records
> are distributed among peers rather than in a central location server.
>
> ?
>
> Am I missing any important aspect?
>
> I agree, this is what P2P-SIP is trying to do.  I'm just not certain why,
> because mDNS can do the same thing in a way that is already standardized,
> already deployed, and already works.  Basically, for every real-world case,
> it seems mDNS is adequate for ad-hoc unauthenticated namespaces, and DNS is
> adequate for global, authenticated namespaces.  There is already a seamless
> interface to graduate from mDNS to DNS depending upon what is available; it
> already covers the whole range of use cases (minus the extreme security
> cases, which I believe remain unsolved problems by any means).
>
>
>
> I don't dispute you could make a DHT bootstrapped with UPnP or whatever.
> You can build a lot of things.  I just don't see why we should bother when
> there are easier, adequate technologies available.
>
>
>
> -david
>
>
>
> _______________________________________________
>
> p2p-sip mailing list
>
> p2p-sip at cs.columbia.edu
>
> https://lists.cs.columbia.edu/cucslists/listinfo/p2p-sip
>
>
>  ------------------------------
>
> _______________________________________________
> p2p-sip mailing list
> p2p-sip at cs.columbia.edu
> https://lists.cs.columbia.edu/cucslists/listinfo/p2p-sip
>
>
> _______________________________________________
> p2p-sip mailing list
> p2p-sip at cs.columbia.edu
> https://lists.cs.columbia.edu/cucslists/listinfo/p2p-sip
>
>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: https://lists.cs.columbia.edu/pipermail/p2p-sip/attachments/20061207/9d9ba147/attachment-0001.html

From philip_matthews at magma.ca  Thu Dec  7 12:48:06 2006
From: philip_matthews at magma.ca (Philip Matthews)
Date: Thu, 7 Dec 2006 12:48:06 -0500
Subject: [p2p-sip] Concepts draft: Open Issue #1: Searching for services
Message-ID: <B784DF3B-10C8-401B-BCA4-149F7AF0ECD0@magma.ca>

Open issue #1:  Can different peers in a P2PSIP Overlay offer
different services to the Overlay? If so, do we need the
concept of  searching for the set of peers that offer a service?


Proposed Resolution: Yes and Yes. Though we may require
that all peers offer some minimal set of services, there is likely
to be at least one service  that is not offered by every peer.
Thus we need the concept of searching for the set of peers
that offer a service.

For now, we are proposing to leave undefined the detailed
properties of the search. Thus we are proposing not
to answer questions like whether the search finds all peers
that offer the service, or whether the search returns the
"closest" peer. Instead, we propose to add these questions to the
design questions section of the document.

Comments?

- Philip



From philip_matthews at magma.ca  Thu Dec  7 12:37:29 2006
From: philip_matthews at magma.ca (Philip Matthews)
Date: Thu, 7 Dec 2006 12:37:29 -0500
Subject: [p2p-sip] Open issues in Concepts draft
Message-ID: <B40CB53A-F5C4-4F60-8F75-CBE7E13D950F@magma.ca>

Folks:

The authors of the Concepts draft (draft-willis-p2psip-concepts)
have started talking about the next revision of the document.

One of the main goals of the next revision is to try to address some of
the open issues in the document. The authors have been discussing
these issues, and have come up with some proposed resolutions.

I am going to start posting these issues and the authors' proposed  
resolution.
I'll do these as separate threads so that the we can discuss each one  
individually.

- Philip


From philip_matthews at magma.ca  Thu Dec  7 12:48:21 2006
From: philip_matthews at magma.ca (Philip Matthews)
Date: Thu, 7 Dec 2006 12:48:21 -0500
Subject: [p2p-sip] Concepts draft: Open Issue #2: Adding NATs
Message-ID: <24892751-5955-4316-A6B6-0C1CE5F29DD5@magma.ca>

Open Issue #2:  Right now, the Concepts document says
there can be NATs between peers. However, the examples
in section 3.4 ignore the fact that it is generally not possible
for an arbitrary node to send a message (e.g. a SIP INVITE)
to destination node that is behind a NAT.

Proposed Resolution:  The P2PSIP Overlay provides a message
relaying service. In other words, the sending peer/client can send
a SIP message through the Overlay in some manner that causes
it to be delivered to the destination peer or client.

Consider the example in section 3.4.2 where client C sends an
INVITE to the user U residing on peer F. Since client C connects
to the Overlay via peer Q, client C might pass this INVITE to peer Q
which then passes somehow through the overlay so that it gets
delivered to peer F.  (If client C somehow knows how to reach
another peer R, it might choose to send the message to that peer
for relaying instead).

There are a number of ways that such a relaying service might
be implemented. Some have suggested that the lookup might return
a route vector to the destination. Others have suggested that the
routing might be done as part of the search. However, in the
next revision, we are proposing to not give details, but just say
that such a mechanism exists.

Comments?

- Philip

From philip_matthews at magma.ca  Thu Dec  7 12:48:35 2006
From: philip_matthews at magma.ca (Philip Matthews)
Date: Thu, 7 Dec 2006 12:48:35 -0500
Subject: [p2p-sip] Concepts draft: Open Issue #3: Enrollment as a
	Peer/Client Protocol operation?
Message-ID: <3E1F6E23-B2EF-43F2-B9EE-3C175384922B@magma.ca>

Open issue #3: In version -03 of the Concepts  draft, I added some  
text that stated
that Enrollment of peers and resources/users was an operation  
supported by the Peer and Client Protocols.
However, a couple of weeks ago, Brian Rosen and Henry Sinnreich  
suggested that
we not standardize the protocol for obtaining credentials, but simply  
specify the acceptable
format of a credential.

Proposed Resolution: The authors propose to adopt Brian and Henry's  
view of the world.
Thus we propose to remove enrollment of peers and resources/users as  
operations supported
by the Peer and Client protocols, and state that the method of  
obtaining credentials is
not being standardized for now.

Comments?

- Philip




From dbarrett at quinthar.com  Thu Dec  7 16:37:01 2006
From: dbarrett at quinthar.com (David Barrett)
Date: Thu, 7 Dec 2006 13:37:01 -0800
Subject: [p2p-sip] DNS for inter-domain? Yes; intra-domain? No
In-Reply-To: <7b683f3f0612070412h51f5d30er4cc6726e07139ee4@mail.gmail.com>
Message-ID: <09a701c71a47$d529b8f0$0202a8c0@Quinthar>

Hm, I will agree that the conversation has covered all the useful ground
it's going to cover.  As for the confusions you raise:

 

-          DNS is used when available (internet or private network), and
mDNS otherwise (ad hoc subnet)



-          Multicast only works on subnets.  Ad hoc networks are typically
single-subnets.  Thus mDNS works on ad hoc networks.



-          Multicast does not work between subnets.  The internet and
private networks use multiple subnets.  Thus mDNS will not work across the
internet or on private networks spanning multiple subnets.  However, DNS is
typically available in this case, and thus DNS is used instead. 



-          DNS and mDNS cover the total range of real-world networking
environments.



-          I understand P2P-SIP is currently proposed to use DNS as an
optional extension to the mandatory DHT.  I think it'd be more sensible to
have mandatory support for DNS with optional DHT, particularly because the
DHT is hard work for no real-world gain and it seems wasteful to force
everyone to pay it.



-          The "P2P" system I propose as a replacement for DHT is merely
what Adam originally proposed: configure P2P-SIP UAs to update a DNS record
that points to another P2P-SIP UA as a proxy.  



-          Whether or not that is P2P "enough" is a long boring religious
discussion.  But so far as we've been able to determine, this solves all the
same use cases as the DHT, but with less work and better integration with
existing SIP.



-          P2P works; it's what pays my bills.  But it's a question of when
and where is appropriate.



Please let me know if you have any more questions about this idea or need me
to clarify it more, but I think we've covered it exhaustively.

 

-david

 

  _____  

From: Eunsoo Shim [mailto:eunsooshim at gmail.com] 
Sent: Thursday, December 07, 2006 4:12 AM
To: David Barrett
Cc: Peter Pan; Henning Schulzrinne; P2P-SIP
Subject: Re: [p2p-sip] DNS for inter-domain? Yes; intra-domain? No

 

mDNS works only in a subnet. If you limit P2PSIP applications only to that
environment, mDNS should work. You can go to SIPPING to talk about it. Many
of us understand it.
If you suggest mDNS even for other environments, please have the Internet
changed to support multicast almost everywhere first. Then we can talk about
using mDNS for P2PSIP in a general sense. 

dynDNS is just another DNS system, i.e. relying on DNS SERVERs. Using dynDNS
is same as retrieving a redirect server information from DNS and getting the
location of a particular user from the redirect server. That means, even we
don't have to update the DNS infrastructure! Just use existing DNS and SIP
and you can get what you propose.

You continuously try to imply that P2PSIP does not use DNS at all. I pointed
out that DNS was assumed for inter-domain connections for P2PSIP (and
actually SIP). All left from your argument is that we got to use DNS servers
even for intra-domain lookups, which is equivalent to using a central
location server. 

Now it sounds to me that you are saying "p2p does not work. Stick to central
servers!"
I have no interest in arguing whether C/S architecture or P2P architecture
is better. Neither are many people in this list, I believe. We understand
there are tradeoffs. Maybe you can enjoy that kind of discussion with people
from Skype or eDonkey. 

I think we passed already time for that kind of discussion here.

Already the current SIP systems use central servers for location
registration and voicemail storage. 
If you are interested in another way/protocol of using central servers for
SIP, this is not a right place to propose it. 
If you have a problem with DHT, please propose a different 'P2P' mechanism
for 'P2PSIP'.


Thanks.

Eunsoo





On 12/6/06, David Barrett <dbarrett at quinthar.com> wrote:

I'm not sure the DNS infrastructure needs much updating at all.  Perhaps
there will be more demand for DNS registrars to offer automated (aka
"dynamic") record updating, but that's really just standardizing what they
already do.

 

As for voicemail, no, I wouldn't expect to put that into DNS; that's
something best handled by some real storage provider.  You *could* do it in
a DHT, but it'd just be slower and less reliable, and the engineering cost
of building that would pay for central storage many, many times over.

 

Again, in a vacuum without taking anything else into consideration, be it
engineering cost, performance, reliability, existing systems, standards
support and so on, yes, a DHT could be used.  But when everything else is
factored in, it just makes no sense at all.

 

-david

 

  _____  

From: Peter Pan [mailto:huang-ming.pan at comcast.net] 
Sent: Wednesday, December 06, 2006 12:09 PM
To: David Barrett; 'Henning Schulzrinne'
Cc: 'P2P-SIP'


Subject: Re: [p2p-sip] DNS for inter-domain? Yes; intra-domain? No

 

Great summary. Please see my feedback inline.

 

Peter

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

From: David Barrett <mailto:dbarrett at quinthar.com>  

To: 'Henning Schulzrinne' <mailto:hgs at cs.columbia.edu>  

Cc: 'P2P-SIP' <mailto:p2p-sip at cs.columbia.edu>  

Sent: Wednesday, December 06, 2006 12:07 AM

Subject: Re: [p2p-sip] DNS for inter-domain? Yes; intra-domain? No

 

I think the value of this discussion is in identifying and clarifying
misimpressions about how DNS and mDNS work (and do not work) so we can make
a more informed choice on whether or not to use them.  Already we've
clarified:

 

How can DNS records be automatically updated? (Through dynamic DNS
protocols)

- [pp]: what is the cost and feasibility to upgrade existing DNS
infrastructure to support dynamic DNS globally? is it similiar to the
deployment of ipv6 and ip mobility?

- How fast can DNS propagate?  (Extremely fast with short TTLs)

- What if I have multiple SIP proxies? (Put multiple IPs into the DNS
record) 

Is DNS only for IP addresses? (No, can store other records) 

- [pp]: can/will these DNS server also store data like voicemail or other
p2p application data?

- How does DNS work in on ad-hoc networks? (Using mDNS via multicast) 

- Where is multicast supported? (Only everywhere it's needed - on subnets)

- Is mDNS an IETF standard? (It's on the standards track.)

- Is mDNS actually in use? (It's the foundation of zeroconf, which is used
extensively by Apple)

- Are there any use-cases that DHT supports that DNS/mDNS cannot?
(Apparently not.)

 

Lack of clarity on these points could call for the unnecessary dismissal of
these technologies on false premises, which I think nobody wants.

 

-david

 


  _____  


From: Henning Schulzrinne [mailto:hgs at cs.columbia.edu] 
Sent: Tuesday, December 05, 2006 4:20 PM
To: David Barrett
Cc: 'Eunsoo Shim'; 'P2P-SIP'
Subject: Re: [p2p-sip] DNS for inter-domain? Yes; intra-domain? No

 

I have to admit that I'm confused by this whole discussion. I thought we had
agreement that

 

(1) specifying mDNS + SIP operation is useful (and hopefully simple), but
easily handled within the framework of SIPPING;

 

(2) specifying dynamic DNS + SIP might be useful, subject to working out
issues such as proxy discovery, authentication and the like and that
interested parties should work out, at least to the extent of identifying
the pieces that might need work (as a random example: can I expect my SIP
Digest secret to work as a DNS update secret?). If, as claimed, this is easy
to do, SIPPING is again the natural home for such a BCP effort, with the
DNS-related WGs as the home for any DNS extensions that might prove
necessary.

 

Maybe those individuals interested in solving (2) should form an informal
design team and get beyond the hand-waving stage and identify the problems,
if any, that would have to be addressed in either protocol or operational
standards. I have unfortunately seen little such effort so far, although I
suspect that such level of detail isn't terribly well suited to be hashed
out in front of several hundred mailing list members.

 

 

We also have had 2 BOFs and numerous ad-hoc meetings on this topic, with a
rough consensus on the current charter. Are you trying to keep people from
working on a specific topic you consider inferior?

 

We seem to be rehashing the same arguments again and again. 

 

Henning

 

 

 

On Dec 5, 2006, at 7:02 PM, David Barrett wrote:

 

Does the actual, real-world availability of multicast factor into the
decision as to whether or not to use it?

 

I mean, we *could* assume that every odd IP address won't work, and then
design around that.  But odd IP addresses *do* work.

 

Likewise, multicast *is* (or could easily be made) available for all of the
real-world use cases that demand it.

 

It just seems like there are much more interesting *actual* problems to
solve than those that don't affect real users.  But if you're excited about
solving the non-multicast ad-hoc P2P-SIP problem on the assumption that
someday, somebody will have reason to use it (even though we can't really
brainstorm anybody who would use it today), then I guess that's a cool
research project.

 

But it's a *research* project, not an IETF standardization exercise.  DNS
and mDNS work everywhere the *actual* users are, and thus I think the IETF
should focus on them, first.

 

-david

 


  _____  


From: Eunsoo Shim [ mailto:eunsooshim at gmail.com] 
Sent: Tuesday, December 05, 2006 3:04 PM
To: David Barrett
Cc: P2P-SIP
Subject: Re: [p2p-sip] DNS for inter-domain? Yes; intra-domain? No

 

I understand that mDNS can be used in ad-hoc networks where multicast works.
Some of us, including me, have discussed about using mDNS and noticed that
there is little to do to use mDNS or any other multicast-based discovery
protocol. 
Those people decided not to focus on the approach because assuming multicast
availability was not general and they did not want to limit P2PSIP into such
network environments.
They said such an approach (using mDNS or similar ones) could be discussed
maybe in SIPPING. 

The proposed charter is targeting to address intra-domain lookup that works
even if there is no multicast support.

Using DNS for intra-domain lookup is not different from using a central
server, which is not interesting to many of us, I believe. So transition
from mDNS to DNS is not appealing, either. 

What do you say?
Thanks.

Eunsoo

On 12/5/06, David Barrett <dbarrett at quinthar.com> wrote: 

 


  _____  


From: Eunsoo Shim 
Subject: [p2p-sip] DNS for inter-domain? Yes; intra-domain? No

 

So my understanding is that what P2PSIP group is trying to do is designing
mechanisms for P2P-style intra-domain location lookup where location records
are distributed among peers rather than in a central location server.

.

Am I missing any important aspect?

I agree, this is what P2P-SIP is trying to do.  I'm just not certain why,
because mDNS can do the same thing in a way that is already standardized,
already deployed, and already works.  Basically, for every real-world case,
it seems mDNS is adequate for ad-hoc unauthenticated namespaces, and DNS is
adequate for global, authenticated namespaces.  There is already a seamless
interface to graduate from mDNS to DNS depending upon what is available; it
already covers the whole range of use cases (minus the extreme security
cases, which I believe remain unsolved problems by any means).

 

I don't dispute you could make a DHT bootstrapped with UPnP or whatever.
You can build a lot of things.  I just don't see why we should bother when
there are easier, adequate technologies available.

 

-david

 

_______________________________________________

p2p-sip mailing list

p2p-sip at cs.columbia.edu 

https://lists.cs.columbia.edu/cucslists/listinfo/p2p-sip 

 


  _____  


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


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



 

-------------- next part --------------
An HTML attachment was scrubbed...
URL: https://lists.cs.columbia.edu/pipermail/p2p-sip/attachments/20061207/2d28e1bf/attachment-0001.html

From RADHIKA.R.ROY at saic.com  Thu Dec  7 17:44:56 2006
From: RADHIKA.R.ROY at saic.com (Roy, Radhika R.)
Date: Thu, 7 Dec 2006 17:44:56 -0500
Subject: [p2p-sip] DNS for inter-domain? Yes; intra-domain? No
References: <09a701c71a47$d529b8f0$0202a8c0@Quinthar>
Message-ID: <62D92A9A02BCC845B202323D49A48426E1CF54@0591-ITS-EXMP02.us.saic.com>

David:
 
Do DNS and mDNS also cover P2P Mobile Ad Hoc Networks (MANET)?
 
We are looking for MANET support right now for P2P-SIP networking.
 
Best regards,
Radhika
 
PS: We are using DNS right now, but no insufficient.

________________________________

From: p2p-sip-bounces at cs.columbia.edu on behalf of David Barrett
Sent: Thu 12/7/2006 4:37 PM
To: 'Eunsoo Shim'
Cc: 'P2P-SIP'
Subject: Re: [p2p-sip] DNS for inter-domain? Yes; intra-domain? No



Hm, I will agree that the conversation has covered all the useful ground it's going to cover.  As for the confusions you raise:

 

-          DNS is used when available (internet or private network), and mDNS otherwise (ad hoc subnet)



-          Multicast only works on subnets.  Ad hoc networks are typically single-subnets.  Thus mDNS works on ad hoc networks.



-          Multicast does not work between subnets.  The internet and private networks use multiple subnets.  Thus mDNS will not work across the internet or on private networks spanning multiple subnets.  However, DNS is typically available in this case, and thus DNS is used instead. 



-          DNS and mDNS cover the total range of real-world networking environments.



-          I understand P2P-SIP is currently proposed to use DNS as an optional extension to the mandatory DHT.  I think it'd be more sensible to have mandatory support for DNS with optional DHT, particularly because the DHT is hard work for no real-world gain and it seems wasteful to force everyone to pay it.



-          The "P2P" system I propose as a replacement for DHT is merely what Adam originally proposed: configure P2P-SIP UAs to update a DNS record that points to another P2P-SIP UA as a proxy.  



-          Whether or not that is P2P "enough" is a long boring religious discussion.  But so far as we've been able to determine, this solves all the same use cases as the DHT, but with less work and better integration with existing SIP.



-          P2P works; it's what pays my bills.  But it's a question of when and where is appropriate.



Please let me know if you have any more questions about this idea or need me to clarify it more, but I think we've covered it exhaustively.

 

-david

 

________________________________

From: Eunsoo Shim [mailto:eunsooshim at gmail.com] 
Sent: Thursday, December 07, 2006 4:12 AM
To: David Barrett
Cc: Peter Pan; Henning Schulzrinne; P2P-SIP
Subject: Re: [p2p-sip] DNS for inter-domain? Yes; intra-domain? No

 

mDNS works only in a subnet. If you limit P2PSIP applications only to that environment, mDNS should work. You can go to SIPPING to talk about it. Many of us understand it.
If you suggest mDNS even for other environments, please have the Internet changed to support multicast almost everywhere first. Then we can talk about using mDNS for P2PSIP in a general sense. 

dynDNS is just another DNS system, i.e. relying on DNS SERVERs. Using dynDNS is same as retrieving a redirect server information from DNS and getting the location of a particular user from the redirect server. That means, even we don't have to update the DNS infrastructure! Just use existing DNS and SIP and you can get what you propose.

You continuously try to imply that P2PSIP does not use DNS at all. I pointed out that DNS was assumed for inter-domain connections for P2PSIP (and actually SIP). All left from your argument is that we got to use DNS servers even for intra-domain lookups, which is equivalent to using a central location server. 

Now it sounds to me that you are saying "p2p does not work. Stick to central servers!"
I have no interest in arguing whether C/S architecture or P2P architecture is better. Neither are many people in this list, I believe. We understand there are tradeoffs. Maybe you can enjoy that kind of discussion with people from Skype or eDonkey. 

I think we passed already time for that kind of discussion here.

Already the current SIP systems use central servers for location registration and voicemail storage. 
If you are interested in another way/protocol of using central servers for SIP, this is not a right place to propose it. 
If you have a problem with DHT, please propose a different 'P2P' mechanism for 'P2PSIP'.


Thanks.

Eunsoo





On 12/6/06, David Barrett <dbarrett at quinthar.com> wrote:

I'm not sure the DNS infrastructure needs much updating at all.  Perhaps there will be more demand for DNS registrars to offer automated (aka "dynamic") record updating, but that's really just standardizing what they already do.

 

As for voicemail, no, I wouldn't expect to put that into DNS; that's something best handled by some real storage provider.  You *could* do it in a DHT, but it'd just be slower and less reliable, and the engineering cost of building that would pay for central storage many, many times over.

 

Again, in a vacuum without taking anything else into consideration, be it engineering cost, performance, reliability, existing systems, standards support and so on, yes, a DHT could be used.  But when everything else is factored in, it just makes no sense at all.

 

-david

 

________________________________

From: Peter Pan [mailto:huang-ming.pan at comcast.net] 
Sent: Wednesday, December 06, 2006 12:09 PM
To: David Barrett; 'Henning Schulzrinne'
Cc: 'P2P-SIP'


Subject: Re: [p2p-sip] DNS for inter-domain? Yes; intra-domain? No

 

Great summary. Please see my feedback inline.

 

Peter

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

	From: David Barrett <mailto:dbarrett at quinthar.com>  

	To: 'Henning Schulzrinne' <mailto:hgs at cs.columbia.edu>  

	Cc: 'P2P-SIP' <mailto:p2p-sip at cs.columbia.edu>  

	Sent: Wednesday, December 06, 2006 12:07 AM

	Subject: Re: [p2p-sip] DNS for inter-domain? Yes; intra-domain? No

	 

	I think the value of this discussion is in identifying and clarifying misimpressions about how DNS and mDNS work (and do not work) so we can make a more informed choice on whether or not to use them.  Already we've clarified:

	 

	How can DNS records be automatically updated? (Through dynamic DNS protocols)

	- [pp]: what is the cost and feasibility to upgrade existing DNS infrastructure to support dynamic DNS globally? is it similiar to the deployment of ipv6 and ip mobility?

	- How fast can DNS propagate?  (Extremely fast with short TTLs)

	- What if I have multiple SIP proxies? (Put multiple IPs into the DNS record) 

	Is DNS only for IP addresses? (No, can store other records) 

	- [pp]: can/will these DNS server also store data like voicemail or other p2p application data?

	- How does DNS work in on ad-hoc networks? (Using mDNS via multicast) 

	- Where is multicast supported? (Only everywhere it's needed - on subnets)

	- Is mDNS an IETF standard? (It's on the standards track.)

	- Is mDNS actually in use? (It's the foundation of zeroconf, which is used extensively by Apple)

	- Are there any use-cases that DHT supports that DNS/mDNS cannot? (Apparently not.)

	 

	Lack of clarity on these points could call for the unnecessary dismissal of these technologies on false premises, which I think nobody wants.

	 

	-david

	 

	
________________________________


	From: Henning Schulzrinne [mailto:hgs at cs.columbia.edu] 
	Sent: Tuesday, December 05, 2006 4:20 PM
	To: David Barrett
	Cc: 'Eunsoo Shim'; 'P2P-SIP'
	Subject: Re: [p2p-sip] DNS for inter-domain? Yes; intra-domain? No

	 

	I have to admit that I'm confused by this whole discussion. I thought we had agreement that

	 

	(1) specifying mDNS + SIP operation is useful (and hopefully simple), but easily handled within the framework of SIPPING;

	 

	(2) specifying dynamic DNS + SIP might be useful, subject to working out issues such as proxy discovery, authentication and the like and that interested parties should work out, at least to the extent of identifying the pieces that might need work (as a random example: can I expect my SIP Digest secret to work as a DNS update secret?). If, as claimed, this is easy to do, SIPPING is again the natural home for such a BCP effort, with the DNS-related WGs as the home for any DNS extensions that might prove necessary.

	 

	Maybe those individuals interested in solving (2) should form an informal design team and get beyond the hand-waving stage and identify the problems, if any, that would have to be addressed in either protocol or operational standards. I have unfortunately seen little such effort so far, although I suspect that such level of detail isn't terribly well suited to be hashed out in front of several hundred mailing list members.

	 

	 

	We also have had 2 BOFs and numerous ad-hoc meetings on this topic, with a rough consensus on the current charter. Are you trying to keep people from working on a specific topic you consider inferior?

	 

	We seem to be rehashing the same arguments again and again. 

	 

	Henning

	 

	 

	 

	On Dec 5, 2006, at 7:02 PM, David Barrett wrote:

	 

	Does the actual, real-world availability of multicast factor into the decision as to whether or not to use it?

	 

	I mean, we *could* assume that every odd IP address won't work, and then design around that.  But odd IP addresses *do* work.

	 

	Likewise, multicast *is* (or could easily be made) available for all of the real-world use cases that demand it.

	 

	It just seems like there are much more interesting *actual* problems to solve than those that don't affect real users.  But if you're excited about solving the non-multicast ad-hoc P2P-SIP problem on the assumption that someday, somebody will have reason to use it (even though we can't really brainstorm anybody who would use it today), then I guess that's a cool research project.

	 

	But it's a *research* project, not an IETF standardization exercise.  DNS and mDNS work everywhere the *actual* users are, and thus I think the IETF should focus on them, first.

	 

	-david

	 

	
________________________________


	From: Eunsoo Shim [ mailto:eunsooshim at gmail.com] 
	Sent: Tuesday, December 05, 2006 3:04 PM
	To: David Barrett
	Cc: P2P-SIP
	Subject: Re: [p2p-sip] DNS for inter-domain? Yes; intra-domain? No

	 

	I understand that mDNS can be used in ad-hoc networks where multicast works.
	Some of us, including me, have discussed about using mDNS and noticed that there is little to do to use mDNS or any other multicast-based discovery protocol. 
	Those people decided not to focus on the approach because assuming multicast availability was not general and they did not want to limit P2PSIP into such network environments.
	They said such an approach (using mDNS or similar ones) could be discussed maybe in SIPPING. 
	
	The proposed charter is targeting to address intra-domain lookup that works even if there is no multicast support.
	
	Using DNS for intra-domain lookup is not different from using a central server, which is not interesting to many of us, I believe. So transition from mDNS to DNS is not appealing, either. 
	
	What do you say?
	Thanks.
	
	Eunsoo

	On 12/5/06, David Barrett <dbarrett at quinthar.com> wrote: 

	 

	
________________________________


	From: Eunsoo Shim 
	Subject: [p2p-sip] DNS for inter-domain? Yes; intra-domain? No

	 

	So my understanding is that what P2PSIP group is trying to do is designing mechanisms for P2P-style intra-domain location lookup where location records are distributed among peers rather than in a central location server.

	...

	Am I missing any important aspect?

	I agree, this is what P2P-SIP is trying to do.  I'm just not certain why, because mDNS can do the same thing in a way that is already standardized, already deployed, and already works.  Basically, for every real-world case, it seems mDNS is adequate for ad-hoc unauthenticated namespaces, and DNS is adequate for global, authenticated namespaces.  There is already a seamless interface to graduate from mDNS to DNS depending upon what is available; it already covers the whole range of use cases (minus the extreme security cases, which I believe remain unsolved problems by any means).

	 

	I don't dispute you could make a DHT bootstrapped with UPnP or whatever.  You can build a lot of things.  I just don't see why we should bother when there are easier, adequate technologies available.

	 

	-david

	 

	_______________________________________________

	p2p-sip mailing list

	p2p-sip at cs.columbia.edu 

	https://lists.cs.columbia.edu/cucslists/listinfo/p2p-sip 

	 

	
________________________________


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


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



 

-------------- next part --------------
An HTML attachment was scrubbed...
URL: https://lists.cs.columbia.edu/pipermail/p2p-sip/attachments/20061207/7c7dacb8/attachment-0001.html

From dbarrett at quinthar.com  Thu Dec  7 20:23:47 2006
From: dbarrett at quinthar.com (David Barrett)
Date: Thu, 7 Dec 2006 17:23:47 -0800
Subject: [p2p-sip] DNS for inter-domain? Yes; intra-domain? No
In-Reply-To: <62D92A9A02BCC845B202323D49A48426E1CF54@0591-ITS-EXMP02.us.saic.com>
Message-ID: <0a2601c71a67$8386ed40$0202a8c0@Quinthar>

I'm not very familiar with MANET, but I think it supports multicast, so mDNS
could probably play a role in it.

 

Either way, if someone goes to the trouble of setting up some kind of
grid/mesh network, they'll probably want to do more than just VoIP.  They'll
need *some* kind of DNS to use for everything else.  It's probable they'll
find a way to tunnel normal DNS.  It's possible they'll find a way to use
normal mDNS.  Before inventing a custom, non-DNS solution, it seems sensible
to see how they're going to solve the DNS problem first and put our weight
behind that, rather than inventing a new, competing solution.

 

-david

 

PS: What is inadequate about DNS and mDNS for your network today?

 

  _____  

From: Roy, Radhika R. [mailto:RADHIKA.R.ROY at saic.com] 
Sent: Thursday, December 07, 2006 2:45 PM
To: David Barrett; Eunsoo Shim
Cc: P2P-SIP
Subject: RE: [p2p-sip] DNS for inter-domain? Yes; intra-domain? No

 

David:

 

Do DNS and mDNS also cover P2P Mobile Ad Hoc Networks (MANET)?

 

We are looking for MANET support right now for P2P-SIP networking.

 

Best regards,

Radhika

 

PS: We are using DNS right now, but no insufficient.

 

  _____  

From: p2p-sip-bounces at cs.columbia.edu on behalf of David Barrett
Sent: Thu 12/7/2006 4:37 PM
To: 'Eunsoo Shim'
Cc: 'P2P-SIP'
Subject: Re: [p2p-sip] DNS for inter-domain? Yes; intra-domain? No

Hm, I will agree that the conversation has covered all the useful ground
it's going to cover.  As for the confusions you raise:

 

-          DNS is used when available (internet or private network), and
mDNS otherwise (ad hoc subnet)

-          Multicast only works on subnets.  Ad hoc networks are typically
single-subnets.  Thus mDNS works on ad hoc networks.

-          Multicast does not work between subnets.  The internet and
private networks use multiple subnets.  Thus mDNS will not work across the
internet or on private networks spanning multiple subnets.  However, DNS is
typically available in this case, and thus DNS is used instead. 

-          DNS and mDNS cover the total range of real-world networking
environments.

-          I understand P2P-SIP is currently proposed to use DNS as an
optional extension to the mandatory DHT.  I think it'd be more sensible to
have mandatory support for DNS with optional DHT, particularly because the
DHT is hard work for no real-world gain and it seems wasteful to force
everyone to pay it.

-          The "P2P" system I propose as a replacement for DHT is merely
what Adam originally proposed: configure P2P-SIP UAs to update a DNS record
that points to another P2P-SIP UA as a proxy.  

-          Whether or not that is P2P "enough" is a long boring religious
discussion.  But so far as we've been able to determine, this solves all the
same use cases as the DHT, but with less work and better integration with
existing SIP.

-          P2P works; it's what pays my bills.  But it's a question of when
and where is appropriate.

Please let me know if you have any more questions about this idea or need me
to clarify it more, but I think we've covered it exhaustively.

 

-david

 

  _____  

From: Eunsoo Shim [mailto:eunsooshim at gmail.com] 
Sent: Thursday, December 07, 2006 4:12 AM
To: David Barrett
Cc: Peter Pan; Henning Schulzrinne; P2P-SIP
Subject: Re: [p2p-sip] DNS for inter-domain? Yes; intra-domain? No

 

mDNS works only in a subnet. If you limit P2PSIP applications only to that
environment, mDNS should work. You can go to SIPPING to talk about it. Many
of us understand it.
If you suggest mDNS even for other environments, please have the Internet
changed to support multicast almost everywhere first. Then we can talk about
using mDNS for P2PSIP in a general sense. 

dynDNS is just another DNS system, i.e. relying on DNS SERVERs. Using dynDNS
is same as retrieving a redirect server information from DNS and getting the
location of a particular user from the redirect server. That means, even we
don't have to update the DNS infrastructure! Just use existing DNS and SIP
and you can get what you propose.

You continuously try to imply that P2PSIP does not use DNS at all. I pointed
out that DNS was assumed for inter-domain connections for P2PSIP (and
actually SIP). All left from your argument is that we got to use DNS servers
even for intra-domain lookups, which is equivalent to using a central
location server. 

Now it sounds to me that you are saying "p2p does not work. Stick to central
servers!"
I have no interest in arguing whether C/S architecture or P2P architecture
is better. Neither are many people in this list, I believe. We understand
there are tradeoffs. Maybe you can enjoy that kind of discussion with people
from Skype or eDonkey. 

I think we passed already time for that kind of discussion here.

Already the current SIP systems use central servers for location
registration and voicemail storage. 
If you are interested in another way/protocol of using central servers for
SIP, this is not a right place to propose it. 
If you have a problem with DHT, please propose a different 'P2P' mechanism
for 'P2PSIP'.


Thanks.

Eunsoo




On 12/6/06, David Barrett <dbarrett at quinthar.com> wrote:

I'm not sure the DNS infrastructure needs much updating at all.  Perhaps
there will be more demand for DNS registrars to offer automated (aka
"dynamic") record updating, but that's really just standardizing what they
already do.

 

As for voicemail, no, I wouldn't expect to put that into DNS; that's
something best handled by some real storage provider.  You *could* do it in
a DHT, but it'd just be slower and less reliable, and the engineering cost
of building that would pay for central storage many, many times over.

 

Again, in a vacuum without taking anything else into consideration, be it
engineering cost, performance, reliability, existing systems, standards
support and so on, yes, a DHT could be used.  But when everything else is
factored in, it just makes no sense at all.

 

-david

 

  _____  

From: Peter Pan [mailto:huang-ming.pan at comcast.net] 
Sent: Wednesday, December 06, 2006 12:09 PM
To: David Barrett; 'Henning Schulzrinne'
Cc: 'P2P-SIP'


Subject: Re: [p2p-sip] DNS for inter-domain? Yes; intra-domain? No

 

Great summary. Please see my feedback inline.

 

Peter

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

From: David Barrett <mailto:dbarrett at quinthar.com>  

To: 'Henning Schulzrinne' <mailto:hgs at cs.columbia.edu>  

Cc: 'P2P-SIP' <mailto:p2p-sip at cs.columbia.edu>  

Sent: Wednesday, December 06, 2006 12:07 AM

Subject: Re: [p2p-sip] DNS for inter-domain? Yes; intra-domain? No

 

I think the value of this discussion is in identifying and clarifying
misimpressions about how DNS and mDNS work (and do not work) so we can make
a more informed choice on whether or not to use them.  Already we've
clarified:

 

How can DNS records be automatically updated? (Through dynamic DNS
protocols)

- [pp]: what is the cost and feasibility to upgrade existing DNS
infrastructure to support dynamic DNS globally? is it similiar to the
deployment of ipv6 and ip mobility?

- How fast can DNS propagate?  (Extremely fast with short TTLs)

- What if I have multiple SIP proxies? (Put multiple IPs into the DNS
record) 

Is DNS only for IP addresses? (No, can store other records) 

- [pp]: can/will these DNS server also store data like voicemail or other
p2p application data?

- How does DNS work in on ad-hoc networks? (Using mDNS via multicast) 

- Where is multicast supported? (Only everywhere it's needed - on subnets)

- Is mDNS an IETF standard? (It's on the standards track.)

- Is mDNS actually in use? (It's the foundation of zeroconf, which is used
extensively by Apple)

- Are there any use-cases that DHT supports that DNS/mDNS cannot?
(Apparently not.)

 

Lack of clarity on these points could call for the unnecessary dismissal of
these technologies on false premises, which I think nobody wants.

 

-david

 


  _____  


From: Henning Schulzrinne [mailto:hgs at cs.columbia.edu] 
Sent: Tuesday, December 05, 2006 4:20 PM
To: David Barrett
Cc: 'Eunsoo Shim'; 'P2P-SIP'
Subject: Re: [p2p-sip] DNS for inter-domain? Yes; intra-domain? No

 

I have to admit that I'm confused by this whole discussion. I thought we had
agreement that

 

(1) specifying mDNS + SIP operation is useful (and hopefully simple), but
easily handled within the framework of SIPPING;

 

(2) specifying dynamic DNS + SIP might be useful, subject to working out
issues such as proxy discovery, authentication and the like and that
interested parties should work out, at least to the extent of identifying
the pieces that might need work (as a random example: can I expect my SIP
Digest secret to work as a DNS update secret?). If, as claimed, this is easy
to do, SIPPING is again the natural home for such a BCP effort, with the
DNS-related WGs as the home for any DNS extensions that might prove
necessary.

 

Maybe those individuals interested in solving (2) should form an informal
design team and get beyond the hand-waving stage and identify the problems,
if any, that would have to be addressed in either protocol or operational
standards. I have unfortunately seen little such effort so far, although I
suspect that such level of detail isn't terribly well suited to be hashed
out in front of several hundred mailing list members.

 

 

We also have had 2 BOFs and numerous ad-hoc meetings on this topic, with a
rough consensus on the current charter. Are you trying to keep people from
working on a specific topic you consider inferior?

 

We seem to be rehashing the same arguments again and again. 

 

Henning

 

 

 

On Dec 5, 2006, at 7:02 PM, David Barrett wrote:

 

Does the actual, real-world availability of multicast factor into the
decision as to whether or not to use it?

 

I mean, we *could* assume that every odd IP address won't work, and then
design around that.  But odd IP addresses *do* work.

 

Likewise, multicast *is* (or could easily be made) available for all of the
real-world use cases that demand it.

 

It just seems like there are much more interesting *actual* problems to
solve than those that don't affect real users.  But if you're excited about
solving the non-multicast ad-hoc P2P-SIP problem on the assumption that
someday, somebody will have reason to use it (even though we can't really
brainstorm anybody who would use it today), then I guess that's a cool
research project.

 

But it's a *research* project, not an IETF standardization exercise.  DNS
and mDNS work everywhere the *actual* users are, and thus I think the IETF
should focus on them, first.

 

-david

 


  _____  


From: Eunsoo Shim [ mailto:eunsooshim at gmail.com] 
Sent: Tuesday, December 05, 2006 3:04 PM
To: David Barrett
Cc: P2P-SIP
Subject: Re: [p2p-sip] DNS for inter-domain? Yes; intra-domain? No

 

I understand that mDNS can be used in ad-hoc networks where multicast works.
Some of us, including me, have discussed about using mDNS and noticed that
there is little to do to use mDNS or any other multicast-based discovery
protocol. 
Those people decided not to focus on the approach because assuming multicast
availability was not general and they did not want to limit P2PSIP into such
network environments.
They said such an approach (using mDNS or similar ones) could be discussed
maybe in SIPPING. 

The proposed charter is targeting to address intra-domain lookup that works
even if there is no multicast support.

Using DNS for intra-domain lookup is not different from using a central
server, which is not interesting to many of us, I believe. So transition
from mDNS to DNS is not appealing, either. 

What do you say?
Thanks.

Eunsoo

On 12/5/06, David Barrett <dbarrett at quinthar.com> wrote: 

 


  _____  


From: Eunsoo Shim 
Subject: [p2p-sip] DNS for inter-domain? Yes; intra-domain? No

 

So my understanding is that what P2PSIP group is trying to do is designing
mechanisms for P2P-style intra-domain location lookup where location records
are distributed among peers rather than in a central location server.

.

Am I missing any important aspect?

I agree, this is what P2P-SIP is trying to do.  I'm just not certain why,
because mDNS can do the same thing in a way that is already standardized,
already deployed, and already works.  Basically, for every real-world case,
it seems mDNS is adequate for ad-hoc unauthenticated namespaces, and DNS is
adequate for global, authenticated namespaces.  There is already a seamless
interface to graduate from mDNS to DNS depending upon what is available; it
already covers the whole range of use cases (minus the extreme security
cases, which I believe remain unsolved problems by any means).

 

I don't dispute you could make a DHT bootstrapped with UPnP or whatever.
You can build a lot of things.  I just don't see why we should bother when
there are easier, adequate technologies available.

 

-david

 

_______________________________________________

p2p-sip mailing list

p2p-sip at cs.columbia.edu 

https://lists.cs.columbia.edu/cucslists/listinfo/p2p-sip 

 


  _____  


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


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

 

-------------- next part --------------
An HTML attachment was scrubbed...
URL: https://lists.cs.columbia.edu/pipermail/p2p-sip/attachments/20061207/ddcab807/attachment-0001.html

From RADHIKA.R.ROY at saic.com  Thu Dec  7 23:51:07 2006
From: RADHIKA.R.ROY at saic.com (Roy, Radhika R.)
Date: Thu, 7 Dec 2006 23:51:07 -0500 
Subject: [p2p-sip] DNS for inter-domain? Yes; intra-domain? No
Message-ID: <62D92A9A02BCC845B202323D49A48426E1CF56@0591-ITS-EXMP02.us.saic.com>

Issue is this: Mobile ad hoc network - no fixed infrastructure where
everything is P2P. No one knows who will be the next peer after a certain
period of time. The speed of each peer may vary.
 
DNS is used assuming that there is at least one fixed non-mobile node. It is
the problem that we want to get rid of.
 
First, let us understand what works in this environment.
 
IETF MANET WG has all the drafts and RFCs.
 
Best regards,
Radhika
 
PS: 
 
1. It is not about DNS or mDNS or multicasting per se.
2. Let us find our what works in the above environment, i.e., MANET.
Everything is distributive as P2P in mobile ad hoc environment. Period.
3. The routing protocol for MANET that will be used in the about environment
best commercially (to be accepted as standard soon as all alternative
protocols have provided their performance and simulations results, but this
is expected to be chosen as s finalist - draft is available in the IETF
MANET WG Archive): OSPFv3-MDR.
4. We may have to enhance the charter for the P2P-SIP in the future as the
present scope of this would be WG is not enough. However, it is a different
story as a first step let us accomplish what is in the present charter.
 
 
 

  _____  

From: David Barrett [mailto:dbarrett at quinthar.com]
Sent: Thu 12/7/2006 8:23 PM
To: 'Roy, Radhika R.'; 'Eunsoo Shim'
Cc: 'P2P-SIP'
Subject: RE: [p2p-sip] DNS for inter-domain? Yes; intra-domain? No



I'm not very familiar with MANET, but I think it supports multicast, so mDNS
could probably play a role in it.

 

Either way, if someone goes to the trouble of setting up some kind of
grid/mesh network, they'll probably want to do more than just VoIP.  They'll
need *some* kind of DNS to use for everything else.  It's probable they'll
find a way to tunnel normal DNS.  It's possible they'll find a way to use
normal mDNS.  Before inventing a custom, non-DNS solution, it seems sensible
to see how they're going to solve the DNS problem first and put our weight
behind that, rather than inventing a new, competing solution.

 

-david

 

PS: What is inadequate about DNS and mDNS for your network today?

 

  _____  

From: Roy, Radhika R. [mailto:RADHIKA.R.ROY at saic.com] 
Sent: Thursday, December 07, 2006 2:45 PM
To: David Barrett; Eunsoo Shim
Cc: P2P-SIP
Subject: RE: [p2p-sip] DNS for inter-domain? Yes; intra-domain? No

 

David:

 

Do DNS and mDNS also cover P2P Mobile Ad Hoc Networks (MANET)?

 

We are looking for MANET support right now for P2P-SIP networking.

 

Best regards,

Radhika

 

PS: We are using DNS right now, but no insufficient.

 

  _____  

From: p2p-sip-bounces at cs.columbia.edu on behalf of David Barrett
Sent: Thu 12/7/2006 4:37 PM
To: 'Eunsoo Shim'
Cc: 'P2P-SIP'
Subject: Re: [p2p-sip] DNS for inter-domain? Yes; intra-domain? No

Hm, I will agree that the conversation has covered all the useful ground
it's going to cover.  As for the confusions you raise:

 

-          DNS is used when available (internet or private network), and
mDNS otherwise (ad hoc subnet)

-          Multicast only works on subnets.  Ad hoc networks are typically
single-subnets.  Thus mDNS works on ad hoc networks.

-          Multicast does not work between subnets.  The internet and
private networks use multiple subnets.  Thus mDNS will not work across the
internet or on private networks spanning multiple subnets.  However, DNS is
typically available in this case, and thus DNS is used instead. 

-          DNS and mDNS cover the total range of real-world networking
environments.

-          I understand P2P-SIP is currently proposed to use DNS as an
optional extension to the mandatory DHT.  I think it'd be more sensible to
have mandatory support for DNS with optional DHT, particularly because the
DHT is hard work for no real-world gain and it seems wasteful to force
everyone to pay it.

-          The "P2P" system I propose as a replacement for DHT is merely
what Adam originally proposed: configure P2P-SIP UAs to update a DNS record
that points to another P2P-SIP UA as a proxy.  

-          Whether or not that is P2P "enough" is a long boring religious
discussion.  But so far as we've been able to determine, this solves all the
same use cases as the DHT, but with less work and better integration with
existing SIP.

-          P2P works; it's what pays my bills.  But it's a question of when
and where is appropriate.

Please let me know if you have any more questions about this idea or need me
to clarify it more, but I think we've covered it exhaustively.

 

-david

 

  _____  

From: Eunsoo Shim [mailto:eunsooshim at gmail.com] 
Sent: Thursday, December 07, 2006 4:12 AM
To: David Barrett
Cc: Peter Pan; Henning Schulzrinne; P2P-SIP
Subject: Re: [p2p-sip] DNS for inter-domain? Yes; intra-domain? No

 

mDNS works only in a subnet. If you limit P2PSIP applications only to that
environment, mDNS should work. You can go to SIPPING to talk about it. Many
of us understand it.
If you suggest mDNS even for other environments, please have the Internet
changed to support multicast almost everywhere first. Then we can talk about
using mDNS for P2PSIP in a general sense. 

dynDNS is just another DNS system, i.e. relying on DNS SERVERs. Using dynDNS
is same as retrieving a redirect server information from DNS and getting the
location of a particular user from the redirect server. That means, even we
don't have to update the DNS infrastructure! Just use existing DNS and SIP
and you can get what you propose.

You continuously try to imply that P2PSIP does not use DNS at all. I pointed
out that DNS was assumed for inter-domain connections for P2PSIP (and
actually SIP). All left from your argument is that we got to use DNS servers
even for intra-domain lookups, which is equivalent to using a central
location server. 

Now it sounds to me that you are saying "p2p does not work. Stick to central
servers!"
I have no interest in arguing whether C/S architecture or P2P architecture
is better. Neither are many people in this list, I believe. We understand
there are tradeoffs. Maybe you can enjoy that kind of discussion with people
from Skype or eDonkey. 

I think we passed already time for that kind of discussion here.

Already the current SIP systems use central servers for location
registration and voicemail storage. 
If you are interested in another way/protocol of using central servers for
SIP, this is not a right place to propose it. 
If you have a problem with DHT, please propose a different 'P2P' mechanism
for 'P2PSIP'.


Thanks.

Eunsoo




On 12/6/06, David Barrett <dbarrett at quinthar.com
<mailto:dbarrett at quinthar.com> > wrote:

I'm not sure the DNS infrastructure needs much updating at all.  Perhaps
there will be more demand for DNS registrars to offer automated (aka
"dynamic") record updating, but that's really just standardizing what they
already do.

 

As for voicemail, no, I wouldn't expect to put that into DNS; that's
something best handled by some real storage provider.  You *could* do it in
a DHT, but it'd just be slower and less reliable, and the engineering cost
of building that would pay for central storage many, many times over.

 

Again, in a vacuum without taking anything else into consideration, be it
engineering cost, performance, reliability, existing systems, standards
support and so on, yes, a DHT could be used.  But when everything else is
factored in, it just makes no sense at all.

 

-david

 

  _____  

From: Peter Pan [mailto:huang-ming.pan at comcast.net
<mailto:huang-ming.pan at comcast.net> ] 
Sent: Wednesday, December 06, 2006 12:09 PM
To: David Barrett; 'Henning Schulzrinne'
Cc: 'P2P-SIP'


Subject: Re: [p2p-sip] DNS for inter-domain? Yes; intra-domain? No

 

Great summary. Please see my feedback inline.

 

Peter

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

From: David Barrett <mailto:dbarrett at quinthar.com>  

To: 'Henning Schulzrinne' <mailto:hgs at cs.columbia.edu>  

Cc: 'P2P-SIP' <mailto:p2p-sip at cs.columbia.edu>  

Sent: Wednesday, December 06, 2006 12:07 AM

Subject: Re: [p2p-sip] DNS for inter-domain? Yes; intra-domain? No

 

I think the value of this discussion is in identifying and clarifying
misimpressions about how DNS and mDNS work (and do not work) so we can make
a more informed choice on whether or not to use them.  Already we've
clarified:

 

How can DNS records be automatically updated? (Through dynamic DNS
protocols)

- [pp]: what is the cost and feasibility to upgrade existing DNS
infrastructure to support dynamic DNS globally? is it similiar to the
deployment of ipv6 and ip mobility?

- How fast can DNS propagate?  (Extremely fast with short TTLs)

- What if I have multiple SIP proxies? (Put multiple IPs into the DNS
record) 

Is DNS only for IP addresses? (No, can store other records) 

- [pp]: can/will these DNS server also store data like voicemail or other
p2p application data?

- How does DNS work in on ad-hoc networks? (Using mDNS via multicast) 

- Where is multicast supported? (Only everywhere it's needed - on subnets)

- Is mDNS an IETF standard? (It's on the standards track.)

- Is mDNS actually in use? (It's the foundation of zeroconf, which is used
extensively by Apple)

- Are there any use-cases that DHT supports that DNS/mDNS cannot?
(Apparently not.)

 

Lack of clarity on these points could call for the unnecessary dismissal of
these technologies on false premises, which I think nobody wants.

 

-david

 


  _____  


From: Henning Schulzrinne [mailto:hgs at cs.columbia.edu
<mailto:hgs at cs.columbia.edu> ] 
Sent: Tuesday, December 05, 2006 4:20 PM
To: David Barrett
Cc: 'Eunsoo Shim'; 'P2P-SIP'
Subject: Re: [p2p-sip] DNS for inter-domain? Yes; intra-domain? No

 

I have to admit that I'm confused by this whole discussion. I thought we had
agreement that

 

(1) specifying mDNS + SIP operation is useful (and hopefully simple), but
easily handled within the framework of SIPPING;

 

(2) specifying dynamic DNS + SIP might be useful, subject to working out
issues such as proxy discovery, authentication and the like and that
interested parties should work out, at least to the extent of identifying
the pieces that might need work (as a random example: can I expect my SIP
Digest secret to work as a DNS update secret?). If, as claimed, this is easy
to do, SIPPING is again the natural home for such a BCP effort, with the
DNS-related WGs as the home for any DNS extensions that might prove
necessary.

 

Maybe those individuals interested in solving (2) should form an informal
design team and get beyond the hand-waving stage and identify the problems,
if any, that would have to be addressed in either protocol or operational
standards. I have unfortunately seen little such effort so far, although I
suspect that such level of detail isn't terribly well suited to be hashed
out in front of several hundred mailing list members.

 

 

We also have had 2 BOFs and numerous ad-hoc meetings on this topic, with a
rough consensus on the current charter. Are you trying to keep people from
working on a specific topic you consider inferior?

 

We seem to be rehashing the same arguments again and again. 

 

Henning

 

 

 

On Dec 5, 2006, at 7:02 PM, David Barrett wrote:

 

Does the actual, real-world availability of multicast factor into the
decision as to whether or not to use it?

 

I mean, we *could* assume that every odd IP address won't work, and then
design around that.  But odd IP addresses *do* work.

 

Likewise, multicast *is* (or could easily be made) available for all of the
real-world use cases that demand it.

 

It just seems like there are much more interesting *actual* problems to
solve than those that don't affect real users.  But if you're excited about
solving the non-multicast ad-hoc P2P-SIP problem on the assumption that
someday, somebody will have reason to use it (even though we can't really
brainstorm anybody who would use it today), then I guess that's a cool
research project.

 

But it's a *research* project, not an IETF standardization exercise.  DNS
and mDNS work everywhere the *actual* users are, and thus I think the IETF
should focus on them, first.

 

-david

 


  _____  


From: Eunsoo Shim [ mailto:eunsooshim at gmail.com
<mailto:eunsooshim at gmail.com> ] 
Sent: Tuesday, December 05, 2006 3:04 PM
To: David Barrett
Cc: P2P-SIP
Subject: Re: [p2p-sip] DNS for inter-domain? Yes; intra-domain? No

 

I understand that mDNS can be used in ad-hoc networks where multicast works.
Some of us, including me, have discussed about using mDNS and noticed that
there is little to do to use mDNS or any other multicast-based discovery
protocol. 
Those people decided not to focus on the approach because assuming multicast
availability was not general and they did not want to limit P2PSIP into such
network environments.
They said such an approach (using mDNS or similar ones) could be discussed
maybe in SIPPING. 

The proposed charter is targeting to address intra-domain lookup that works
even if there is no multicast support.

Using DNS for intra-domain lookup is not different from using a central
server, which is not interesting to many of us, I believe. So transition
from mDNS to DNS is not appealing, either. 

What do you say?
Thanks.

Eunsoo

On 12/5/06, David Barrett <dbarrett at quinthar.com
<mailto:dbarrett at quinthar.com> > wrote: 

 


  _____  


From: Eunsoo Shim 
Subject: [p2p-sip] DNS for inter-domain? Yes; intra-domain? No

 

So my understanding is that what P2PSIP group is trying to do is designing
mechanisms for P2P-style intra-domain location lookup where location records
are distributed among peers rather than in a central location server.

...

Am I missing any important aspect?

I agree, this is what P2P-SIP is trying to do.  I'm just not certain why,
because mDNS can do the same thing in a way that is already standardized,
already deployed, and already works.  Basically, for every real-world case,
it seems mDNS is adequate for ad-hoc unauthenticated namespaces, and DNS is
adequate for global, authenticated namespaces.  There is already a seamless
interface to graduate from mDNS to DNS depending upon what is available; it
already covers the whole range of use cases (minus the extreme security
cases, which I believe remain unsolved problems by any means).

 

I don't dispute you could make a DHT bootstrapped with UPnP or whatever.
You can build a lot of things.  I just don't see why we should bother when
there are easier, adequate technologies available.

 

-david

 

_______________________________________________

p2p-sip mailing list

p2p-sip at cs.columbia.edu <mailto:p2p-sip at cs.columbia.edu>  

https://lists.cs.columbia.edu/cucslists/listinfo/p2p-sip
<https://lists.cs.columbia.edu/cucslists/listinfo/p2p-sip> 

 


  _____  


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


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

 


From jiang.x.f at huawei.com  Fri Dec  8 02:47:24 2006
From: jiang.x.f at huawei.com (JiangXingFeng)
Date: Fri, 08 Dec 2006 15:47:24 +0800
Subject: [p2p-sip] Concepts draft: Open Issue #1: Searching for services
In-Reply-To: <B784DF3B-10C8-401B-BCA4-149F7AF0ECD0@magma.ca>
Message-ID: <000001c71a9d$1a0ac6d0$7205a40a@china.huawei.com>

Hi, Philip:
  I have some comments, please see inline.
Regards
--
Jiang XingFeng
--------------------------------------------------------------------

CONFIDENTIALITY NOTICE

This e-mail and its attachments contain confidential information from
HUAWEI, which is intended only for the person or entity whose address is
listed above. Any use of the information contained herein in any way
(including, but not limited to, total or partial disclosure, reproduction,
or dissemination) by persons other than the intended recipient(s) is
prohibited. If you receive this e-mail in error, please notify the sender by
phone or email immediately and delete it!

--------------------------------------------------------------------
> -----Original Message-----
> From: p2p-sip-bounces at cs.columbia.edu
> [mailto:p2p-sip-bounces at cs.columbia.edu] On Behalf Of Philip Matthews
> Sent: Friday, December 08, 2006 1:48 AM
> To: P2P-SIP
> Cc: David A. Bryan
> Subject: [p2p-sip] Concepts draft: Open Issue #1: Searching for services
> 
> Open issue #1:  Can different peers in a P2PSIP Overlay offer
> different services to the Overlay? If so, do we need the
> concept of  searching for the set of peers that offer a service?
> 
> 

> Proposed Resolution: Yes and Yes. Though we may require
> that all peers offer some minimal set of services, there is likely
> to be at least one service  that is not offered by every peer.
> Thus we need the concept of searching for the set of peers
> that offer a service.
For the first issue, I agree with the proposed resolution. Peers in the
overlay have different capabilities other than basic storage and routing
service. These capabilities could be used to make some applications easier
to implement.
For the second one, I think if we think of different service as resource
like SIP URI and put these resources into overlay, it is easy to get these
resource by common search mechanism of DHT. 
A problem following the above open issues is: if there are several peers
providing the same service, how does the system choose the appropriate one
or more to serve other peers? Maybe the criteria are net-path closest away
from service requesting peers, or performance of service providing peer? I
think the criteria are dependent on the specific application. Is this
problem in the scope of P2PSIP or out of scope? 
> 
> For now, we are proposing to leave undefined the detailed
> properties of the search. Thus we are proposing not
> to answer questions like whether the search finds all peers
> that offer the service, or whether the search returns the
> "closest" peer. Instead, we propose to add these questions to the
> design questions section of the document.
> 
> Comments?
> 
> - Philip
> 
> 
> _______________________________________________
> p2p-sip mailing list
> p2p-sip at cs.columbia.edu
> https://lists.cs.columbia.edu/cucslists/listinfo/p2p-sip



From dbarrett at quinthar.com  Fri Dec  8 03:37:06 2006
From: dbarrett at quinthar.com (David Barrett)
Date: Fri, 8 Dec 2006 00:37:06 -0800
Subject: [p2p-sip] DNS for inter-domain? Yes; intra-domain? No
In-Reply-To: <62D92A9A02BCC845B202323D49A48426E1CF56@0591-ITS-EXMP02.us.saic.com>
Message-ID: <000401c71aa4$13449950$0202a8c0@Quinthar>

I'm not sure I follow you here.

The "m" in "mDNS" stands for multicast because you send your DNS requests to
a multicast channel instead of to a fixed DNS server.  Anyone who knows the
answer broadcasts a reply.  It's designed precisely for an ad hoc network
with no fixed nodes.

Does this address your concern?

-david

> -----Original Message-----
> From: Roy, Radhika R. [mailto:RADHIKA.R.ROY at saic.com]
> Sent: Thursday, December 07, 2006 8:51 PM
> To: David Barrett; Eunsoo Shim
> Cc: P2P-SIP
> Subject: RE: [p2p-sip] DNS for inter-domain? Yes; intra-domain? No
> 
> Issue is this: Mobile ad hoc network - no fixed infrastructure where
> everything is P2P. No one knows who will be the next peer after a certain
> period of time. The speed of each peer may vary.
> 
> DNS is used assuming that there is at least one fixed non-mobile node. It
> is
> the problem that we want to get rid of.
> 
> First, let us understand what works in this environment.
> 
> IETF MANET WG has all the drafts and RFCs.
> 
> Best regards,
> Radhika
> 
> PS:
> 
> 1. It is not about DNS or mDNS or multicasting per se.
> 2. Let us find our what works in the above environment, i.e., MANET.
> Everything is distributive as P2P in mobile ad hoc environment. Period.
> 3. The routing protocol for MANET that will be used in the about
> environment
> best commercially (to be accepted as standard soon as all alternative
> protocols have provided their performance and simulations results, but
> this
> is expected to be chosen as s finalist - draft is available in the IETF
> MANET WG Archive): OSPFv3-MDR.
> 4. We may have to enhance the charter for the P2P-SIP in the future as the
> present scope of this would be WG is not enough. However, it is a
> different
> story as a first step let us accomplish what is in the present charter.
> 
> 
> 
> 
>   _____
> 
> From: David Barrett [mailto:dbarrett at quinthar.com]
> Sent: Thu 12/7/2006 8:23 PM
> To: 'Roy, Radhika R.'; 'Eunsoo Shim'
> Cc: 'P2P-SIP'
> Subject: RE: [p2p-sip] DNS for inter-domain? Yes; intra-domain? No
> 
> 
> 
> I'm not very familiar with MANET, but I think it supports multicast, so
> mDNS
> could probably play a role in it.
> 
> 
> 
> Either way, if someone goes to the trouble of setting up some kind of
> grid/mesh network, they'll probably want to do more than just VoIP.
> They'll
> need *some* kind of DNS to use for everything else.  It's probable they'll
> find a way to tunnel normal DNS.  It's possible they'll find a way to use
> normal mDNS.  Before inventing a custom, non-DNS solution, it seems
> sensible
> to see how they're going to solve the DNS problem first and put our weight
> behind that, rather than inventing a new, competing solution.
> 
> 
> 
> -david
> 
> 
> 
> PS: What is inadequate about DNS and mDNS for your network today?
> 
> 
> 
>   _____
> 
> From: Roy, Radhika R. [mailto:RADHIKA.R.ROY at saic.com]
> Sent: Thursday, December 07, 2006 2:45 PM
> To: David Barrett; Eunsoo Shim
> Cc: P2P-SIP
> Subject: RE: [p2p-sip] DNS for inter-domain? Yes; intra-domain? No
> 
> 
> 
> David:
> 
> 
> 
> Do DNS and mDNS also cover P2P Mobile Ad Hoc Networks (MANET)?
> 
> 
> 
> We are looking for MANET support right now for P2P-SIP networking.
> 
> 
> 
> Best regards,
> 
> Radhika
> 
> 
> 
> PS: We are using DNS right now, but no insufficient.
> 
> 
> 
>   _____
> 
> From: p2p-sip-bounces at cs.columbia.edu on behalf of David Barrett
> Sent: Thu 12/7/2006 4:37 PM
> To: 'Eunsoo Shim'
> Cc: 'P2P-SIP'
> Subject: Re: [p2p-sip] DNS for inter-domain? Yes; intra-domain? No
> 
> Hm, I will agree that the conversation has covered all the useful ground
> it's going to cover.  As for the confusions you raise:
> 
> 
> 
> -          DNS is used when available (internet or private network), and
> mDNS otherwise (ad hoc subnet)
> 
> -          Multicast only works on subnets.  Ad hoc networks are typically
> single-subnets.  Thus mDNS works on ad hoc networks.
> 
> -          Multicast does not work between subnets.  The internet and
> private networks use multiple subnets.  Thus mDNS will not work across the
> internet or on private networks spanning multiple subnets.  However, DNS
> is
> typically available in this case, and thus DNS is used instead.
> 
> -          DNS and mDNS cover the total range of real-world networking
> environments.
> 
> -          I understand P2P-SIP is currently proposed to use DNS as an
> optional extension to the mandatory DHT.  I think it'd be more sensible to
> have mandatory support for DNS with optional DHT, particularly because the
> DHT is hard work for no real-world gain and it seems wasteful to force
> everyone to pay it.
> 
> -          The "P2P" system I propose as a replacement for DHT is merely
> what Adam originally proposed: configure P2P-SIP UAs to update a DNS
> record
> that points to another P2P-SIP UA as a proxy.
> 
> -          Whether or not that is P2P "enough" is a long boring religious
> discussion.  But so far as we've been able to determine, this solves all
> the
> same use cases as the DHT, but with less work and better integration with
> existing SIP.
> 
> -          P2P works; it's what pays my bills.  But it's a question of
> when
> and where is appropriate.
> 
> Please let me know if you have any more questions about this idea or need
> me
> to clarify it more, but I think we've covered it exhaustively.
> 
> 
> 
> -david
> 
> 
> 
>   _____
> 
> From: Eunsoo Shim [mailto:eunsooshim at gmail.com]
> Sent: Thursday, December 07, 2006 4:12 AM
> To: David Barrett
> Cc: Peter Pan; Henning Schulzrinne; P2P-SIP
> Subject: Re: [p2p-sip] DNS for inter-domain? Yes; intra-domain? No
> 
> 
> 
> mDNS works only in a subnet. If you limit P2PSIP applications only to that
> environment, mDNS should work. You can go to SIPPING to talk about it.
> Many
> of us understand it.
> If you suggest mDNS even for other environments, please have the Internet
> changed to support multicast almost everywhere first. Then we can talk
> about
> using mDNS for P2PSIP in a general sense.
> 
> dynDNS is just another DNS system, i.e. relying on DNS SERVERs. Using
> dynDNS
> is same as retrieving a redirect server information from DNS and getting
> the
> location of a particular user from the redirect server. That means, even
> we
> don't have to update the DNS infrastructure! Just use existing DNS and SIP
> and you can get what you propose.
> 
> You continuously try to imply that P2PSIP does not use DNS at all. I
> pointed
> out that DNS was assumed for inter-domain connections for P2PSIP (and
> actually SIP). All left from your argument is that we got to use DNS
> servers
> even for intra-domain lookups, which is equivalent to using a central
> location server.
> 
> Now it sounds to me that you are saying "p2p does not work. Stick to
> central
> servers!"
> I have no interest in arguing whether C/S architecture or P2P architecture
> is better. Neither are many people in this list, I believe. We understand
> there are tradeoffs. Maybe you can enjoy that kind of discussion with
> people
> from Skype or eDonkey.
> 
> I think we passed already time for that kind of discussion here.
> 
> Already the current SIP systems use central servers for location
> registration and voicemail storage.
> If you are interested in another way/protocol of using central servers for
> SIP, this is not a right place to propose it.
> If you have a problem with DHT, please propose a different 'P2P' mechanism
> for 'P2PSIP'.
> 
> 
> Thanks.
> 
> Eunsoo
> 
> 
> 
> 
> On 12/6/06, David Barrett <dbarrett at quinthar.com
> <mailto:dbarrett at quinthar.com> > wrote:
> 
> I'm not sure the DNS infrastructure needs much updating at all.  Perhaps
> there will be more demand for DNS registrars to offer automated (aka
> "dynamic") record updating, but that's really just standardizing what they
> already do.
> 
> 
> 
> As for voicemail, no, I wouldn't expect to put that into DNS; that's
> something best handled by some real storage provider.  You *could* do it
> in
> a DHT, but it'd just be slower and less reliable, and the engineering cost
> of building that would pay for central storage many, many times over.
> 
> 
> 
> Again, in a vacuum without taking anything else into consideration, be it
> engineering cost, performance, reliability, existing systems, standards
> support and so on, yes, a DHT could be used.  But when everything else is
> factored in, it just makes no sense at all.
> 
> 
> 
> -david
> 
> 
> 
>   _____
> 
> From: Peter Pan [mailto:huang-ming.pan at comcast.net
> <mailto:huang-ming.pan at comcast.net> ]
> Sent: Wednesday, December 06, 2006 12:09 PM
> To: David Barrett; 'Henning Schulzrinne'
> Cc: 'P2P-SIP'
> 
> 
> Subject: Re: [p2p-sip] DNS for inter-domain? Yes; intra-domain? No
> 
> 
> 
> Great summary. Please see my feedback inline.
> 
> 
> 
> Peter
> 
> ----- Original Message -----
> 
> From: David Barrett <mailto:dbarrett at quinthar.com>
> 
> To: 'Henning Schulzrinne' <mailto:hgs at cs.columbia.edu>
> 
> Cc: 'P2P-SIP' <mailto:p2p-sip at cs.columbia.edu>
> 
> Sent: Wednesday, December 06, 2006 12:07 AM
> 
> Subject: Re: [p2p-sip] DNS for inter-domain? Yes; intra-domain? No
> 
> 
> 
> I think the value of this discussion is in identifying and clarifying
> misimpressions about how DNS and mDNS work (and do not work) so we can
> make
> a more informed choice on whether or not to use them.  Already we've
> clarified:
> 
> 
> 
> How can DNS records be automatically updated? (Through dynamic DNS
> protocols)
> 
> - [pp]: what is the cost and feasibility to upgrade existing DNS
> infrastructure to support dynamic DNS globally? is it similiar to the
> deployment of ipv6 and ip mobility?
> 
> - How fast can DNS propagate?  (Extremely fast with short TTLs)
> 
> - What if I have multiple SIP proxies? (Put multiple IPs into the DNS
> record)
> 
> Is DNS only for IP addresses? (No, can store other records)
> 
> - [pp]: can/will these DNS server also store data like voicemail or other
> p2p application data?
> 
> - How does DNS work in on ad-hoc networks? (Using mDNS via multicast)
> 
> - Where is multicast supported? (Only everywhere it's needed - on subnets)
> 
> - Is mDNS an IETF standard? (It's on the standards track.)
> 
> - Is mDNS actually in use? (It's the foundation of zeroconf, which is used
> extensively by Apple)
> 
> - Are there any use-cases that DHT supports that DNS/mDNS cannot?
> (Apparently not.)
> 
> 
> 
> Lack of clarity on these points could call for the unnecessary dismissal
> of
> these technologies on false premises, which I think nobody wants.
> 
> 
> 
> -david
> 
> 
> 
> 
>   _____
> 
> 
> From: Henning Schulzrinne [mailto:hgs at cs.columbia.edu
> <mailto:hgs at cs.columbia.edu> ]
> Sent: Tuesday, December 05, 2006 4:20 PM
> To: David Barrett
> Cc: 'Eunsoo Shim'; 'P2P-SIP'
> Subject: Re: [p2p-sip] DNS for inter-domain? Yes; intra-domain? No
> 
> 
> 
> I have to admit that I'm confused by this whole discussion. I thought we
> had
> agreement that
> 
> 
> 
> (1) specifying mDNS + SIP operation is useful (and hopefully simple), but
> easily handled within the framework of SIPPING;
> 
> 
> 
> (2) specifying dynamic DNS + SIP might be useful, subject to working out
> issues such as proxy discovery, authentication and the like and that
> interested parties should work out, at least to the extent of identifying
> the pieces that might need work (as a random example: can I expect my SIP
> Digest secret to work as a DNS update secret?). If, as claimed, this is
> easy
> to do, SIPPING is again the natural home for such a BCP effort, with the
> DNS-related WGs as the home for any DNS extensions that might prove
> necessary.
> 
> 
> 
> Maybe those individuals interested in solving (2) should form an informal
> design team and get beyond the hand-waving stage and identify the
> problems,
> if any, that would have to be addressed in either protocol or operational
> standards. I have unfortunately seen little such effort so far, although I
> suspect that such level of detail isn't terribly well suited to be hashed
> out in front of several hundred mailing list members.
> 
> 
> 
> 
> 
> We also have had 2 BOFs and numerous ad-hoc meetings on this topic, with a
> rough consensus on the current charter. Are you trying to keep people from
> working on a specific topic you consider inferior?
> 
> 
> 
> We seem to be rehashing the same arguments again and again.
> 
> 
> 
> Henning
> 
> 
> 
> 
> 
> 
> 
> On Dec 5, 2006, at 7:02 PM, David Barrett wrote:
> 
> 
> 
> Does the actual, real-world availability of multicast factor into the
> decision as to whether or not to use it?
> 
> 
> 
> I mean, we *could* assume that every odd IP address won't work, and then
> design around that.  But odd IP addresses *do* work.
> 
> 
> 
> Likewise, multicast *is* (or could easily be made) available for all of
> the
> real-world use cases that demand it.
> 
> 
> 
> It just seems like there are much more interesting *actual* problems to
> solve than those that don't affect real users.  But if you're excited
> about
> solving the non-multicast ad-hoc P2P-SIP problem on the assumption that
> someday, somebody will have reason to use it (even though we can't really
> brainstorm anybody who would use it today), then I guess that's a cool
> research project.
> 
> 
> 
> But it's a *research* project, not an IETF standardization exercise.  DNS
> and mDNS work everywhere the *actual* users are, and thus I think the IETF
> should focus on them, first.
> 
> 
> 
> -david
> 
> 
> 
> 
>   _____
> 
> 
> From: Eunsoo Shim [ mailto:eunsooshim at gmail.com
> <mailto:eunsooshim at gmail.com> ]
> Sent: Tuesday, December 05, 2006 3:04 PM
> To: David Barrett
> Cc: P2P-SIP
> Subject: Re: [p2p-sip] DNS for inter-domain? Yes; intra-domain? No
> 
> 
> 
> I understand that mDNS can be used in ad-hoc networks where multicast
> works.
> Some of us, including me, have discussed about using mDNS and noticed that
> there is little to do to use mDNS or any other multicast-based discovery
> protocol.
> Those people decided not to focus on the approach because assuming
> multicast
> availability was not general and they did not want to limit P2PSIP into
> such
> network environments.
> They said such an approach (using mDNS or similar ones) could be discussed
> maybe in SIPPING.
> 
> The proposed charter is targeting to address intra-domain lookup that
> works
> even if there is no multicast support.
> 
> Using DNS for intra-domain lookup is not different from using a central
> server, which is not interesting to many of us, I believe. So transition
> from mDNS to DNS is not appealing, either.
> 
> What do you say?
> Thanks.
> 
> Eunsoo
> 
> On 12/5/06, David Barrett <dbarrett at quinthar.com
> <mailto:dbarrett at quinthar.com> > wrote:
> 
> 
> 
> 
>   _____
> 
> 
> From: Eunsoo Shim
> Subject: [p2p-sip] DNS for inter-domain? Yes; intra-domain? No
> 
> 
> 
> So my understanding is that what P2PSIP group is trying to do is designing
> mechanisms for P2P-style intra-domain location lookup where location
> records
> are distributed among peers rather than in a central location server.
> 
> ...
> 
> Am I missing any important aspect?
> 
> I agree, this is what P2P-SIP is trying to do.  I'm just not certain why,
> because mDNS can do the same thing in a way that is already standardized,
> already deployed, and already works.  Basically, for every real-world
> case,
> it seems mDNS is adequate for ad-hoc unauthenticated namespaces, and DNS
> is
> adequate for global, authenticated namespaces.  There is already a
> seamless
> interface to graduate from mDNS to DNS depending upon what is available;
> it
> already covers the whole range of use cases (minus the extreme security
> cases, which I believe remain unsolved problems by any means).
> 
> 
> 
> I don't dispute you could make a DHT bootstrapped with UPnP or whatever.
> You can build a lot of things.  I just don't see why we should bother when
> there are easier, adequate technologies available.
> 
> 
> 
> -david
> 
> 
> 
> _______________________________________________
> 
> p2p-sip mailing list
> 
> p2p-sip at cs.columbia.edu <mailto:p2p-sip at cs.columbia.edu>
> 
> https://lists.cs.columbia.edu/cucslists/listinfo/p2p-sip
> <https://lists.cs.columbia.edu/cucslists/listinfo/p2p-sip>
> 
> 
> 
> 
>   _____
> 
> 
> _______________________________________________
> p2p-sip mailing list
> p2p-sip at cs.columbia.edu <mailto:p2p-sip at cs.columbia.edu>
> https://lists.cs.columbia.edu/cucslists/listinfo/p2p-sip
> <https://lists.cs.columbia.edu/cucslists/listinfo/p2p-sip>
> 
> 
> _______________________________________________
> p2p-sip mailing list
> p2p-sip at cs.columbia.edu <mailto:p2p-sip at cs.columbia.edu>
> https://lists.cs.columbia.edu/cucslists/listinfo/p2p-sip
> <https://lists.cs.columbia.edu/cucslists/listinfo/p2p-sip>
> 
> 


From jani.hautakorpi at ericsson.com  Fri Dec  8 06:15:51 2006
From: jani.hautakorpi at ericsson.com (Jani Hautakorpi (JO/LMF))
Date: Fri, 08 Dec 2006 13:15:51 +0200
Subject: [p2p-sip] Concepts draft: Open Issue #1: Searching for services
In-Reply-To: <B784DF3B-10C8-401B-BCA4-149F7AF0ECD0@magma.ca>
References: <B784DF3B-10C8-401B-BCA4-149F7AF0ECD0@magma.ca>
Message-ID: <45794967.6090302@ericsson.com>

Hi Philip,

Philip Matthews wrote:
> Open issue #1:  Can different peers in a P2PSIP Overlay offer
> different services to the Overlay? If so, do we need the
> concept of  searching for the set of peers that offer a service?

Do you have some concrete services in mind?

My initial impression is that a search functionality would make sense 
only if the services would be such that they couldn't be identified with 
a SIP URI.


-- 
Jani


From spencer at mcsr-labs.org  Fri Dec  8 08:23:33 2006
From: spencer at mcsr-labs.org (Spencer Dawkins)
Date: Fri, 8 Dec 2006 07:23:33 -0600
Subject: [p2p-sip] Concepts draft: Open Issue #1: Searching for services
References: <B784DF3B-10C8-401B-BCA4-149F7AF0ECD0@magma.ca>
Message-ID: <0ddd01c71acc$0fc97520$6701a8c0@china.huawei.com>

I agree with this proposed resolution. It seems important that we identify 
the need for the capability, but we should not go too far down the road in 
the concepts document.

We DO need to go down the road in the protocol documents later.

Thanks,

Spencer


> Open issue #1:  Can different peers in a P2PSIP Overlay offer
> different services to the Overlay? If so, do we need the
> concept of  searching for the set of peers that offer a service?
>
>
> Proposed Resolution: Yes and Yes. Though we may require
> that all peers offer some minimal set of services, there is likely
> to be at least one service  that is not offered by every peer.
> Thus we need the concept of searching for the set of peers
> that offer a service.
>
> For now, we are proposing to leave undefined the detailed
> properties of the search. Thus we are proposing not
> to answer questions like whether the search finds all peers
> that offer the service, or whether the search returns the
> "closest" peer. Instead, we propose to add these questions to the
> design questions section of the document.
>
> Comments?
>
> - Philip 



From spencer at mcsr-labs.org  Fri Dec  8 08:26:50 2006
From: spencer at mcsr-labs.org (Spencer Dawkins)
Date: Fri, 8 Dec 2006 07:26:50 -0600
Subject: [p2p-sip] Concepts draft: Open Issue #2: Adding NATs
References: <24892751-5955-4316-A6B6-0C1CE5F29DD5@magma.ca>
Message-ID: <0de301c71acc$851bca30$6701a8c0@china.huawei.com>

Hi, Philip,

What would you see as alternative proposed resolutions?

Obviously, one alternative is dropping support for NATs in arbitrary places 
in the overlay topology, and I agree that that's the wrong alternative.

Are there other alternatives?

Thanks,

Spencer


> Open Issue #2:  Right now, the Concepts document says
> there can be NATs between peers. However, the examples
> in section 3.4 ignore the fact that it is generally not possible
> for an arbitrary node to send a message (e.g. a SIP INVITE)
> to destination node that is behind a NAT.
>
> Proposed Resolution:  The P2PSIP Overlay provides a message
> relaying service. In other words, the sending peer/client can send
> a SIP message through the Overlay in some manner that causes
> it to be delivered to the destination peer or client.
>
> Consider the example in section 3.4.2 where client C sends an
> INVITE to the user U residing on peer F. Since client C connects
> to the Overlay via peer Q, client C might pass this INVITE to peer Q
> which then passes somehow through the overlay so that it gets
> delivered to peer F.  (If client C somehow knows how to reach
> another peer R, it might choose to send the message to that peer
> for relaying instead).
>
> There are a number of ways that such a relaying service might
> be implemented. Some have suggested that the lookup might return
> a route vector to the destination. Others have suggested that the
> routing might be done as part of the search. However, in the
> next revision, we are proposing to not give details, but just say
> that such a mechanism exists.
>
> Comments?
>
> - Philip 



From spencer at mcsr-labs.org  Fri Dec  8 08:31:10 2006
From: spencer at mcsr-labs.org (Spencer Dawkins)
Date: Fri, 8 Dec 2006 07:31:10 -0600
Subject: [p2p-sip] Concepts draft: Open Issue #3: Enrollment as
	aPeer/Client Protocol operation?
References: <3E1F6E23-B2EF-43F2-B9EE-3C175384922B@magma.ca>
Message-ID: <0de901c71acd$1fdb1da0$6701a8c0@china.huawei.com>

Hi, Philip,

Sort-of agree. Agree that we need to specify the acceptable format of a 
credential.

My understanding of this proposed resolution is basically

- we don't HAVE to specify the method of obtaining credentials now, and

- if we DO standardize the method of obtaining credentials, it's not likely 
to be part of the peer or client protocols, so would be done separately in 
any event.

Do you guys agree? If so, I'd like this understanding to be included in the 
document.

Thanks,

Spencer


> Open issue #3: In version -03 of the Concepts  draft, I added some
> text that stated
> that Enrollment of peers and resources/users was an operation
> supported by the Peer and Client Protocols.
> However, a couple of weeks ago, Brian Rosen and Henry Sinnreich
> suggested that
> we not standardize the protocol for obtaining credentials, but simply
> specify the acceptable
> format of a credential.
>
> Proposed Resolution: The authors propose to adopt Brian and Henry's
> view of the world.
> Thus we propose to remove enrollment of peers and resources/users as
> operations supported
> by the Peer and Client protocols, and state that the method of
> obtaining credentials is
> not being standardized for now.
>
> Comments?
>
> - Philip
>
>
>
> _______________________________________________
> p2p-sip mailing list
> p2p-sip at cs.columbia.edu
> https://lists.cs.columbia.edu/cucslists/listinfo/p2p-sip
> 



From philip_matthews at magma.ca  Fri Dec  8 09:56:15 2006
From: philip_matthews at magma.ca (Philip Matthews)
Date: Fri, 8 Dec 2006 09:56:15 -0500
Subject: [p2p-sip] Concepts draft: Open Issue #1: Searching for services
In-Reply-To: <000001c71a9d$1a0ac6d0$7205a40a@china.huawei.com>
References: <000001c71a9d$1a0ac6d0$7205a40a@china.huawei.com>
Message-ID: <8F4FF595-DEAE-413E-B5F7-7486C5555B26@magma.ca>

See inline.

On 8-Dec-06, at 02:47 , JiangXingFeng wrote:
[snip]

>> Proposed Resolution: Yes and Yes. Though we may require
>> that all peers offer some minimal set of services, there is likely
>> to be at least one service  that is not offered by every peer.
>> Thus we need the concept of searching for the set of peers
>> that offer a service.
> For the first issue, I agree with the proposed resolution. Peers in  
> the
> overlay have different capabilities other than basic storage and  
> routing
> service. These capabilities could be used to make some applications  
> easier
> to implement.
> For the second one, I think if we think of different service as  
> resource
> like SIP URI and put these resources into overlay, it is easy to  
> get these
> resource by common search mechanism of DHT.
> A problem following the above open issues is: if there are several  
> peers
> providing the same service, how does the system choose the  
> appropriate one
> or more to serve other peers? Maybe the criteria are net-path  
> closest away
> from service requesting peers, or performance of service providing  
> peer? I
> think the criteria are dependent on the specific application. Is this
> problem in the scope of P2PSIP or out of scope?

I agree that the criterion of selecting a peer that is net-path  
closest would
be good for some services. Of course, for other services, different  
criteria might
be more appropriate (e.g., for a voicemail service, voicemail storage  
capacity
might be more important).

Ultimately, we might well want very sophisticated search schemes.  
However,
some have suggested that we stick with a simple scheme for now, in  
the same
way that we are sticking with a simple lookup scheme finding users.

Since the goal of the Concept draft is to specify a basic framework  
and give some terminology
to help the group discuss ideas, it seems best to us (the authors) if  
the draft
is as general as possible about the search mechanism, at least until  
the group
starts trying to narrow down the search mechanism details.

- Philip

From philip_matthews at magma.ca  Fri Dec  8 10:22:24 2006
From: philip_matthews at magma.ca (Philip Matthews)
Date: Fri, 8 Dec 2006 10:22:24 -0500
Subject: [p2p-sip] Concepts draft: Open Issue #1: Searching for services
In-Reply-To: <45794967.6090302@ericsson.com>
References: <B784DF3B-10C8-401B-BCA4-149F7AF0ECD0@magma.ca>
	<45794967.6090302@ericsson.com>
Message-ID: <AE316798-72CA-4C38-913C-389E398A24EB@magma.ca>


On 8-Dec-06, at 06:15 , Jani Hautakorpi (JO/LMF) wrote:

> Hi Philip,
>
> Philip Matthews wrote:
>> Open issue #1:  Can different peers in a P2PSIP Overlay offer
>> different services to the Overlay? If so, do we need the
>> concept of  searching for the set of peers that offer a service?
>
> Do you have some concrete services in mind?
>
> My initial impression is that a search functionality would make  
> sense only if the services would be such that they couldn't be  
> identified with a SIP URI.

Hi Jani:

Here are my opinions of what such services might be (my co-authors  
may not agree! ;-).

One service might be STUN and TURN server service. Some peers might  
offer both STUN and TURN, some might offer just STUN, and some might  
offer neither.
Other services might be SIP Proxy server service, or SIP Redirect  
server service, etc.  Other services might be the willingness to host  
clients.

There has been some debate over which services all peers must offer,  
and which peers can optionally offer.  Rather than continue that  
debate right now, the authors are simply saying:
              "there will be at least one service that not all peers  
will offer, and thus a search mechanism is needed".

- Philip

From bryan at ethernot.org  Fri Dec  8 10:45:19 2006
From: bryan at ethernot.org (David A. Bryan)
Date: Fri, 8 Dec 2006 10:45:19 -0500
Subject: [p2p-sip] Concepts draft: Open Issue #3: Enrollment as
	aPeer/Client Protocol operation?
In-Reply-To: <0de901c71acd$1fdb1da0$6701a8c0@china.huawei.com>
References: <3E1F6E23-B2EF-43F2-B9EE-3C175384922B@magma.ca>
	<0de901c71acd$1fdb1da0$6701a8c0@china.huawei.com>
Message-ID: <8b2769930612080745m4936bb32mb1538b83dab2b49b@mail.gmail.com>

I think the idea was just to reduce the work, and not take on
something that could be much broader than P2PSIP only use. Both of
Spencer's statements below are true, and I'll add one more:

- if we DO standardize, it could be this is useful to others outside
P2PSIP and might even be some work we push into another WG.

David


On 12/8/06, Spencer Dawkins <spencer at mcsr-labs.org> wrote:
> Hi, Philip,
>
> Sort-of agree. Agree that we need to specify the acceptable format of a
> credential.
>
> My understanding of this proposed resolution is basically
>
> - we don't HAVE to specify the method of obtaining credentials now, and
>
> - if we DO standardize the method of obtaining credentials, it's not likely
> to be part of the peer or client protocols, so would be done separately in
> any event.
>
> Do you guys agree? If so, I'd like this understanding to be included in the
> document.
>
> Thanks,
>
> Spencer
>
>
> > Open issue #3: In version -03 of the Concepts  draft, I added some
> > text that stated
> > that Enrollment of peers and resources/users was an operation
> > supported by the Peer and Client Protocols.
> > However, a couple of weeks ago, Brian Rosen and Henry Sinnreich
> > suggested that
> > we not standardize the protocol for obtaining credentials, but simply
> > specify the acceptable
> > format of a credential.
> >
> > Proposed Resolution: The authors propose to adopt Brian and Henry's
> > view of the world.
> > Thus we propose to remove enrollment of peers and resources/users as
> > operations supported
> > by the Peer and Client protocols, and state that the method of
> > obtaining credentials is
> > not being standardized for now.
> >
> > Comments?
> >
> > - Philip
> >
> >
> >
> > _______________________________________________
> > p2p-sip mailing list
> > p2p-sip at cs.columbia.edu
> > https://lists.cs.columbia.edu/cucslists/listinfo/p2p-sip
> >
>
>
> _______________________________________________
> p2p-sip mailing list
> p2p-sip at cs.columbia.edu
> https://lists.cs.columbia.edu/cucslists/listinfo/p2p-sip
>

From bryan at ethernot.org  Fri Dec  8 10:50:43 2006
From: bryan at ethernot.org (David A. Bryan)
Date: Fri, 8 Dec 2006 10:50:43 -0500
Subject: [p2p-sip] Concepts draft: Open Issue #1: Searching for services
In-Reply-To: <AE316798-72CA-4C38-913C-389E398A24EB@magma.ca>
References: <B784DF3B-10C8-401B-BCA4-149F7AF0ECD0@magma.ca>
	<45794967.6090302@ericsson.com>
	<AE316798-72CA-4C38-913C-389E398A24EB@magma.ca>
Message-ID: <8b2769930612080750h79f512a1x46976b504110f245@mail.gmail.com>

I think I had in mind something like what Jiang had in mind -- for
now, with a simple system, things are stored by URIs, and the method
for choosing between is out of scope. I soould like a broken record
today, but I think again that a protocol for selecting the best relay
from a list is something with broader applicability than P2PSIP --
might be able to reuse something or if we have to make one, push it to
a group that can make a more general one.

And although I missed this earlier...Philip: can we use a word other
than search, since it is simple lookup we are really talking about
here? Otherwise wording looks great to me.

David

On 12/8/06, Philip Matthews <philip_matthews at magma.ca> wrote:
>
> On 8-Dec-06, at 06:15 , Jani Hautakorpi (JO/LMF) wrote:
>
> > Hi Philip,
> >
> > Philip Matthews wrote:
> >> Open issue #1:  Can different peers in a P2PSIP Overlay offer
> >> different services to the Overlay? If so, do we need the
> >> concept of  searching for the set of peers that offer a service?
> >
> > Do you have some concrete services in mind?
> >
> > My initial impression is that a search functionality would make
> > sense only if the services would be such that they couldn't be
> > identified with a SIP URI.
>
> Hi Jani:
>
> Here are my opinions of what such services might be (my co-authors
> may not agree! ;-).
>
> One service might be STUN and TURN server service. Some peers might
> offer both STUN and TURN, some might offer just STUN, and some might
> offer neither.
> Other services might be SIP Proxy server service, or SIP Redirect
> server service, etc.  Other services might be the willingness to host
> clients.
>
> There has been some debate over which services all peers must offer,
> and which peers can optionally offer.  Rather than continue that
> debate right now, the authors are simply saying:
>               "there will be at least one service that not all peers
> will offer, and thus a search mechanism is needed".
>
> - Philip
>

From mlepinski at bbn.com  Fri Dec  8 11:01:39 2006
From: mlepinski at bbn.com (Matt Lepinski)
Date: Fri, 08 Dec 2006 11:01:39 -0500
Subject: [p2p-sip] Concepts draft: Open Issue #1: Searching for services
In-Reply-To: <AE316798-72CA-4C38-913C-389E398A24EB@magma.ca>
References: <B784DF3B-10C8-401B-BCA4-149F7AF0ECD0@magma.ca>	<45794967.6090302@ericsson.com>
	<AE316798-72CA-4C38-913C-389E398A24EB@magma.ca>
Message-ID: <45798C63.7040209@bbn.com>

Philip,

Thanks a lot for the clarification. In light of your examples, the 
service search issue makes a lot more sense.

However, from the text of the -03 version of the concepts draft, it is 
unclear to me what things are "services" (as opposed to say, 
"resources"). Are you planning to add a definition of a "service" to 
section 2 of the draft?

- Matt Lepinski

Philip Matthews wrote:

>On 8-Dec-06, at 06:15 , Jani Hautakorpi (JO/LMF) wrote:
>
>  
>
>>Hi Philip,
>>
>>Philip Matthews wrote:
>>    
>>
>>>Open issue #1:  Can different peers in a P2PSIP Overlay offer
>>>different services to the Overlay? If so, do we need the
>>>concept of  searching for the set of peers that offer a service?
>>>      
>>>
>>Do you have some concrete services in mind?
>>
>>My initial impression is that a search functionality would make  
>>sense only if the services would be such that they couldn't be  
>>identified with a SIP URI.
>>    
>>
>
>Hi Jani:
>
>Here are my opinions of what such services might be (my co-authors  
>may not agree! ;-).
>
>One service might be STUN and TURN server service. Some peers might  
>offer both STUN and TURN, some might offer just STUN, and some might  
>offer neither.
>Other services might be SIP Proxy server service, or SIP Redirect  
>server service, etc.  Other services might be the willingness to host  
>clients.
>
>There has been some debate over which services all peers must offer,  
>and which peers can optionally offer.  Rather than continue that  
>debate right now, the authors are simply saying:
>              "there will be at least one service that not all peers  
>will offer, and thus a search mechanism is needed".
>
>- Philip
>_______________________________________________
>p2p-sip mailing list
>p2p-sip at cs.columbia.edu
>https://lists.cs.columbia.edu/cucslists/listinfo/p2p-sip
>
>  
>



From enrico.marocco at telecomitalia.it  Fri Dec  8 11:15:51 2006
From: enrico.marocco at telecomitalia.it (Enrico Marocco)
Date: Fri, 08 Dec 2006 17:15:51 +0100
Subject: [p2p-sip] Concepts draft: Open Issue #1: Searching for services
In-Reply-To: <000001c71a9d$1a0ac6d0$7205a40a@china.huawei.com>
References: <000001c71a9d$1a0ac6d0$7205a40a@china.huawei.com>
Message-ID: <45798FB7.3010105@telecomitalia.it>

JiangXingFeng wrote:
>> Open issue #1:  Can different peers in a P2PSIP Overlay offer
>> different services to the Overlay? If so, do we need the
>> concept of  searching for the set of peers that offer a service?
>>
>>
> 
>> Proposed Resolution: Yes and Yes. Though we may require
>> that all peers offer some minimal set of services, there is likely
>> to be at least one service  that is not offered by every peer.
>> Thus we need the concept of searching for the set of peers
>> that offer a service.
> For the first issue, I agree with the proposed resolution. Peers in the
> overlay have different capabilities other than basic storage and routing
> service. These capabilities could be used to make some applications easier
> to implement.
> For the second one, I think if we think of different service as resource
> like SIP URI and put these resources into overlay, it is easy to get these
> resource by common search mechanism of DHT. 

We must be very careful in taking such a decision, because, with such a
naive approach, we would store all the istances of the same service
(apart from replicas) on the same peer and cause dangerous displacements
to the DHT traffic.  For example, the peer authoritative for the URL
representing the STUN service, would receive many more queries than
almost any other peer.

> A problem following the above open issues is: if there are several peers
> providing the same service, how does the system choose the appropriate one
> or more to serve other peers? Maybe the criteria are net-path closest away
> from service requesting peers, or performance of service providing peer? I
> think the criteria are dependent on the specific application. Is this
> problem in the scope of P2PSIP or out of scope? 

This is a very general problem, probably out of scope for p2psip.

>> For now, we are proposing to leave undefined the detailed
>> properties of the search. Thus we are proposing not
>> to answer questions like whether the search finds all peers
>> that offer the service, or whether the search returns the
>> "closest" peer. Instead, we propose to add these questions to the
>> design questions section of the document.

I agree with Philip; at the moment we don't have enough ground to give a
solid definition for the generic "search" operation.

-- 
Ciao,
Enrico

From philip_matthews at magma.ca  Fri Dec  8 11:18:20 2006
From: philip_matthews at magma.ca (Philip Matthews)
Date: Fri, 8 Dec 2006 11:18:20 -0500
Subject: [p2p-sip] Concepts draft: Open Issue #2: Adding NATs
In-Reply-To: <0de301c71acc$851bca30$6701a8c0@china.huawei.com>
References: <24892751-5955-4316-A6B6-0C1CE5F29DD5@magma.ca>
	<0de301c71acc$851bca30$6701a8c0@china.huawei.com>
Message-ID: <7ED352B8-53CB-4859-8B03-DF7644B44C9F@magma.ca>


On 8-Dec-06, at 08:26 , Spencer Dawkins wrote:

> Hi, Philip,
>
> What would you see as alternative proposed resolutions?
>
> Obviously, one alternative is dropping support for NATs in  
> arbitrary places
> in the overlay topology, and I agree that that's the wrong  
> alternative.
>
> Are there other alternatives?

One alternative, of course, is to say that all peer must reside in  
the same address realm
(for example, all peers must have public IP addresses). At this  
point, this would
be a charter change.

The other alternatives that I can think of are impractical or  
undesirable.
They include such things as: (a) requiring all peers behind NATs to  
maintain connections
to all other peers and clients, or (b) requiring a peer behind  a NAT  
(of any kind) to have
an associated TURN server, but a TURN server with endpoint- 
independent filtering
(something deliberately ruled out by the current  TURN spec).

Note: If you assume that there is a reasonable number of peers in the  
same address realm X
(e.g., the public Internet), then we could require that each peer  
behind a NAT maintain
a connection to a peer in realm X. Then a message from peer A to peer  
D is sent to peer C
in realm  X, to which peer D is connected to. The phrase "routing  
through the overlay" is
intended to be general enough to include this approach.


- Philip

PS. Personally, I don't think that all P2PSIP networks will have such  
an address realm X available,
which is  what I think is a problem for the DNS-based approach that  
has been discussed
so much in the past two weeks.


From philip_matthews at magma.ca  Fri Dec  8 11:29:11 2006
From: philip_matthews at magma.ca (Philip Matthews)
Date: Fri, 8 Dec 2006 11:29:11 -0500
Subject: [p2p-sip] Concepts draft: Open Issue #3: Enrollment as
	aPeer/Client Protocol operation?
In-Reply-To: <0de901c71acd$1fdb1da0$6701a8c0@china.huawei.com>
References: <3E1F6E23-B2EF-43F2-B9EE-3C175384922B@magma.ca>
	<0de901c71acd$1fdb1da0$6701a8c0@china.huawei.com>
Message-ID: <61D7DDBB-54C1-46CB-A74F-D3DEA084059E@magma.ca>

On 8-Dec-06, at 08:31 , Spencer Dawkins wrote:

> Hi, Philip,
>
> Sort-of agree. Agree that we need to specify the acceptable format  
> of a
> credential.
>
> My understanding of this proposed resolution is basically
>
> - we don't HAVE to specify the method of obtaining credentials now,  
> and
>
> - if we DO standardize the method of obtaining credentials, it's  
> not likely
> to be part of the peer or client protocols, so would be done  
> separately in
> any event.
>
> Do you guys agree? If so, I'd like this understanding to be  
> included in the
> document.
>
> Thanks,
>
> Spencer

Personally, I don't have a strong opinion on this, and I don't think  
any of my
co-authors do either. We were simply influenced by the fact that (a)  
some people
in the group DO seem to have a strong opinion, and (b) the proposed  
resolution
reduces the scope of the initial work of the group, which focuses the  
group's efforts
more tightly.

I mostly agree with what you said above. However, is it really true  
that we would
not ever want to add this capability to the peer or client protocols?

- Philip

From philip_matthews at magma.ca  Fri Dec  8 11:42:50 2006
From: philip_matthews at magma.ca (Philip Matthews)
Date: Fri, 8 Dec 2006 11:42:50 -0500
Subject: [p2p-sip] Concepts draft: Open Issue #1: Searching for services
In-Reply-To: <8b2769930612080750h79f512a1x46976b504110f245@mail.gmail.com>
References: <B784DF3B-10C8-401B-BCA4-149F7AF0ECD0@magma.ca>
	<45794967.6090302@ericsson.com>
	<AE316798-72CA-4C38-913C-389E398A24EB@magma.ca>
	<8b2769930612080750h79f512a1x46976b504110f245@mail.gmail.com>
Message-ID: <A72226BC-FB7F-43EF-8782-2189AA879BF0@magma.ca>


On 8-Dec-06, at 10:50 , David A. Bryan wrote:

> I think I had in mind something like what Jiang had in mind -- for
> now, with a simple system, things are stored by URIs, and the method
> for choosing between is out of scope. I soould like a broken record
> today, but I think again that a protocol for selecting the best relay
> from a list is something with broader applicability than P2PSIP --
> might be able to reuse something or if we have to make one, push it to
> a group that can make a more general one.
>
> And although I missed this earlier...Philip: can we use a word other
> than search, since it is simple lookup we are really talking about
> here? Otherwise wording looks great to me.
>
> David

How about we say:
"a lookup of a service returns at least one peer that supports that  
service".

First, this changes "search" to "lookup".

Second, it doesn't take a position on how many peers are returned.
If 5 peers support the service, then the lookup may return just one  
peer,
or it may return all 5. (Of course, when we define the mechanism, we may
need to answer this question, but for now we leave this question as  
open).

- Philip

From RADHIKA.R.ROY at saic.com  Fri Dec  8 11:43:49 2006
From: RADHIKA.R.ROY at saic.com (Roy, Radhika R.)
Date: Fri, 8 Dec 2006 11:43:49 -0500 
Subject: [p2p-sip] DNS for inter-domain? Yes; intra-domain? No
Message-ID: <62D92A9A02BCC845B202323D49A48426E1CF58@0591-ITS-EXMP02.us.saic.com>

To some extent just because it uses the multicast property, but is not good
enough in mobile ad hoc network.
 
The issue is similar the way OSPFv3 was not good enough for mobile ad hoc
environment where P2P properties are important (here is something missing:
peering in ad hoc environment and its impact on DNS or mDNS). Also, it is
for SIP: audio/video and other features are also required.
 
Let us stop here.
 
We will revisit this issue later.
 
Best regards,
Radhika

  _____  

From: David Barrett [mailto:dbarrett at quinthar.com]
Sent: Fri 12/8/2006 3:37 AM
To: 'Roy, Radhika R.'; 'Eunsoo Shim'
Cc: 'P2P-SIP'
Subject: RE: [p2p-sip] DNS for inter-domain? Yes; intra-domain? No



I'm not sure I follow you here.

The "m" in "mDNS" stands for multicast because you send your DNS requests to
a multicast channel instead of to a fixed DNS server.  Anyone who knows the
answer broadcasts a reply.  It's designed precisely for an ad hoc network
with no fixed nodes.

Does this address your concern?

-david

> -----Original Message-----
> From: Roy, Radhika R. [mailto:RADHIKA.R.ROY at saic.com
<mailto:RADHIKA.R.ROY at saic.com> ]
> Sent: Thursday, December 07, 2006 8:51 PM
> To: David Barrett; Eunsoo Shim
> Cc: P2P-SIP
> Subject: RE: [p2p-sip] DNS for inter-domain? Yes; intra-domain? No
>
> Issue is this: Mobile ad hoc network - no fixed infrastructure where
> everything is P2P. No one knows who will be the next peer after a certain
> period of time. The speed of each peer may vary.
>
> DNS is used assuming that there is at least one fixed non-mobile node. It
> is
> the problem that we want to get rid of.
>
> First, let us understand what works in this environment.
>
> IETF MANET WG has all the drafts and RFCs.
>
> Best regards,
> Radhika
>
> PS:
>
> 1. It is not about DNS or mDNS or multicasting per se.
> 2. Let us find our what works in the above environment, i.e., MANET.
> Everything is distributive as P2P in mobile ad hoc environment. Period.
> 3. The routing protocol for MANET that will be used in the about
> environment
> best commercially (to be accepted as standard soon as all alternative
> protocols have provided their performance and simulations results, but
> this
> is expected to be chosen as s finalist - draft is available in the IETF
> MANET WG Archive): OSPFv3-MDR.
> 4. We may have to enhance the charter for the P2P-SIP in the future as the
> present scope of this would be WG is not enough. However, it is a
> different
> story as a first step let us accomplish what is in the present charter.
>
>
>
>
>   _____
>
> From: David Barrett [mailto:dbarrett at quinthar.com
<mailto:dbarrett at quinthar.com> ]
> Sent: Thu 12/7/2006 8:23 PM
> To: 'Roy, Radhika R.'; 'Eunsoo Shim'
> Cc: 'P2P-SIP'
> Subject: RE: [p2p-sip] DNS for inter-domain? Yes; intra-domain? No
>
>
>
> I'm not very familiar with MANET, but I think it supports multicast, so
> mDNS
> could probably play a role in it.
>
>
>
> Either way, if someone goes to the trouble of setting up some kind of
> grid/mesh network, they'll probably want to do more than just VoIP.
> They'll
> need *some* kind of DNS to use for everything else.  It's probable they'll
> find a way to tunnel normal DNS.  It's possible they'll find a way to use
> normal mDNS.  Before inventing a custom, non-DNS solution, it seems
> sensible
> to see how they're going to solve the DNS problem first and put our weight
> behind that, rather than inventing a new, competing solution.
>
>
>
> -david
>
>
>
> PS: What is inadequate about DNS and mDNS for your network today?
>
>
>
>   _____
>
> From: Roy, Radhika R. [mailto:RADHIKA.R.ROY at saic.com
<mailto:RADHIKA.R.ROY at saic.com> ]
> Sent: Thursday, December 07, 2006 2:45 PM
> To: David Barrett; Eunsoo Shim
> Cc: P2P-SIP
> Subject: RE: [p2p-sip] DNS for inter-domain? Yes; intra-domain? No
>
>
>
> David:
>
>
>
> Do DNS and mDNS also cover P2P Mobile Ad Hoc Networks (MANET)?
>
>
>
> We are looking for MANET support right now for P2P-SIP networking.
>
>
>
> Best regards,
>
> Radhika
>
>
>
> PS: We are using DNS right now, but no insufficient.
>
>
>
>   _____
>
> From: p2p-sip-bounces at cs.columbia.edu on behalf of David Barrett
> Sent: Thu 12/7/2006 4:37 PM
> To: 'Eunsoo Shim'
> Cc: 'P2P-SIP'
> Subject: Re: [p2p-sip] DNS for inter-domain? Yes; intra-domain? No
>
> Hm, I will agree that the conversation has covered all the useful ground
> it's going to cover.  As for the confusions you raise:
>
>
>
> -          DNS is used when available (internet or private network), and
> mDNS otherwise (ad hoc subnet)
>
> -          Multicast only works on subnets.  Ad hoc networks are typically
> single-subnets.  Thus mDNS works on ad hoc networks.
>
> -          Multicast does not work between subnets.  The internet and
> private networks use multiple subnets.  Thus mDNS will not work across the
> internet or on private networks spanning multiple subnets.  However, DNS
> is
> typically available in this case, and thus DNS is used instead.
>
> -          DNS and mDNS cover the total range of real-world networking
> environments.
>
> -          I understand P2P-SIP is currently proposed to use DNS as an
> optional extension to the mandatory DHT.  I think it'd be more sensible to
> have mandatory support for DNS with optional DHT, particularly because the
> DHT is hard work for no real-world gain and it seems wasteful to force
> everyone to pay it.
>
> -          The "P2P" system I propose as a replacement for DHT is merely
> what Adam originally proposed: configure P2P-SIP UAs to update a DNS
> record
> that points to another P2P-SIP UA as a proxy.
>
> -          Whether or not that is P2P "enough" is a long boring religious
> discussion.  But so far as we've been able to determine, this solves all
> the
> same use cases as the DHT, but with less work and better integration with
> existing SIP.
>
> -          P2P works; it's what pays my bills.  But it's a question of
> when
> and where is appropriate.
>
> Please let me know if you have any more questions about this idea or need
> me
> to clarify it more, but I think we've covered it exhaustively.
>
>
>
> -david
>
>
>
>   _____
>
> From: Eunsoo Shim [mailto:eunsooshim at gmail.com
<mailto:eunsooshim at gmail.com> ]
> Sent: Thursday, December 07, 2006 4:12 AM
> To: David Barrett
> Cc: Peter Pan; Henning Schulzrinne; P2P-SIP
> Subject: Re: [p2p-sip] DNS for inter-domain? Yes; intra-domain? No
>
>
>
> mDNS works only in a subnet. If you limit P2PSIP applications only to that
> environment, mDNS should work. You can go to SIPPING to talk about it.
> Many
> of us understand it.
> If you suggest mDNS even for other environments, please have the Internet
> changed to support multicast almost everywhere first. Then we can talk
> about
> using mDNS for P2PSIP in a general sense.
>
> dynDNS is just another DNS system, i.e. relying on DNS SERVERs. Using
> dynDNS
> is same as retrieving a redirect server information from DNS and getting
> the
> location of a particular user from the redirect server. That means, even
> we
> don't have to update the DNS infrastructure! Just use existing DNS and SIP
> and you can get what you propose.
>
> You continuously try to imply that P2PSIP does not use DNS at all. I
> pointed
> out that DNS was assumed for inter-domain connections for P2PSIP (and
> actually SIP). All left from your argument is that we got to use DNS
> servers
> even for intra-domain lookups, which is equivalent to using a central
> location server.
>
> Now it sounds to me that you are saying "p2p does not work. Stick to
> central
> servers!"
> I have no interest in arguing whether C/S architecture or P2P architecture
> is better. Neither are many people in this list, I believe. We understand
> there are tradeoffs. Maybe you can enjoy that kind of discussion with
> people
> from Skype or eDonkey.
>
> I think we passed already time for that kind of discussion here.
>
> Already the current SIP systems use central servers for location
> registration and voicemail storage.
> If you are interested in another way/protocol of using central servers for
> SIP, this is not a right place to propose it.
> If you have a problem with DHT, please propose a different 'P2P' mechanism
> for 'P2PSIP'.
>
>
> Thanks.
>
> Eunsoo
>
>
>
>
> On 12/6/06, David Barrett <dbarrett at quinthar.com
> <mailto:dbarrett at quinthar.com <mailto:dbarrett at quinthar.com> > > wrote:
>
> I'm not sure the DNS infrastructure needs much updating at all.  Perhaps
> there will be more demand for DNS registrars to offer automated (aka
> "dynamic") record updating, but that's really just standardizing what they
> already do.
>
>
>
> As for voicemail, no, I wouldn't expect to put that into DNS; that's
> something best handled by some real storage provider.  You *could* do it
> in
> a DHT, but it'd just be slower and less reliable, and the engineering cost
> of building that would pay for central storage many, many times over.
>
>
>
> Again, in a vacuum without taking anything else into consideration, be it
> engineering cost, performance, reliability, existing systems, standards
> support and so on, yes, a DHT could be used.  But when everything else is
> factored in, it just makes no sense at all.
>
>
>
> -david
>
>
>
>   _____
>
> From: Peter Pan [mailto:huang-ming.pan at comcast.net
<mailto:huang-ming.pan at comcast.net> 
> <mailto:huang-ming.pan at comcast.net <mailto:huang-ming.pan at comcast.net> > ]
> Sent: Wednesday, December 06, 2006 12:09 PM
> To: David Barrett; 'Henning Schulzrinne'
> Cc: 'P2P-SIP'
>
>
> Subject: Re: [p2p-sip] DNS for inter-domain? Yes; intra-domain? No
>
>
>
> Great summary. Please see my feedback inline.
>
>
>
> Peter
>
> ----- Original Message -----
>
> From: David Barrett <mailto:dbarrett at quinthar.com
<mailto:dbarrett at quinthar.com> >
>
> To: 'Henning Schulzrinne' <mailto:hgs at cs.columbia.edu
<mailto:hgs at cs.columbia.edu> >
>
> Cc: 'P2P-SIP' <mailto:p2p-sip at cs.columbia.edu
<mailto:p2p-sip at cs.columbia.edu> >
>
> Sent: Wednesday, December 06, 2006 12:07 AM
>
> Subject: Re: [p2p-sip] DNS for inter-domain? Yes; intra-domain? No
>
>
>
> I think the value of this discussion is in identifying and clarifying
> misimpressions about how DNS and mDNS work (and do not work) so we can
> make
> a more informed choice on whether or not to use them.  Already we've
> clarified:
>
>
>
> How can DNS records be automatically updated? (Through dynamic DNS
> protocols)
>
> - [pp]: what is the cost and feasibility to upgrade existing DNS
> infrastructure to support dynamic DNS globally? is it similiar to the
> deployment of ipv6 and ip mobility?
>
> - How fast can DNS propagate?  (Extremely fast with short TTLs)
>
> - What if I have multiple SIP proxies? (Put multiple IPs into the DNS
> record)
>
> Is DNS only for IP addresses? (No, can store other records)
>
> - [pp]: can/will these DNS server also store data like voicemail or other
> p2p application data?
>
> - How does DNS work in on ad-hoc networks? (Using mDNS via multicast)
>
> - Where is multicast supported? (Only everywhere it's needed - on subnets)
>
> - Is mDNS an IETF standard? (It's on the standards track.)
>
> - Is mDNS actually in use? (It's the foundation of zeroconf, which is used
> extensively by Apple)
>
> - Are there any use-cases that DHT supports that DNS/mDNS cannot?
> (Apparently not.)
>
>
>
> Lack of clarity on these points could call for the unnecessary dismissal
> of
> these technologies on false premises, which I think nobody wants.
>
>
>
> -david
>
>
>
>
>   _____
>
>
> From: Henning Schulzrinne [mailto:hgs at cs.columbia.edu
<mailto:hgs at cs.columbia.edu> 
> <mailto:hgs at cs.columbia.edu <mailto:hgs at cs.columbia.edu> > ]
> Sent: Tuesday, December 05, 2006 4:20 PM
> To: David Barrett
> Cc: 'Eunsoo Shim'; 'P2P-SIP'
> Subject: Re: [p2p-sip] DNS for inter-domain? Yes; intra-domain? No
>
>
>
> I have to admit that I'm confused by this whole discussion. I thought we
> had
> agreement that
>
>
>
> (1) specifying mDNS + SIP operation is useful (and hopefully simple), but
> easily handled within the framework of SIPPING;
>
>
>
> (2) specifying dynamic DNS + SIP might be useful, subject to working out
> issues such as proxy discovery, authentication and the like and that
> interested parties should work out, at least to the extent of identifying
> the pieces that might need work (as a random example: can I expect my SIP
> Digest secret to work as a DNS update secret?). If, as claimed, this is
> easy
> to do, SIPPING is again the natural home for such a BCP effort, with the
> DNS-related WGs as the home for any DNS extensions that might prove
> necessary.
>
>
>
> Maybe those individuals interested in solving (2) should form an informal
> design team and get beyond the hand-waving stage and identify the
> problems,
> if any, that would have to be addressed in either protocol or operational
> standards. I have unfortunately seen little such effort so far, although I
> suspect that such level of detail isn't terribly well suited to be hashed
> out in front of several hundred mailing list members.
>
>
>
>
>
> We also have had 2 BOFs and numerous ad-hoc meetings on this topic, with a
> rough consensus on the current charter. Are you trying to keep people from
> working on a specific topic you consider inferior?
>
>
>
> We seem to be rehashing the same arguments again and again.
>
>
>
> Henning
>
>
>
>
>
>
>
> On Dec 5, 2006, at 7:02 PM, David Barrett wrote:
>
>
>
> Does the actual, real-world availability of multicast factor into the
> decision as to whether or not to use it?
>
>
>
> I mean, we *could* assume that every odd IP address won't work, and then
> design around that.  But odd IP addresses *do* work.
>
>
>
> Likewise, multicast *is* (or could easily be made) available for all of
> the
> real-world use cases that demand it.
>
>
>
> It just seems like there are much more interesting *actual* problems to
> solve than those that don't affect real users.  But if you're excited
> about
> solving the non-multicast ad-hoc P2P-SIP problem on the assumption that
> someday, somebody will have reason to use it (even though we can't really
> brainstorm anybody who would use it today), then I guess that's a cool
> research project.
>
>
>
> But it's a *research* project, not an IETF standardization exercise.  DNS
> and mDNS work everywhere the *actual* users are, and thus I think the IETF
> should focus on them, first.
>
>
>
> -david
>
>
>
>
>   _____
>
>
> From: Eunsoo Shim [ mailto:eunsooshim at gmail.com
<mailto:eunsooshim at gmail.com> 
> <mailto:eunsooshim at gmail.com <mailto:eunsooshim at gmail.com> > ]
> Sent: Tuesday, December 05, 2006 3:04 PM
> To: David Barrett
> Cc: P2P-SIP
> Subject: Re: [p2p-sip] DNS for inter-domain? Yes; intra-domain? No
>
>
>
> I understand that mDNS can be used in ad-hoc networks where multicast
> works.
> Some of us, including me, have discussed about using mDNS and noticed that
> there is little to do to use mDNS or any other multicast-based discovery
> protocol.
> Those people decided not to focus on the approach because assuming
> multicast
> availability was not general and they did not want to limit P2PSIP into
> such
> network environments.
> They said such an approach (using mDNS or similar ones) could be discussed
> maybe in SIPPING.
>
> The proposed charter is targeting to address intra-domain lookup that
> works
> even if there is no multicast support.
>
> Using DNS for intra-domain lookup is not different from using a central
> server, which is not interesting to many of us, I believe. So transition
> from mDNS to DNS is not appealing, either.
>
> What do you say?
> Thanks.
>
> Eunsoo
>
> On 12/5/06, David Barrett <dbarrett at quinthar.com
> <mailto:dbarrett at quinthar.com <mailto:dbarrett at quinthar.com> > > wrote:
>
>
>
>
>   _____
>
>
> From: Eunsoo Shim
> Subject: [p2p-sip] DNS for inter-domain? Yes; intra-domain? No
>
>
>
> So my understanding is that what P2PSIP group is trying to do is designing
> mechanisms for P2P-style intra-domain location lookup where location
> records
> are distributed among peers rather than in a central location server.
>
> ...
>
> Am I missing any important aspect?
>
> I agree, this is what P2P-SIP is trying to do.  I'm just not certain why,
> because mDNS can do the same thing in a way that is already standardized,
> already deployed, and already works.  Basically, for every real-world
> case,
> it seems mDNS is adequate for ad-hoc unauthenticated namespaces, and DNS
> is
> adequate for global, authenticated namespaces.  There is already a
> seamless
> interface to graduate from mDNS to DNS depending upon what is available;
> it
> already covers the whole range of use cases (minus the extreme security
> cases, which I believe remain unsolved problems by any means).
>
>
>
> I don't dispute you could make a DHT bootstrapped with UPnP or whatever.
> You can build a lot of things.  I just don't see why we should bother when
> there are easier, adequate technologies available.
>
>
>
> -david
>
>
>
> _______________________________________________
>
> p2p-sip mailing list
>
> p2p-sip at cs.columbia.edu <mailto:p2p-sip at cs.columbia.edu
<mailto:p2p-sip at cs.columbia.edu> >
>
> https://lists.cs.columbia.edu/cucslists/listinfo/p2p-sip
<https://lists.cs.columbia.edu/cucslists/listinfo/p2p-sip> 
> <https://lists.cs.columbia.edu/cucslists/listinfo/p2p-sip
<https://lists.cs.columbia.edu/cucslists/listinfo/p2p-sip> >
>
>
>
>
>   _____
>
>
> _______________________________________________
> p2p-sip mailing list
> p2p-sip at cs.columbia.edu <mailto:p2p-sip at cs.columbia.edu
<mailto:p2p-sip at cs.columbia.edu> >
> https://lists.cs.columbia.edu/cucslists/listinfo/p2p-sip
<https://lists.cs.columbia.edu/cucslists/listinfo/p2p-sip> 
> <https://lists.cs.columbia.edu/cucslists/listinfo/p2p-sip
<https://lists.cs.columbia.edu/cucslists/listinfo/p2p-sip> >
>
>
> _______________________________________________
> p2p-sip mailing list
> p2p-sip at cs.columbia.edu <mailto:p2p-sip at cs.columbia.edu
<mailto:p2p-sip at cs.columbia.edu> >
> https://lists.cs.columbia.edu/cucslists/listinfo/p2p-sip
<https://lists.cs.columbia.edu/cucslists/listinfo/p2p-sip> 
> <https://lists.cs.columbia.edu/cucslists/listinfo/p2p-sip
<https://lists.cs.columbia.edu/cucslists/listinfo/p2p-sip> >
>
>



From philip_matthews at magma.ca  Fri Dec  8 11:46:02 2006
From: philip_matthews at magma.ca (Philip Matthews)
Date: Fri, 8 Dec 2006 11:46:02 -0500
Subject: [p2p-sip] Concepts draft: Open Issue #1: Searching for services
In-Reply-To: <45798C63.7040209@bbn.com>
References: <B784DF3B-10C8-401B-BCA4-149F7AF0ECD0@magma.ca>	<45794967.6090302@ericsson.com>
	<AE316798-72CA-4C38-913C-389E398A24EB@magma.ca>
	<45798C63.7040209@bbn.com>
Message-ID: <C7978002-90DC-4AB4-9669-A8FA7B8E0A8C@magma.ca>


On 8-Dec-06, at 11:01 , Matt Lepinski wrote:

> Philip,
>
> Thanks a lot for the clarification. In light of your examples, the  
> service search issue makes a lot more sense.
>
> However, from the text of the -03 version of the concepts draft, it  
> is unclear to me what things are "services" (as opposed to say,  
> "resources"). Are you planning to add a definition of a "service"  
> to section 2 of the draft?
>
> - Matt Lepinski
>

Another one of the open issues in the Concept draft is to clarify the  
distinction
(if any) between a "user" and a "resource". I haven't posted this one  
yet because
I don't have a proposed resolution. However, it seems that we should add
"service" to this list.

- Philip


From RADHIKA.R.ROY at saic.com  Fri Dec  8 11:46:09 2006
From: RADHIKA.R.ROY at saic.com (Roy, Radhika R.)
Date: Fri, 8 Dec 2006 11:46:09 -0500
Subject: [p2p-sip] Concepts draft: Open Issue #1: Searching for services
References: <B784DF3B-10C8-401B-BCA4-149F7AF0ECD0@magma.ca><45794967.6090302@ericsson.com><AE316798-72CA-4C38-913C-389E398A24EB@magma.ca><8b2769930612080750h79f512a1x46976b504110f245@mail.gmail.com>
	<A72226BC-FB7F-43EF-8782-2189AA879BF0@magma.ca>
Message-ID: <62D92A9A02BCC845B202323D49A48426E1CF59@0591-ITS-EXMP02.us.saic.com>

Inline [RRR].

________________________________

From: p2p-sip-bounces at cs.columbia.edu on behalf of Philip Matthews
Sent: Fri 12/8/2006 11:42 AM
To: David A. Bryan
Cc: P2P-SIP
Subject: Re: [p2p-sip] Concepts draft: Open Issue #1: Searching for services




On 8-Dec-06, at 10:50 , David A. Bryan wrote:

> I think I had in mind something like what Jiang had in mind -- for
> now, with a simple system, things are stored by URIs, and the method
> for choosing between is out of scope. I soould like a broken record
> today, but I think again that a protocol for selecting the best relay
> from a list is something with broader applicability than P2PSIP --
> might be able to reuse something or if we have to make one, push it to
> a group that can make a more general one.
>
> And although I missed this earlier...Philip: can we use a word other
> than search, since it is simple lookup we are really talking about
> here? Otherwise wording looks great to me.
>
> David

How about we say:
"a lookup of a service returns at least one peer that supports that 
service".

First, this changes "search" to "lookup".

Second, it doesn't take a position on how many peers are returned.
If 5 peers support the service, then the lookup may return just one 
peer,
or it may return all 5. (Of course, when we define the mechanism, we may
need to answer this question, but for now we leave this question as 
open).

[RRR] This is an important issue. In MANET, when the peer is selected to create the route toplogy (e.g., OSPFv3-MDR) in the IP layer in P2P environment, there is an algorithms which is the best one based on cost-performance trade-off distributively.

[RRR] There can be many solutions: 1. Combine application layer P2P logic with just IP network topology logic, 2. An indpendent topology optimization in the SIP application layer, and/or 3. Others.

- Philip
_______________________________________________


-------------- next part --------------
An HTML attachment was scrubbed...
URL: https://lists.cs.columbia.edu/pipermail/p2p-sip/attachments/20061208/f879fe18/attachment.html

From spencer at mcsr-labs.org  Fri Dec  8 11:49:54 2006
From: spencer at mcsr-labs.org (Spencer Dawkins)
Date: Fri, 8 Dec 2006 10:49:54 -0600
Subject: [p2p-sip] Concepts draft: Open Issue #3: Enrollment as
	aPeer/Client Protocol operation?
References: <3E1F6E23-B2EF-43F2-B9EE-3C175384922B@magma.ca>
	<0de901c71acd$1fdb1da0$6701a8c0@china.huawei.com>
	<61D7DDBB-54C1-46CB-A74F-D3DEA084059E@magma.ca>
Message-ID: <111901c71ae8$e3b34520$6701a8c0@china.huawei.com>

Hi, Philip,

Reply inline below...

Thanks!

Spencer

From: "Philip Matthews" <philip_matthews at magma.ca>


> On 8-Dec-06, at 08:31 , Spencer Dawkins wrote:
>
>> Hi, Philip,
>>
>> Sort-of agree. Agree that we need to specify the acceptable format  of a
>> credential.
>>
>> My understanding of this proposed resolution is basically
>>
>> - we don't HAVE to specify the method of obtaining credentials now,  and
>>
>> - if we DO standardize the method of obtaining credentials, it's  not 
>> likely
>> to be part of the peer or client protocols, so would be done  separately 
>> in
>> any event.
>>
>> Do you guys agree? If so, I'd like this understanding to be  included in 
>> the
>> document.
>>
>> Thanks,
>>
>> Spencer
>
> Personally, I don't have a strong opinion on this, and I don't think  any 
> of my
> co-authors do either. We were simply influenced by the fact that (a)  some 
> people
> in the group DO seem to have a strong opinion, and (b) the proposed 
> resolution
> reduces the scope of the initial work of the group, which focuses the 
> group's efforts
> more tightly.
>
> I mostly agree with what you said above. However, is it really true  that 
> we would
> not ever want to add this capability to the peer or client protocols?
>
> - Philip

I liked David's additional statement, which was

"- if we DO standardize, it could be this is useful to others outside
   P2PSIP and might even be some work we push into another WG."

so, taking David's comment and your comment above into account, my list ends 
up as

- we don't HAVE to specify the method of obtaining credentials now, and

- if we DO standardize the method of obtaining credentials, it may not
  be part of the peer and/or client protocol,

- and might developed in another working group, if the resulting
  standardized method would be useful outside P2PSIP.

Does this work for people?

Thanks,

Spencer 



From philip_matthews at magma.ca  Fri Dec  8 12:06:49 2006
From: philip_matthews at magma.ca (Philip Matthews)
Date: Fri, 8 Dec 2006 12:06:49 -0500
Subject: [p2p-sip] Concepts draft: Open Issue #1: Searching for services
In-Reply-To: <45798FB7.3010105@telecomitalia.it>
References: <000001c71a9d$1a0ac6d0$7205a40a@china.huawei.com>
	<45798FB7.3010105@telecomitalia.it>
Message-ID: <D735FF13-6AD4-4A41-B891-2A14580CF3BF@magma.ca>


On 8-Dec-06, at 11:15 , Enrico Marocco wrote:

> JiangXingFeng wrote:
>> For the second one, I think if we think of different service as  
>> resource
>> like SIP URI and put these resources into overlay, it is easy to  
>> get these
>> resource by common search mechanism of DHT.
>
> We must be very careful in taking such a decision, because, with  
> such a
> naive approach, we would store all the istances of the same service
> (apart from replicas) on the same peer and cause dangerous  
> displacements
> to the DHT traffic.  For example, the peer authoritative for the URL
> representing the STUN service, would receive many more queries than
> almost any other peer.

Enrico:

I don't totally understand your concern.

Say I implement service lookup as follows. For each service, there is  
a unique URI for the service (e.g., "STUNServers@<Overlay  
Identifier>" or whatever). The information associated with this URI  
is stored in a single peer in the Overlay (ignoring replicas for  
robustness). Say that peer is X. Then all the peers that offer the  
service "register" with X saying that they offer the service.  When  
peer A does a lookup on "STUNServers@<Overlay Identifier>", then X  
returns one or more peers supporting the service. Peer X might return  
just three peers for each lookup query, and randomly select those  
peers each time, to make sure that no one peer is overloaded.

Is your concern that many peers would want STUN services, and thus  
there would be many lookups on "STUNServers@<Overlay Identifier>" and  
thus peer X would get overloaded?  Or are you concerned about the  
peers offering the service?

- Philip





From enrico.marocco at telecomitalia.it  Fri Dec  8 12:20:52 2006
From: enrico.marocco at telecomitalia.it (Enrico Marocco)
Date: Fri, 08 Dec 2006 18:20:52 +0100
Subject: [p2p-sip] Concepts draft: Open Issue #1: Searching for services
In-Reply-To: <D735FF13-6AD4-4A41-B891-2A14580CF3BF@magma.ca>
References: <000001c71a9d$1a0ac6d0$7205a40a@china.huawei.com>
	<45798FB7.3010105@telecomitalia.it>
	<D735FF13-6AD4-4A41-B891-2A14580CF3BF@magma.ca>
Message-ID: <45799EF4.3050702@telecomitalia.it>

Philip Matthews wrote:
> Say I implement service lookup as follows. For each service, there is a
> unique URI for the service (e.g., "STUNServers@<Overlay Identifier>" or
> whatever). The information associated with this URI is stored in a
> single peer in the Overlay (ignoring replicas for robustness). Say that
> peer is X. Then all the peers that offer the service "register" with X
> saying that they offer the service.  When peer A does a lookup on
> "STUNServers@<Overlay Identifier>", then X returns one or more peers
> supporting the service. Peer X might return just three peers for each
> lookup query, and randomly select those peers each time, to make sure
> that no one peer is overloaded.
> 
> Is your concern that many peers would want STUN services, and thus there
> would be many lookups on "STUNServers@<Overlay Identifier>" and thus
> peer X would get overloaded?  Or are you concerned about the peers
> offering the service?

My concern is the former; in the case you describe, node X, not only
would get overloaded -- and that's already a good reason for being
careful in the decision -- but also, having a more important role, would
be the preferred target of attacks and, in general, a weak point of the
system.

-- 
Ciao,
Enrico

From philip_matthews at magma.ca  Fri Dec  8 12:31:58 2006
From: philip_matthews at magma.ca (Philip Matthews)
Date: Fri, 8 Dec 2006 12:31:58 -0500
Subject: [p2p-sip] Concepts draft: Open Issue #1: Searching for services
In-Reply-To: <62D92A9A02BCC845B202323D49A48426E1CF59@0591-ITS-EXMP02.us.saic.com>
References: <B784DF3B-10C8-401B-BCA4-149F7AF0ECD0@magma.ca><45794967.6090302@ericsson.com><AE316798-72CA-4C38-913C-389E398A24EB@magma.ca><8b2769930612080750h79f512a1x46976b504110f245@mail.gmail.com>
	<A72226BC-FB7F-43EF-8782-2189AA879BF0@magma.ca>
	<62D92A9A02BCC845B202323D49A48426E1CF59@0591-ITS-EXMP02.us.saic.com>
Message-ID: <BB340594-6EC3-4CD8-A119-61A5A55F6D6F@magma.ca>

Radhika:

Let me draw an analogy with searching for a user. Last March, I  
started arguing that it was important that a P2PSIP Overlay support  
searching for a user by attributes other than his/her AoR. For  
example, I wanted to be able to say: show me all users with a first  
name of "Jennifer", and then select the appropriate user to  dial  
from a list. However, David Bryan convinced me that if I had a simple  
AoR to user record lookup, then I could build a more sophisticated  
search mechanism on top of it. For this reason, and because I  
understood the need to reduce the scope, I agreed to the current  
charter saying that more sophisticated searching mechanisms are out- 
of-scope for now.

Similarly, if I provide you with a mechanism to get back the list of  
all peers that support a service, then you could implement a  
sophisticated selection algorithm on top of this. For example, you  
could ping all (or some fraction) of the peers, and select the one  
that was net-path closest. Meanwhile, the overall work of the group  
is reduced for now.

For that reason, I propose that the mechanism to find a service be a  
simple lookup scheme, and not something more sophisticated (for now).  
Later, once we have a peer protocol defined, we can start working on  
extensions that provide a sophisticated search scheme.

- Philip



On 8-Dec-06, at 11:46 , Roy, Radhika R. wrote:

> Inline [RRR].
>
> From: p2p-sip-bounces at cs.columbia.edu on behalf of Philip Matthews
> Sent: Fri 12/8/2006 11:42 AM
> To: David A. Bryan
> Cc: P2P-SIP
> Subject: Re: [p2p-sip] Concepts draft: Open Issue #1: Searching for  
> services
>
>
> On 8-Dec-06, at 10:50 , David A. Bryan wrote:
>
> > I think I had in mind something like what Jiang had in mind -- for
> > now, with a simple system, things are stored by URIs, and the method
> > for choosing between is out of scope. I soould like a broken record
> > today, but I think again that a protocol for selecting the best  
> relay
> > from a list is something with broader applicability than P2PSIP --
> > might be able to reuse something or if we have to make one, push  
> it to
> > a group that can make a more general one.
> >
> > And although I missed this earlier...Philip: can we use a word other
> > than search, since it is simple lookup we are really talking about
> > here? Otherwise wording looks great to me.
> >
> > David
>
> How about we say:
> "a lookup of a service returns at least one peer that supports that
> service".
>
> First, this changes "search" to "lookup".
>
> Second, it doesn't take a position on how many peers are returned.
> If 5 peers support the service, then the lookup may return just one
> peer,
> or it may return all 5. (Of course, when we define the mechanism,  
> we may
> need to answer this question, but for now we leave this question as
> open).
>
> [RRR] This is an important issue. In MANET, when the peer is  
> selected to create the route toplogy (e.g., OSPFv3-MDR) in the IP  
> layer in P2P environment, there is an algorithms which is the best  
> one based on cost-performance trade-off distributively.
>
> [RRR] There can be many solutions: 1. Combine application layer P2P  
> logic with just IP network topology logic, 2. An indpendent  
> topology optimization in the SIP application layer, and/or 3. Others.
>
> - Philip
> _______________________________________________
>

-------------- next part --------------
An HTML attachment was scrubbed...
URL: https://lists.cs.columbia.edu/pipermail/p2p-sip/attachments/20061208/fbc0a3fc/attachment-0001.html

From philip_matthews at magma.ca  Fri Dec  8 15:03:15 2006
From: philip_matthews at magma.ca (Philip Matthews)
Date: Fri, 8 Dec 2006 15:03:15 -0500
Subject: [p2p-sip] Concepts draft: Open Issue #1: Searching for services
In-Reply-To: <45799EF4.3050702@telecomitalia.it>
References: <000001c71a9d$1a0ac6d0$7205a40a@china.huawei.com>
	<45798FB7.3010105@telecomitalia.it>
	<D735FF13-6AD4-4A41-B891-2A14580CF3BF@magma.ca>
	<45799EF4.3050702@telecomitalia.it>
Message-ID: <C7CD7926-4D4E-45EF-B8FD-557B3F64C430@magma.ca>


On 8-Dec-06, at 12:20 , Enrico Marocco wrote:

> Philip Matthews wrote:
>> Say I implement service lookup as follows. For each service, there  
>> is a
>> unique URI for the service (e.g., "STUNServers@<Overlay  
>> Identifier>" or
>> whatever). The information associated with this URI is stored in a
>> single peer in the Overlay (ignoring replicas for robustness). Say  
>> that
>> peer is X. Then all the peers that offer the service "register"  
>> with X
>> saying that they offer the service.  When peer A does a lookup on
>> "STUNServers@<Overlay Identifier>", then X returns one or more peers
>> supporting the service. Peer X might return just three peers for each
>> lookup query, and randomly select those peers each time, to make sure
>> that no one peer is overloaded.
>>
>> Is your concern that many peers would want STUN services, and thus  
>> there
>> would be many lookups on "STUNServers@<Overlay Identifier>" and thus
>> peer X would get overloaded?  Or are you concerned about the peers
>> offering the service?
>
> My concern is the former; in the case you describe, node X, not only
> would get overloaded -- and that's already a good reason for being
> careful in the decision -- but also, having a more important role,  
> would
> be the preferred target of attacks and, in general, a weak point of  
> the
> system.

I now see the problem you are concerned about.
I haven't thought a lot about it, but I am aware that some DHTs,
Kademlia for example, attempt to deal with this by caching the
results of a lookup along the lookup path so that frequent queries
can be answered faster.

- Philip

From dean.willis at softarmor.com  Sat Dec  9 04:37:04 2006
From: dean.willis at softarmor.com (Dean Willis)
Date: Sat, 9 Dec 2006 03:37:04 -0600
Subject: [p2p-sip] Where are we "getting wrapped around the DNS axle"
In-Reply-To: <05ef01c718b5$506df880$0202a8c0@Quinthar>
References: <05ef01c718b5$506df880$0202a8c0@Quinthar>
Message-ID: <041B9C9D-76F0-409A-9914-FA1CEBFB869E@softarmor.com>



You seem to be missing the point.

You CANNOT put, using anything like current resource records, a multi- 
hop SIP route into DNS.

Furthermore, you cannot make DNS return a different route from Alice  
to Bob than from Charlie to Bob.

So when Alice needs to route her requests through U, V, and W to  
reach Bob, but Charlie needs to route his requests through X, Y and Z  
to reach Bob, DNS (or mDNS) or anything vaguely DNS-like is a  
completely inappropriate tool for discovering those routes.

Sure, it works to use DNS to discover the thing that will help you  
find an entry into the routing topology (which we seem to be calling  
a "bootstrap"), and DNS works for some subset of the P2P problems  
when one assumes artificially simple topologies, but DNS is NOT an  
answer for the general problem here.


--
Dean

On Dec 5, 2006, at 3:35 PM, David Barrett wrote:

>> -----Original Message-----
>> From: Enrico Marocco [mailto:enrico.marocco at telecomitalia.it]
>> Subject: Re: [p2p-sip] Where are we "getting wrapped around the  
>> DNS axle"
>>
>>> In other words, do you agree:
>>>
>>> 	1) P2P-SIP must guarantee that each name resolves to a single user
>>
>> Generally speaking, no.  The peer protocol must guarantee that the  
>> name
>> used by a user enrolled within an overlay could not be used by  
>> another
>> user in the _same_ overlay.
>
> I'll grant it can be useful to have a totally unauthenticated,  
> insecure
> namespace where you can be Enrico for a while and then the moment  
> you log
> off I can log on as Enrico and be indistinguishable to everyone  
> else.  I'd
> wager this most often works in an ad hoc or "trade show"  
> environment, where
> mDNS is totally adequate.
>
> However, I'm assuming a more valuable use case is an authenticated  
> world
> that has long-lived names verified in a secure manner.  And in this  
> world,
> DNS is totally adequate, for networks as small as a few fixed nodes  
> in the
> jungle, or as big as the entire internet.
>
> What I'm waiting to hear about is a critical set of users who:
>
> 	- Don't need authentication
> 	- Can't use mDNS
>
> Or, a critical set of users who:
>
> 	- Do need authentication
> 	- Can't use DNS
>
> I'll totally grant we can construct a network to fit either of  
> these.  For
> the first, set up a network in the jungle and explicitly disable  
> multicast.
> For the second, set up a private network with a thousand nodes, and  
> refuse
> to install a DNS server (basically, disable all other network activity
> except Gnutella and eDonkey).
>
> But these don't seem like realistic use cases.  In the first, just  
> tell
> these jungle warriors "enable multicast" and they'll be fine.  In the
> second, they're going to need a DNS server anyway, so just tell  
> them to use
> that.
>
>
>>> 	2) Multiple users want the same name
>>
>> Possibly, but if an enrollment mechanism is _deployed_ in the overlay
>> (please note that it may be standardized as for jabber, or not), only
>> the firstly enrolled user can use it.
>
> Until he logs off or temporarily loses network connectivity, in  
> which anyone
> can steal his name.  (Though I guess we could have him generate a  
> private
> key and the first to register a public key with a name "wins" and  
> if you
> lose that private key you're basically screwed because there's no  
> "password
> change" feature.)
>
> -david
>
> _______________________________________________
> p2p-sip mailing list
> p2p-sip at cs.columbia.edu
> https://lists.cs.columbia.edu/cucslists/listinfo/p2p-sip
>


From dbarrett at quinthar.com  Sat Dec  9 06:56:24 2006
From: dbarrett at quinthar.com (David Barrett)
Date: Sat, 9 Dec 2006 03:56:24 -0800
Subject: [p2p-sip] Where are we "getting wrapped around the DNS axle"
In-Reply-To: <041B9C9D-76F0-409A-9914-FA1CEBFB869E@softarmor.com>
Message-ID: <006301c71b89$0d17edd0$0202a8c0@Quinthar>

You might consider googling "DNS MANET" and see the variety of interesting
resources on how the DNS problem has been solved for MANET networks.  The
first in the list is:

http://www-users.cs.umn.edu/~jjeong/publications/ietf-internet-draft/draft-j
eong-manet-dns-service-00.txt

  "This document specifies an architecture of DNS service system for 
   mobile ad hoc network which might be connected to the Internet.  The 
   resolution of DNS names of mobile nodes within mobile ad hoc network 
   is performed by multicast DNS and that of DNS names of nodes in the 
   Internet is performed through DNS autoconfiguration of recursive DNS 
   server."

I'm not sure what we're arguing about.  I'm saying if you want to resolve
names to IP addresses, DNS does it, for virtually every type of network
invented, including MANET.  It's been doing the trick a million times a
second for decades, all over the world.  Given this, I'm employing the
logic:

- SIP uses names
- SIP uses IP addresses
- SIP should use DNS to resolve names to IP addresses whenever possible
- DNS works everywhere
- Maybe SIP should use DNS everywhere for name/IP resolution.

Where does this logic break down?

People will continue to invent new types of networks in the future.  And I
don't doubt that with more digging we can come up with a network that
doesn't currently have a clear answer to how DNS will be supported.

But any network that hopes to be adopted in any real way needs to support
DNS, else users won't use it.  Every major network application (email, web,
ftp, ssh, etc) uses DNS, so it seems extremely probable that every network
that will come out in the near or even far future will have some support for
DNS.  We've seen this pattern happen with private networks, with ad hoc
networks, and now with multi-hop ad hoc networks.  I can't visualize what's
more extreme than that (interplanetary links?) but I'd wager DNS will be fit
into that somehow, too.

The best way to future-proof SIP is to just use DNS.

-david

PS: As for multi-hop routing, while I'm certainly not the expert on this
subject, I thought MANET worked at the IP level.  To the application,
everything is still IP addresses.  Thus any facility for using SIP to create
a virtual multi-hop channel wouldn't be necessary, because MANET does the
multi-hop one layer down.  In other words, everyone in a MANET network still
has a normal IP address, so we're back to the problem of resolving names to
IPs, which DNS does exquisitely.

> -----Original Message-----
> From: Dean Willis [mailto:dean.willis at softarmor.com]
> Sent: Saturday, December 09, 2006 1:37 AM
> To: David Barrett
> Cc: 'Enrico Marocco'; 'P2P-SIP'; 'David A. Bryan'
> Subject: Re: [p2p-sip] Where are we "getting wrapped around the DNS axle"
> 
> 
> 
> You seem to be missing the point.
> 
> You CANNOT put, using anything like current resource records, a multi-
> hop SIP route into DNS.
> 
> Furthermore, you cannot make DNS return a different route from Alice
> to Bob than from Charlie to Bob.
> 
> So when Alice needs to route her requests through U, V, and W to
> reach Bob, but Charlie needs to route his requests through X, Y and Z
> to reach Bob, DNS (or mDNS) or anything vaguely DNS-like is a
> completely inappropriate tool for discovering those routes.
> 
> Sure, it works to use DNS to discover the thing that will help you
> find an entry into the routing topology (which we seem to be calling
> a "bootstrap"), and DNS works for some subset of the P2P problems
> when one assumes artificially simple topologies, but DNS is NOT an
> answer for the general problem here.
> 
> 
> --
> Dean
> 
> On Dec 5, 2006, at 3:35 PM, David Barrett wrote:
> 
> >> -----Original Message-----
> >> From: Enrico Marocco [mailto:enrico.marocco at telecomitalia.it]
> >> Subject: Re: [p2p-sip] Where are we "getting wrapped around the
> >> DNS axle"
> >>
> >>> In other words, do you agree:
> >>>
> >>> 	1) P2P-SIP must guarantee that each name resolves to a single user
> >>
> >> Generally speaking, no.  The peer protocol must guarantee that the
> >> name
> >> used by a user enrolled within an overlay could not be used by
> >> another
> >> user in the _same_ overlay.
> >
> > I'll grant it can be useful to have a totally unauthenticated,
> > insecure
> > namespace where you can be Enrico for a while and then the moment
> > you log
> > off I can log on as Enrico and be indistinguishable to everyone
> > else.  I'd
> > wager this most often works in an ad hoc or "trade show"
> > environment, where
> > mDNS is totally adequate.
> >
> > However, I'm assuming a more valuable use case is an authenticated
> > world
> > that has long-lived names verified in a secure manner.  And in this
> > world,
> > DNS is totally adequate, for networks as small as a few fixed nodes
> > in the
> > jungle, or as big as the entire internet.
> >
> > What I'm waiting to hear about is a critical set of users who:
> >
> > 	- Don't need authentication
> > 	- Can't use mDNS
> >
> > Or, a critical set of users who:
> >
> > 	- Do need authentication
> > 	- Can't use DNS
> >
> > I'll totally grant we can construct a network to fit either of
> > these.  For
> > the first, set up a network in the jungle and explicitly disable
> > multicast.
> > For the second, set up a private network with a thousand nodes, and
> > refuse
> > to install a DNS server (basically, disable all other network activity
> > except Gnutella and eDonkey).
> >
> > But these don't seem like realistic use cases.  In the first, just
> > tell
> > these jungle warriors "enable multicast" and they'll be fine.  In the
> > second, they're going to need a DNS server anyway, so just tell
> > them to use
> > that.
> >
> >
> >>> 	2) Multiple users want the same name
> >>
> >> Possibly, but if an enrollment mechanism is _deployed_ in the overlay
> >> (please note that it may be standardized as for jabber, or not), only
> >> the firstly enrolled user can use it.
> >
> > Until he logs off or temporarily loses network connectivity, in
> > which anyone
> > can steal his name.  (Though I guess we could have him generate a
> > private
> > key and the first to register a public key with a name "wins" and
> > if you
> > lose that private key you're basically screwed because there's no
> > "password
> > change" feature.)
> >
> > -david
> >
> > _______________________________________________
> > p2p-sip mailing list
> > p2p-sip at cs.columbia.edu
> > https://lists.cs.columbia.edu/cucslists/listinfo/p2p-sip
> >


From adamfisk at gmail.com  Sun Dec 10 16:50:46 2006
From: adamfisk at gmail.com (Adam Fisk)
Date: Sun, 10 Dec 2006 16:50:46 -0500
Subject: [p2p-sip] First Rule of Distributed Computing:
Message-ID: <f40963db0612101350x4f553627nfec3e0af50a03c57@mail.gmail.com>

Don't distribute your objects.

No specific point to be made here, but I find that rule useful in my own
coding and thinking, and it's surprisingly easy to forget.  If you can avoid
distributing your objects, you're usually left with a much easier system to
implement and maintain.  Is that Radia Perlman?  Not sure.

-Adam
-------------- next part --------------
An HTML attachment was scrubbed...
URL: https://lists.cs.columbia.edu/pipermail/p2p-sip/attachments/20061210/1d1ddee5/attachment.html

From hsinnrei at adobe.com  Sun Dec 10 18:43:47 2006
From: hsinnrei at adobe.com (Henry Sinnreich)
Date: Sun, 10 Dec 2006 15:43:47 -0800
Subject: [p2p-sip] Concepts draft: Open Issue #1: Searching for services
In-Reply-To: <BB340594-6EC3-4CD8-A119-61A5A55F6D6F@magma.ca>
Message-ID: <24CCCC428EFEA2469BF046DB3C7A8D221413E6@namail5.corp.adobe.com>

Range searches such as Philip mentions (last names, etc.) require
different P2P overlays and Google will reveal to you an impressive body
of work.

 

There are good reasons to delay this topic until someone volunteers to
write an I-D with all that's required.

 

Henry

 

________________________________

From: p2p-sip-bounces at cs.columbia.edu
[mailto:p2p-sip-bounces at cs.columbia.edu] On Behalf Of Philip Matthews
Sent: Friday, December 08, 2006 11:32 AM
To: Roy, Radhika R.
Cc: P2P-SIP; David A. Bryan
Subject: Re: [p2p-sip] Concepts draft: Open Issue #1: Searching for
services

 

Radhika:

 

Let me draw an analogy with searching for a user. Last March, I started
arguing that it was important that a P2PSIP Overlay support searching
for a user by attributes other than his/her AoR. For example, I wanted
to be able to say: show me all users with a first name of "Jennifer",
and then select the appropriate user to  dial from a list. However,
David Bryan convinced me that if I had a simple AoR to user record
lookup, then I could build a more sophisticated search mechanism on top
of it. For this reason, and because I understood the need to reduce the
scope, I agreed to the current charter saying that more sophisticated
searching mechanisms are out-of-scope for now.

 

Similarly, if I provide you with a mechanism to get back the list of all
peers that support a service, then you could implement a sophisticated
selection algorithm on top of this. For example, you could ping all (or
some fraction) of the peers, and select the one that was net-path
closest. Meanwhile, the overall work of the group is reduced for now.

 

For that reason, I propose that the mechanism to find a service be a
simple lookup scheme, and not something more sophisticated (for now).
Later, once we have a peer protocol defined, we can start working on
extensions that provide a sophisticated search scheme.

 

- Philip

 

 

 

On 8-Dec-06, at 11:46 , Roy, Radhika R. wrote:





Inline [RRR].

 

________________________________

From: p2p-sip-bounces at cs.columbia.edu on behalf of Philip Matthews
Sent: Fri 12/8/2006 11:42 AM
To: David A. Bryan
Cc: P2P-SIP
Subject: Re: [p2p-sip] Concepts draft: Open Issue #1: Searching for
services

 

On 8-Dec-06, at 10:50 , David A. Bryan wrote:

> I think I had in mind something like what Jiang had in mind -- for
> now, with a simple system, things are stored by URIs, and the method
> for choosing between is out of scope. I soould like a broken record
> today, but I think again that a protocol for selecting the best relay
> from a list is something with broader applicability than P2PSIP --
> might be able to reuse something or if we have to make one, push it to
> a group that can make a more general one.
>
> And although I missed this earlier...Philip: can we use a word other
> than search, since it is simple lookup we are really talking about
> here? Otherwise wording looks great to me.
>
> David

How about we say:
"a lookup of a service returns at least one peer that supports that 
service".

First, this changes "search" to "lookup".

Second, it doesn't take a position on how many peers are returned.
If 5 peers support the service, then the lookup may return just one 
peer,
or it may return all 5. (Of course, when we define the mechanism, we may
need to answer this question, but for now we leave this question as 
open).

[RRR] This is an important issue. In MANET, when the peer is selected to
create the route toplogy (e.g., OSPFv3-MDR) in the IP layer in P2P
environment, there is an algorithms which is the best one based on
cost-performance trade-off distributively.

[RRR] There can be many solutions: 1. Combine application layer P2P
logic with just IP network topology logic, 2. An indpendent topology
optimization in the SIP application layer, and/or 3. Others.

- Philip
_______________________________________________

 

-------------- next part --------------
An HTML attachment was scrubbed...
URL: https://lists.cs.columbia.edu/pipermail/p2p-sip/attachments/20061210/d4ef266d/attachment-0001.html

From dbarrett at quinthar.com  Sun Dec 10 23:44:51 2006
From: dbarrett at quinthar.com (David Barrett)
Date: Sun, 10 Dec 2006 20:44:51 -0800
Subject: [p2p-sip] First Rule of Distributed Computing:
In-Reply-To: <f40963db0612101350x4f553627nfec3e0af50a03c57@mail.gmail.com>
Message-ID: <014001c71cdf$1e0139b0$9d00a8c0@Quinthar>

Heh, this seems like very good advice, but I'm not sure I follow your
meaning.  Can you give us an example?

 

-david

 

  _____  

From: p2p-sip-bounces at cs.columbia.edu
[mailto:p2p-sip-bounces at cs.columbia.edu] On Behalf Of Adam Fisk
Sent: Sunday, December 10, 2006 1:51 PM
To: p2p-sip at cs.columbia.edu
Subject: [p2p-sip] First Rule of Distributed Computing:

 

Don't distribute your objects.  

No specific point to be made here, but I find that rule useful in my own
coding and thinking, and it's surprisingly easy to forget.  If you can avoid
distributing your objects, you're usually left with a much easier system to
implement and maintain.  Is that Radia Perlman?  Not sure. 

-Adam

-------------- next part --------------
An HTML attachment was scrubbed...
URL: https://lists.cs.columbia.edu/pipermail/p2p-sip/attachments/20061210/55905ede/attachment.html

From jani.hautakorpi at ericsson.com  Mon Dec 11 08:28:04 2006
From: jani.hautakorpi at ericsson.com (Jani Hautakorpi (JO/LMF))
Date: Mon, 11 Dec 2006 15:28:04 +0200
Subject: [p2p-sip] Concepts draft: Open Issue #3: Enrollment
 as	aPeer/Client Protocol operation?
In-Reply-To: <111901c71ae8$e3b34520$6701a8c0@china.huawei.com>
References: <3E1F6E23-B2EF-43F2-B9EE-3C175384922B@magma.ca>	<0de901c71acd$1fdb1da0$6701a8c0@china.huawei.com>	<61D7DDBB-54C1-46CB-A74F-D3DEA084059E@magma.ca>
	<111901c71ae8$e3b34520$6701a8c0@china.huawei.com>
Message-ID: <457D5CE4.1070708@ericsson.com>

Hi all,

Spencer Dawkins wrote:
> I liked David's additional statement, which was
> 
> "- if we DO standardize, it could be this is useful to others outside
>    P2PSIP and might even be some work we push into another WG."
> 
> so, taking David's comment and your comment above into account, my list ends 
> up as
> 
> - we don't HAVE to specify the method of obtaining credentials now, and
> 
> - if we DO standardize the method of obtaining credentials, it may not
>   be part of the peer and/or client protocol,
> 
> - and might developed in another working group, if the resulting
>   standardized method would be useful outside P2PSIP.
> 
> Does this work for people?

This makes sense and sounds good at least to me.


-- 
Jani

From adamfisk at gmail.com  Mon Dec 11 10:11:22 2006
From: adamfisk at gmail.com (Adam Fisk)
Date: Mon, 11 Dec 2006 10:11:22 -0500
Subject: [p2p-sip] First Rule of Distributed Computing:
In-Reply-To: <014001c71cdf$1e0139b0$9d00a8c0@Quinthar>
References: <f40963db0612101350x4f553627nfec3e0af50a03c57@mail.gmail.com>
	<014001c71cdf$1e0139b0$9d00a8c0@Quinthar>
Message-ID: <f40963db0612110711r40fa3281we5e061a57fa9cae8@mail.gmail.com>

The example I encounter most frequently is search.  Distributed search is
hard.  It's hard to search a lot of computers without taking up a lot of
bandwidth, not to mention the algorithms are just tough and require massive
coordination amongst many machines.  The bandwidth constraints lead to
architectures that increase latency even if they're scalable, like DHTs or
distributed indexing with random walk-type approaches on top of them.  The
bandwidth constraints and latency also put a constraint on message sizes,
significantly limiting your options for whatever protocol you're working
with.  Updates are also hard.  Instead of updating a server cluster, you
have to update sometimes millions of machines.

With centralized search, these constraints largely dissappear.  You get low
latency, global search, all with much less work.  Things like collaborative
filtering and tagging also become possible, whereas they are extremely
complicated at best with distributed systems.  Sure there's a single point
of failure, but that's rarely an issue in practice if you know what you're
doing.

There have to be really, really good reasons to choose the first,
distributed approach.  In the case of the file sharing networks, that reason
was legal liability.  In other cases, it might be cost.

That's just a rule I frequently have to remind myself of, and it's saved me
a great deal of time in the past.

-Adam


On 12/10/06, David Barrett <dbarrett at quinthar.com> wrote:
>
>  Heh, this seems like very good advice, but I'm not sure I follow your
> meaning.  Can you give us an example?
>
>
>
> -david
>
>
>   ------------------------------
>
> *From:* p2p-sip-bounces at cs.columbia.edu [mailto:
> p2p-sip-bounces at cs.columbia.edu] *On Behalf Of *Adam Fisk
> *Sent:* Sunday, December 10, 2006 1:51 PM
> *To:* p2p-sip at cs.columbia.edu
> *Subject:* [p2p-sip] First Rule of Distributed Computing:
>
>
>
> Don't distribute your objects.
>
> No specific point to be made here, but I find that rule useful in my own
> coding and thinking, and it's surprisingly easy to forget.  If you can avoid
> distributing your objects, you're usually left with a much easier system to
> implement and maintain.  Is that Radia Perlman?  Not sure.
>
> -Adam
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: https://lists.cs.columbia.edu/pipermail/p2p-sip/attachments/20061211/fcb6696b/attachment.html

From eric_d_cooper at sympatico.ca  Mon Dec 11 20:31:52 2006
From: eric_d_cooper at sympatico.ca (Eric Cooper)
Date: Mon, 11 Dec 2006 20:31:52 -0500
Subject: [p2p-sip] Concepts draft: Open Issue #1: Searching for services
References: <B784DF3B-10C8-401B-BCA4-149F7AF0ECD0@magma.ca><45794967.6090302@ericsson.com><AE316798-72CA-4C38-913C-389E398A24EB@magma.ca><8b2769930612080750h79f512a1x46976b504110f245@mail.gmail.com><A72226BC-FB7F-43EF-8782-2189AA879BF0@magma.ca><62D92A9A02BCC845B202323D49A48426E1CF59@0591-ITS-EXMP02.us.saic.com>
	<BB340594-6EC3-4CD8-A119-61A5A55F6D6F@magma.ca>
Message-ID: <BAYC1-PASMTP0893BD362B04F3645CA89FCCD70@CEZ.ICE>

I don't quite see the connection between the AoR to user-record lookup and the subsequent ability to perform more sophisticated searches but leaving that aside, I like the description below.

However, I think the key is the ability to query for *all* the peers.  If you don't know all the peers, you probably can't apply useful filters in the future.

As applied to AoRs & contact data, it's probably not a problem to return all the contact data.  However, the list of all peers supporting a particular service could be quite large.  Possibly a significant fraction of the peers in the overlay.

If large responses are going to be a problem, we may be forced to define mechanisms for more sophisticated searches now.

That said, I don't really see a problem with large responses.  They could just keep rolling in until they are all delivered (or the lookup is cancelled).

Eric Cooper.
 
----- Original Message ----- 
From: "Philip Matthews" <philip_matthews at magma.ca>
To: "Roy, Radhika R." <RADHIKA.R.ROY at saic.com>
Cc: "P2P-SIP" <p2p-sip at cs.columbia.edu>; "David A. Bryan" <bryan at ethernot.org>
Sent: Friday, December 08, 2006 12:31 PM
Subject: Re: [p2p-sip] Concepts draft: Open Issue #1: Searching for services


> Radhika:
> 
> Let me draw an analogy with searching for a user. Last March, I  
> started arguing that it was important that a P2PSIP Overlay support  
> searching for a user by attributes other than his/her AoR. For  
> example, I wanted to be able to say: show me all users with a first  
> name of "Jennifer", and then select the appropriate user to  dial  
> from a list. However, David Bryan convinced me that if I had a simple  
> AoR to user record lookup, then I could build a more sophisticated  
> search mechanism on top of it. For this reason, and because I  
> understood the need to reduce the scope, I agreed to the current  
> charter saying that more sophisticated searching mechanisms are out- 
> of-scope for now.
> 
> Similarly, if I provide you with a mechanism to get back the list of  
> all peers that support a service, then you could implement a  
> sophisticated selection algorithm on top of this. For example, you  
> could ping all (or some fraction) of the peers, and select the one  
> that was net-path closest. Meanwhile, the overall work of the group  
> is reduced for now.
> 
> For that reason, I propose that the mechanism to find a service be a  
> simple lookup scheme, and not something more sophisticated (for now).  
> Later, once we have a peer protocol defined, we can start working on  
> extensions that provide a sophisticated search scheme.
> 
> - Philip
> 
> 
> 
> On 8-Dec-06, at 11:46 , Roy, Radhika R. wrote:
> 
>> Inline [RRR].
>>
>> From: p2p-sip-bounces at cs.columbia.edu on behalf of Philip Matthews
>> Sent: Fri 12/8/2006 11:42 AM
>> To: David A. Bryan
>> Cc: P2P-SIP
>> Subject: Re: [p2p-sip] Concepts draft: Open Issue #1: Searching for  
>> services
>>
>>
>> On 8-Dec-06, at 10:50 , David A. Bryan wrote:
>>
>> > I think I had in mind something like what Jiang had in mind -- for
>> > now, with a simple system, things are stored by URIs, and the method
>> > for choosing between is out of scope. I soould like a broken record
>> > today, but I think again that a protocol for selecting the best  
>> relay
>> > from a list is something with broader applicability than P2PSIP --
>> > might be able to reuse something or if we have to make one, push  
>> it to
>> > a group that can make a more general one.
>> >
>> > And although I missed this earlier...Philip: can we use a word other
>> > than search, since it is simple lookup we are really talking about
>> > here? Otherwise wording looks great to me.
>> >
>> > David
>>
>> How about we say:
>> "a lookup of a service returns at least one peer that supports that
>> service".
>>
>> First, this changes "search" to "lookup".
>>
>> Second, it doesn't take a position on how many peers are returned.
>> If 5 peers support the service, then the lookup may return just one
>> peer,
>> or it may return all 5. (Of course, when we define the mechanism,  
>> we may
>> need to answer this question, but for now we leave this question as
>> open).
>>
>> [RRR] This is an important issue. In MANET, when the peer is  
>> selected to create the route toplogy (e.g., OSPFv3-MDR) in the IP  
>> layer in P2P environment, there is an algorithms which is the best  
>> one based on cost-performance trade-off distributively.
>>
>> [RRR] There can be many solutions: 1. Combine application layer P2P  
>> logic with just IP network topology logic, 2. An indpendent  
>> topology optimization in the SIP application layer, and/or 3. Others.
>>
>> - Philip
>> _______________________________________________
>>
> 
>


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


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


From hsinnrei at adobe.com  Mon Dec 11 22:50:57 2006
From: hsinnrei at adobe.com (Henry Sinnreich)
Date: Mon, 11 Dec 2006 19:50:57 -0800
Subject: [p2p-sip] First Rule of Distributed Computing:
In-Reply-To: <f40963db0612110711r40fa3281we5e061a57fa9cae8@mail.gmail.com>
Message-ID: <24CCCC428EFEA2469BF046DB3C7A8D221413FE@namail5.corp.adobe.com>

Adam Fisk wrote:

>There have to be really, really good reasons to choose the first, distributed approach.

Right. Here are some reasons in the case of Client-Server SIP VoIP. There are about 10-100 VoIP network elements in larger VoIP networks, such as:

* DNS, ENUM servers with VoIP data entries
* SIP registrar and proxy servers, fronting SIP proxies
* Presence servers
* Component servers: 
  - Announcements, voice mail, media servers, conference servers
* Session border controllers (at several points in the network)
* Softswitches: Class 5 switch emulation and/or hosted IP PBXs (Centrex)
* Media servers
* Network elements for QoS
* Policy servers
* Network management
* Voice quality monitoring probes
* Network engineering
* System integration
* VoIP network IT (OSS) systems for all the above
* IT systems provisioned with churning user data for most of the above - this is the most onerous part.
Note that I did not count TDM gateways since gateways would be needed for P2P as well.

This is a nightmare of an infrastructure, since PCs at the endpoints are more than powerful enough to provide these functions and also to move the cost of hosting and bandwidth to the endpoints. There is no P2P VoIP infrastructure and P2P systems are self organizing.

Skype with its >100M users has proven in the market that P2P avoids this complexity and cost (staffed operations 7x24, staffed custom system integration, huge IT systems development, etc.) and most of all, lack of flexibility can be mostly avoided and still provide rich consumer presence, IM, VoIP, video, file transfer and now also conferencing and various enterprise applications. 

I don't have profit numbers for Skype (they had a good business model based on prepaid float I believe before becoming eBay), but I have not seen any profit statements from any VoIP company that I know of, except when VoIP is bundled with broadband Internet or with something else. Does the bundling hide the losses for VoIP service?

VoIP is just an application sitting in the endpoints and that's how it should be on "the dumb network". 
Since when do we need the "Intelligent Network" on the Internet (the IETF's task)? 
This is the rationale for developing standards for P2P SIP.

Leaving VoIP aside, there are roughly 5,000 papers that deal with P2P and for what it is worth, p2p traffic dominates the Internet. Obviously for good reasons.

Thanks, Henry
________________________________________
From: p2p-sip-bounces at cs.columbia.edu [mailto:p2p-sip-bounces at cs.columbia.edu] On Behalf Of Adam Fisk
Sent: Monday, December 11, 2006 9:11 AM
To: David Barrett
Cc: p2p-sip at cs.columbia.edu
Subject: Re: [p2p-sip] First Rule of Distributed Computing:

The example I encounter most frequently is search.? Distributed search is hard.? It's hard to search a lot of computers without taking up a lot of bandwidth, not to mention the algorithms are just tough and require massive coordination amongst many machines.? The bandwidth constraints lead to architectures that increase latency even if they're scalable, like DHTs or distributed indexing with random walk-type approaches on top of them.? The bandwidth constraints and latency also put a constraint on message sizes, significantly limiting your options for whatever protocol you're working with.? Updates are also hard.? Instead of updating a server cluster, you have to update sometimes millions of machines. 

With centralized search, these constraints largely dissappear.? You get low latency, global search, all with much less work.? Things like collaborative filtering and tagging also become possible, whereas they are extremely complicated at best with distributed systems.? Sure there's a single point of failure, but that's rarely an issue in practice if you know what you're doing. 

There have to be really, really good reasons to choose the first, distributed approach.? In the case of the file sharing networks, that reason was legal liability.? In other cases, it might be cost.

That's just a rule I frequently have to remind myself of, and it's saved me a great deal of time in the past. 

-Adam

On 12/10/06, David Barrett <dbarrett at quinthar.com> wrote:
Heh, this seems like very good advice, but I'm not sure I follow your meaning.? Can you give us an example?
?
-david
?
________________________________________
From: p2p-sip-bounces at cs.columbia.edu [mailto: p2p-sip-bounces at cs.columbia.edu] On Behalf Of Adam Fisk
Sent: Sunday, December 10, 2006 1:51 PM
To: p2p-sip at cs.columbia.edu
Subject: [p2p-sip] First Rule of Distributed Computing:
?
Don't distribute your objects.? 

No specific point to be made here, but I find that rule useful in my own coding and thinking, and it's surprisingly easy to forget.? If you can avoid distributing your objects, you're usually left with a much easier system to implement and maintain.? Is that Radia Perlman?? Not sure. 

-Adam



From jiang.x.f at huawei.com  Tue Dec 12 04:32:54 2006
From: jiang.x.f at huawei.com (JiangXingFeng)
Date: Tue, 12 Dec 2006 17:32:54 +0800
Subject: [p2p-sip] Concepts draft: Open Issue #1: Searching for services
In-Reply-To: <8b2769930612080750h79f512a1x46976b504110f245@mail.gmail.com>
Message-ID: <000601c71dd0$80803f40$7205a40a@china.huawei.com>

Hi, David:
  Do you think that the choice from several service nodes is a reasonable
requirement which may be designed more general? If it is, I agree with your
points. The choosing criteria are changing along with different service
requirements, so a general one will be well suited to various applications. 

Regards!
--
Jiang XingFeng
> -----Original Message-----
> From: p2p-sip-bounces at cs.columbia.edu
> [mailto:p2p-sip-bounces at cs.columbia.edu] On Behalf Of David A. Bryan
> Sent: Friday, December 08, 2006 11:51 PM
> To: Philip Matthews
> Cc: P2P-SIP
> Subject: Re: [p2p-sip] Concepts draft: Open Issue #1: Searching for
services
> 
> I think I had in mind something like what Jiang had in mind -- for
> now, with a simple system, things are stored by URIs, and the method
> for choosing between is out of scope. I soould like a broken record
> today, but I think again that a protocol for selecting the best relay
> from a list is something with broader applicability than P2PSIP --
> might be able to reuse something or if we have to make one, push it to
> a group that can make a more general one.
> 
> And although I missed this earlier...Philip: can we use a word other
> than search, since it is simple lookup we are really talking about
> here? Otherwise wording looks great to me.
> 
> David
> 
> On 12/8/06, Philip Matthews <philip_matthews at magma.ca> wrote:
> >
> > On 8-Dec-06, at 06:15 , Jani Hautakorpi (JO/LMF) wrote:
> >
> > > Hi Philip,
> > >
> > > Philip Matthews wrote:
> > >> Open issue #1:  Can different peers in a P2PSIP Overlay offer
> > >> different services to the Overlay? If so, do we need the
> > >> concept of  searching for the set of peers that offer a service?
> > >
> > > Do you have some concrete services in mind?
> > >
> > > My initial impression is that a search functionality would make
> > > sense only if the services would be such that they couldn't be
> > > identified with a SIP URI.
> >
> > Hi Jani:
> >
> > Here are my opinions of what such services might be (my co-authors
> > may not agree! ;-).
> >
> > One service might be STUN and TURN server service. Some peers might
> > offer both STUN and TURN, some might offer just STUN, and some might
> > offer neither.
> > Other services might be SIP Proxy server service, or SIP Redirect
> > server service, etc.  Other services might be the willingness to host
> > clients.
> >
> > There has been some debate over which services all peers must offer,
> > and which peers can optionally offer.  Rather than continue that
> > debate right now, the authors are simply saying:
> >               "there will be at least one service that not all peers
> > will offer, and thus a search mechanism is needed".
> >
> > - Philip
> >
> _______________________________________________
> p2p-sip mailing list
> p2p-sip at cs.columbia.edu
> https://lists.cs.columbia.edu/cucslists/listinfo/p2p-sip



From jiang.x.f at huawei.com  Tue Dec 12 04:46:19 2006
From: jiang.x.f at huawei.com (JiangXingFeng)
Date: Tue, 12 Dec 2006 17:46:19 +0800
Subject: [p2p-sip] Concepts draft: Open Issue #1: Searching for services
In-Reply-To: <45799EF4.3050702@telecomitalia.it>
Message-ID: <000701c71dd2$602a21a0$7205a40a@china.huawei.com>

Hi, Enrico:
  If I do not mistake your points, you mean peer X is exposed to DOS
attacks, which will make X refuse to serve other peers/clients. I agree with
you. Do you have a mature idea of this problem: how to put service resources
into Overlay and avoid to be a weak point of the system?
Regards!
--
Jiang XingFeng
> -----Original Message-----
> From: p2p-sip-bounces at cs.columbia.edu
> [mailto:p2p-sip-bounces at cs.columbia.edu] On Behalf Of Enrico Marocco
> Sent: Saturday, December 09, 2006 1:21 AM
> To: Philip Matthews
> Cc: P2P-SIP
> Subject: Re: [p2p-sip] Concepts draft: Open Issue #1: Searching for
services
> 
> Philip Matthews wrote:
> > Say I implement service lookup as follows. For each service, there is a
> > unique URI for the service (e.g., "STUNServers@<Overlay Identifier>" or
> > whatever). The information associated with this URI is stored in a
> > single peer in the Overlay (ignoring replicas for robustness). Say that
> > peer is X. Then all the peers that offer the service "register" with X
> > saying that they offer the service.  When peer A does a lookup on
> > "STUNServers@<Overlay Identifier>", then X returns one or more peers
> > supporting the service. Peer X might return just three peers for each
> > lookup query, and randomly select those peers each time, to make sure
> > that no one peer is overloaded.
> >
> > Is your concern that many peers would want STUN services, and thus there
> > would be many lookups on "STUNServers@<Overlay Identifier>" and thus
> > peer X would get overloaded?  Or are you concerned about the peers
> > offering the service?
> 
> My concern is the former; in the case you describe, node X, not only
> would get overloaded -- and that's already a good reason for being
> careful in the decision -- but also, having a more important role, would
> be the preferred target of attacks and, in general, a weak point of the
> system.
> 
> --
> Ciao,
> Enrico
> _______________________________________________
> p2p-sip mailing list
> p2p-sip at cs.columbia.edu
> https://lists.cs.columbia.edu/cucslists/listinfo/p2p-sip



From enrico.marocco at telecomitalia.it  Tue Dec 12 06:04:39 2006
From: enrico.marocco at telecomitalia.it (Enrico Marocco)
Date: Tue, 12 Dec 2006 12:04:39 +0100
Subject: [p2p-sip] Concepts draft: Open Issue #1: Searching for services
In-Reply-To: <000701c71dd2$602a21a0$7205a40a@china.huawei.com>
References: <000701c71dd2$602a21a0$7205a40a@china.huawei.com>
Message-ID: <457E8CC7.1010104@telecomitalia.it>

JiangXingFeng wrote:
>   If I do not mistake your points, you mean peer X is exposed to DOS
> attacks, which will make X refuse to serve other peers/clients. I agree with
> you. Do you have a mature idea of this problem: how to put service resources
> into Overlay and avoid to be a weak point of the system?

Peer X would have a more important role than almost any other peer in
the overlay; this leads to think that it would be the preferred target
of any potential attacker.  Moreover, it would also be subject to a load
much higher than the average.

A possible workaround would be to simply ask peers to cache most popular
values, but such a mechanism needs to deal with issues like expiration
and error detection.

A more complex solution would be to adopt some linear replication
mechanism for "special" URIs, so that any node could store the location
of a minimum number of instances and be notified of their departures.
For example, say the minimum number of stun server locations to be
stored in any peer is 2, consider an overlay formed by 6 peers (A, B, C,
D, E and F in sequence) and 4 stun servers (S1, S2, S3 and S4),
distributed as follow:

A: S1
B:
C: S2
D: S3
E: S4
F:

After the first replication cycle, the situation would be the following
(where values for which the node is authoritative are marked):

A: S1*, S4
B: S1, S4
C: S2*, S1
D: S3*, S2
E: S4*, S3
F: S4, S3

The following flow shows what happens when peer A detects S1's failure:

    A        B        C        D        E        F
    |        |        |        |        |        |
    | S?     |        |        |        |        |
    |------------------------------------------->|
    |        |        |        |        | S4, S3 |
    |<-------------------------------------------|
    | S4, S3 |        |        |        |        |
    |------->| S4, S3 |        |        |        |
    |        |------->| S2*, S4|        |        |
    |        |        |------->|        |        |
    |        |        |        |        |        |

And the resulting situation would be:

A: S4, S3
B: S4, S3
C: S2*, S4
D: S3*, S2 (unchanged)
E: S4*, S3 (unchanged)
F: S4, S3 (unchanged)

Note that with such a mechanism the peers are not authoritative on the
URL of the "stun service", but on a specific instance identifier (IP,
contact?) and the complexity isn't logarithmic any more.

-- 
Ciao,
Enrico

From RADHIKA.R.ROY at saic.com  Tue Dec 12 10:46:38 2006
From: RADHIKA.R.ROY at saic.com (Roy, Radhika R.)
Date: Tue, 12 Dec 2006 10:46:38 -0500
Subject: [p2p-sip] Concepts draft: Open Issue #1: Searching for services
In-Reply-To: <457E8CC7.1010104@telecomitalia.it>
Message-ID: <62D92A9A02BCC845B202323D49A48426230F11@0591-ITS-EXMP02.us.saic.com>

Enrico and JiangXingFeng:

Enrico is right.

The bottom line is this: Services will also need to be distributed in
P2P system.

Best regards,
Radhika

PS: 

1. Eventually, SIP endpoints will have distributive services
capabilities in SIP-P2P system. Let us differ those discussions once the
basic P2P-SIP is standardized.

2. Henry's email points to the fact what is coming ahead of us in the
P2P-Internet breaking all business models for the so-called centralized
services.

3. When people will get all those services in mobile ad hoc networks
(MANETs), nothing will exist but only P2P systems.

-----Original Message-----
From: p2p-sip-bounces at cs.columbia.edu
[mailto:p2p-sip-bounces at cs.columbia.edu] On Behalf Of Enrico Marocco
Sent: Tuesday, December 12, 2006 6:05 AM
To: JiangXingFeng
Cc: 'P2P-SIP'
Subject: Re: [p2p-sip] Concepts draft: Open Issue #1: Searching for
services

JiangXingFeng wrote:
>   If I do not mistake your points, you mean peer X is exposed to DOS
> attacks, which will make X refuse to serve other peers/clients. I
agree with
> you. Do you have a mature idea of this problem: how to put service
resources
> into Overlay and avoid to be a weak point of the system?

Peer X would have a more important role than almost any other peer in
the overlay; this leads to think that it would be the preferred target
of any potential attacker.  Moreover, it would also be subject to a load
much higher than the average.

A possible workaround would be to simply ask peers to cache most popular
values, but such a mechanism needs to deal with issues like expiration
and error detection.

A more complex solution would be to adopt some linear replication
mechanism for "special" URIs, so that any node could store the location
of a minimum number of instances and be notified of their departures.
For example, say the minimum number of stun server locations to be
stored in any peer is 2, consider an overlay formed by 6 peers (A, B, C,
D, E and F in sequence) and 4 stun servers (S1, S2, S3 and S4),
distributed as follow:

A: S1
B:
C: S2
D: S3
E: S4
F:

After the first replication cycle, the situation would be the following
(where values for which the node is authoritative are marked):

A: S1*, S4
B: S1, S4
C: S2*, S1
D: S3*, S2
E: S4*, S3
F: S4, S3

The following flow shows what happens when peer A detects S1's failure:

    A        B        C        D        E        F
    |        |        |        |        |        |
    | S?     |        |        |        |        |
    |------------------------------------------->|
    |        |        |        |        | S4, S3 |
    |<-------------------------------------------|
    | S4, S3 |        |        |        |        |
    |------->| S4, S3 |        |        |        |
    |        |------->| S2*, S4|        |        |
    |        |        |------->|        |        |
    |        |        |        |        |        |

And the resulting situation would be:

A: S4, S3
B: S4, S3
C: S2*, S4
D: S3*, S2 (unchanged)
E: S4*, S3 (unchanged)
F: S4, S3 (unchanged)

Note that with such a mechanism the peers are not authoritative on the
URL of the "stun service", but on a specific instance identifier (IP,
contact?) and the complexity isn't logarithmic any more.

-- 
Ciao,
Enrico


From michael.slavitch at e2exchange.com  Tue Dec 12 11:10:02 2006
From: michael.slavitch at e2exchange.com (Michael Slavitch)
Date: Tue, 12 Dec 2006 11:10:02 -0500
Subject: [p2p-sip] First Rule of Distributed Computing:
In-Reply-To: <db25d5be0612120801x69de0b96r912869137ddbff16@mail.gmail.com>
References: <f40963db0612110711r40fa3281we5e061a57fa9cae8@mail.gmail.com>
	<24CCCC428EFEA2469BF046DB3C7A8D221413FE@namail5.corp.adobe.com>
	<1fc32c8f0612120725i4f0ba819t99640d5fdf04151e@mail.gmail.com>
	<db25d5be0612120801x69de0b96r912869137ddbff16@mail.gmail.com>
Message-ID: <db25d5be0612120810t3d79723cvd9f472e7026e8414@mail.gmail.com>

There has to be really good reasons to choose the centralized approach.

Historically, strict centralization works best only when all resources are
scarce, for everything from food distribution to decision making to
economics.  Centralized food distribution goes back to Old Testament times.
The old phrase 'seven years of bumper crop, seven years of famine' might as
well apply to the technology industry with 2001 being the start of famine,
although it may be the telecoms that end up suffering Pharaoh's fate. They
should well consider the consumer's Technicolor Dream-Coat.

Central planning always fails when there is abundance of resource supply,
broad education  and free access to data.  Central planning  fails because
with abundance the central planners can be ignored.  Given that, the first
act of central planners when faced with a revolt of abundance is and always
has been repression and more centralization.

This has been demonstrated over the centuries.

Some successful revolts include the signing of the Magna Carta, the
Reformation,  the Renaissance, and the American Revolution.

In communications terms that analogy relates to long term store, CPU and
bandwidth.

In political terms it looks like a counter-revolution is underway in
Russia.  The telecoms industry is mounting one in the United States.

I suspect the latter will fail.

Classical distributed systems that are used as a solution to scarcity are
inappropriate for comparison to P2P systems that designed to accommodate and
self organize based on resource feast and resource famine. It may be a
mistake to compare the problems seen in distributed systems with the
opportunity in peer to peer systems, especially  if those distributed system
are designed only for scarcity and not abundance.

Scarcity is not a constant. It is now an occasional condition that must be
accommodated and respected. It is no longer the norm.

This means that distributed solutions designed primarily to defeat scarcity
without taking into consideration abundance are inappropriate.
A global calling database for every human being in the world could easily
fit on a PC hard disk, the updates being merely the deltas over time.   A
contact list containing every person in the US can easily fit on a flash
card.   That alone would be an interesting product for smart-phones. Every
sales rep would want one.  A nightly update over WiLAN can update its
database to five-sigma accuracy without cost to the user.  The sixth sigma,
done over an expensive 3G network, will cost a few pennies once in a while
over a dumb network. In economic terms that is a rational system, a rational
consumer will always gravitate towards such a system given a choice.

The search problem is eliminated when distributed systems are not completely
distributed but instead hyper-redundant.

Every hand held smart phone can now contain a class-5 switch database. There
is enough store. There certainly is enough CPU for the call processing. If
every smart phone can be a class-5 switch why do you need centralized
class-5 switches?

I agree that the search problem is a problem for older P2P systems, like
those used for file sharing. By definition a database of stolen goods is
always unstable. That is not true for identities or other legal attributes.

If search is a problem, eliminate search as a problem, and limit the churn
to updates.  The research says that you'll reduce lookups by many orders of
magnitude.

Regards

Michael

Please note the new email and domain, www.e2exchange.com


On 12/12/06, Michael Slavitch <slavitch at gmail.com> wrote:
>
> ---------- Forwarded message ----------
> From: Henry Sinnreich <hsinnrei at adobe.com>
> Date: Dec 11, 2006 10:50 PM
> Subject: Re: [p2p-sip] First Rule of Distributed Computing:
> To: Adam Fisk < adamfisk at gmail.com>
> Cc: p2p-sip at cs.columbia.edu
>
>
> Adam Fisk wrote:
>
> >There have to be really, really good reasons to choose the first,
> distributed approach.
>
> Right. Here are some reasons in the case of Client-Server SIP VoIP.
> There are about 10-100 VoIP network elements in larger VoIP networks,
> such as:
>
> * DNS, ENUM servers with VoIP data entries
> * SIP registrar and proxy servers, fronting SIP proxies
> * Presence servers
> * Component servers:
>   - Announcements, voice mail, media servers, conference servers
> * Session border controllers (at several points in the network)
> * Softswitches: Class 5 switch emulation and/or hosted IP PBXs (Centrex)
> * Media servers
> * Network elements for QoS
> * Policy servers
> * Network management
> * Voice quality monitoring probes
> * Network engineering
> * System integration
> * VoIP network IT (OSS) systems for all the above
> * IT systems provisioned with churning user data for most of the above
> - this is the most onerous part.
> Note that I did not count TDM gateways since gateways would be needed
> for P2P as well.
>
> This is a nightmare of an infrastructure, since PCs at the endpoints
> are more than powerful enough to provide these functions and also to
> move the cost of hosting and bandwidth to the endpoints. There is no
> P2P VoIP infrastructure and P2P systems are self organizing.
>
> Skype with its >100M users has proven in the market that P2P avoids
> this complexity and cost (staffed operations 7x24, staffed custom
> system integration, huge IT systems development, etc.) and most of
> all, lack of flexibility can be mostly avoided and still provide rich
> consumer presence, IM, VoIP, video, file transfer and now also
> conferencing and various enterprise applications.
>
> I don't have profit numbers for Skype (they had a good business model
> based on prepaid float I believe before becoming eBay), but I have not
> seen any profit statements from any VoIP company that I know of,
> except when VoIP is bundled with broadband Internet or with something
> else. Does the bundling hide the losses for VoIP service?
>
> VoIP is just an application sitting in the endpoints and that's how it
> should be on "the dumb network".
> Since when do we need the "Intelligent Network" on the Internet (the
> IETF's task)?
> This is the rationale for developing standards for P2P SIP.
>
> Leaving VoIP aside, there are roughly 5,000 papers that deal with P2P
> and for what it is worth, p2p traffic dominates the Internet.
> Obviously for good reasons.
>
> Thanks, Henry
> ________________________________________
> From: p2p-sip-bounces at cs.columbia.edu
> [mailto:p2p-sip-bounces at cs.columbia.edu] On Behalf Of Adam Fisk
> Sent: Monday, December 11, 2006 9:11 AM
> To: David Barrett
> Cc: p2p-sip at cs.columbia.edu
> Subject: Re: [p2p-sip] First Rule of Distributed Computing:
>
> The example I encounter most frequently is search. Distributed search
> is hard. It's hard to search a lot of computers without taking up a
> lot of bandwidth, not to mention the algorithms are just tough and
> require massive coordination amongst many machines. The bandwidth
> constraints lead to architectures that increase latency even if
> they're scalable, like DHTs or distributed indexing with random
> walk-type approaches on top of them. The bandwidth constraints and
> latency also put a constraint on message sizes, significantly limiting
> your options for whatever protocol you're working with. Updates are
> also hard. Instead of updating a server cluster, you have to update
> sometimes millions of machines.
>
> With centralized search, these constraints largely dissappear. You get
> low latency, global search, all with much less work. Things like
> collaborative filtering and tagging also become possible, whereas they
> are extremely complicated at best with distributed systems. Sure
> there's a single point of failure, but that's rarely an issue in
> practice if you know what you're doing.
>
> There have to be really, really good reasons to choose the first,
> distributed approach. In the case of the file sharing networks, that
> reason was legal liability. In other cases, it might be cost.
>
> That's just a rule I frequently have to remind myself of, and it's
> saved me a great deal of time in the past.
>
> -Adam
>
> On 12/10/06, David Barrett <dbarrett at quinthar.com> wrote:
> Heh, this seems like very good advice, but I'm not sure I follow your
> meaning. Can you give us an example?
>
> -david
>
> ________________________________________
> From: p2p-sip-bounces at cs.columbia.edu [mailto:
> p2p-sip-bounces at cs.columbia.edu ] On Behalf Of Adam Fisk
> Sent: Sunday, December 10, 2006 1:51 PM
> To: p2p-sip at cs.columbia.edu
> Subject: [p2p-sip] First Rule of Distributed Computing:
>
> Don't distribute your objects.
>
> No specific point to be made here, but I find that rule useful in my
> own coding and thinking, and it's surprisingly easy to forget. If you
> can avoid distributing your objects, you're usually left with a much
> easier system to implement and maintain. Is that Radia Perlman? Not
> sure.
>
> -Adam
>
>
> _______________________________________________
> p2p-sip mailing list
> p2p-sip at cs.columbia.edu
> https://lists.cs.columbia.edu/cucslists/listinfo/p2p-sip
-------------- next part --------------
An HTML attachment was scrubbed...
URL: https://lists.cs.columbia.edu/pipermail/p2p-sip/attachments/20061212/a962fca8/attachment-0001.html

From RADHIKA.R.ROY at saic.com  Tue Dec 12 11:25:42 2006
From: RADHIKA.R.ROY at saic.com (Roy, Radhika R.)
Date: Tue, 12 Dec 2006 11:25:42 -0500
Subject: [p2p-sip] First Rule of Distributed Computing:
In-Reply-To: <db25d5be0612120810t3d79723cvd9f472e7026e8414@mail.gmail.com>
Message-ID: <62D92A9A02BCC845B202323D49A48426230F13@0591-ITS-EXMP02.us.saic.com>

It depends what the basic assumptions are. It is true that the
centralized systems will have their uses as it has been pointed by you.
There is no debate on this point.

 

For example, all telcos have their existing centralized systems and
there is no need to change those for them as those business models will
be working for them for many foreseeable years.

 

However, companies like Skype/eBay did not have those luxuries:
P2P-system is a good business model for them.

 

The lesson is this: P2P systems are opening the businesses to millions
of people who can start with nothing and still can have most of it
(i.e., as if, opening the business models for creation of wealth by a
community of large-scale masses).

 

Best regards,

Radhika

 

________________________________

From: p2p-sip-bounces at cs.columbia.edu
[mailto:p2p-sip-bounces at cs.columbia.edu] On Behalf Of Michael Slavitch
Sent: Tuesday, December 12, 2006 11:10 AM
To: p2p-sip at cs.columbia.edu
Subject: Re: [p2p-sip] First Rule of Distributed Computing:

 

There has to be really good reasons to choose the centralized approach.

Historically, strict centralization works best only when all resources
are scarce, for everything from food distribution to decision making to
economics.  Centralized food distribution goes back to Old Testament
times.  The old phrase 'seven years of bumper crop, seven years of
famine' might as well apply to the technology industry with 2001 being
the start of famine, although it may be the telecoms that end up
suffering Pharaoh's fate. They should well consider the consumer's
Technicolor Dream-Coat.  

Central planning always fails when there is abundance of resource
supply,  broad education  and free access to data.  Central planning
fails because with abundance the central planners can be ignored.  Given
that, the first act of central planners when faced with a revolt of
abundance is and always has been repression and more centralization. 

This has been demonstrated over the centuries.  

Some successful revolts include the signing of the Magna Carta, the
Reformation,  the Renaissance, and the American Revolution. 

In communications terms that analogy relates to long term store, CPU and
bandwidth. 

In political terms it looks like a counter-revolution is underway in
Russia.  The telecoms industry is mounting one in the United States. 

I suspect the latter will fail.

Classical distributed systems that are used as a solution to scarcity
are inappropriate for comparison to P2P systems that designed to
accommodate and self organize based on resource feast and resource
famine. It may be a mistake to compare the problems seen in distributed
systems with the opportunity in peer to peer systems, especially  if
those distributed system are designed only for scarcity and not
abundance.  

Scarcity is not a constant. It is now an occasional condition that must
be accommodated and respected. It is no longer the norm.

This means that distributed solutions designed primarily to defeat
scarcity without taking into consideration abundance are inappropriate.

A global calling database for every human being in the world could
easily fit on a PC hard disk, the updates being merely the deltas over
time.   A contact list containing every person in the US can easily fit
on a flash card.   That alone would be an interesting product for
smart-phones. Every sales rep would want one.  A nightly update over
WiLAN can update its database to five-sigma accuracy without cost to the
user.  The sixth sigma, done over an expensive 3G network, will cost a
few pennies once in a while over a dumb network. In economic terms that
is a rational system, a rational consumer will always gravitate towards
such a system given a choice. 

The search problem is eliminated when distributed systems are not
completely distributed but instead hyper-redundant.  

Every hand held smart phone can now contain a class-5 switch database.
There is enough store. There certainly is enough CPU for the call
processing. If every smart phone can be a class-5 switch why do you need
centralized class-5 switches?

I agree that the search problem is a problem for older P2P systems, like
those used for file sharing. By definition a database of stolen goods is
always unstable. That is not true for identities or other legal
attributes. 

If search is a problem, eliminate search as a problem, and limit the
churn to updates.  The research says that you'll reduce lookups by many
orders of magnitude. 

Regards

Michael

Please note the new email and domain, www.e2exchange.com
<http://www.e2exchange.com/> 



On 12/12/06, Michael Slavitch <slavitch at gmail.com> wrote:

---------- Forwarded message ---------- 
From: Henry Sinnreich <hsinnrei at adobe.com>
Date: Dec 11, 2006 10:50 PM
Subject: Re: [p2p-sip] First Rule of Distributed Computing: 
To: Adam Fisk < adamfisk at gmail.com <mailto:adamfisk at gmail.com> >
Cc: p2p-sip at cs.columbia.edu


Adam Fisk wrote:

>There have to be really, really good reasons to choose the first, 
distributed approach. 

Right. Here are some reasons in the case of Client-Server SIP VoIP.
There are about 10-100 VoIP network elements in larger VoIP networks,
such as:

* DNS, ENUM servers with VoIP data entries
* SIP registrar and proxy servers, fronting SIP proxies 
* Presence servers
* Component servers:
  - Announcements, voice mail, media servers, conference servers
* Session border controllers (at several points in the network)
* Softswitches: Class 5 switch emulation and/or hosted IP PBXs (Centrex)

* Media servers
* Network elements for QoS
* Policy servers
* Network management
* Voice quality monitoring probes
* Network engineering
* System integration
* VoIP network IT (OSS) systems for all the above 
* IT systems provisioned with churning user data for most of the above
- this is the most onerous part.
Note that I did not count TDM gateways since gateways would be needed
for P2P as well.

This is a nightmare of an infrastructure, since PCs at the endpoints 
are more than powerful enough to provide these functions and also to
move the cost of hosting and bandwidth to the endpoints. There is no
P2P VoIP infrastructure and P2P systems are self organizing.

Skype with its >100M users has proven in the market that P2P avoids 
this complexity and cost (staffed operations 7x24, staffed custom
system integration, huge IT systems development, etc.) and most of
all, lack of flexibility can be mostly avoided and still provide rich
consumer presence, IM, VoIP, video, file transfer and now also 
conferencing and various enterprise applications.

I don't have profit numbers for Skype (they had a good business model
based on prepaid float I believe before becoming eBay), but I have not
seen any profit statements from any VoIP company that I know of, 
except when VoIP is bundled with broadband Internet or with something
else. Does the bundling hide the losses for VoIP service?

VoIP is just an application sitting in the endpoints and that's how it
should be on "the dumb network". 
Since when do we need the "Intelligent Network" on the Internet (the
IETF's task)?
This is the rationale for developing standards for P2P SIP.

Leaving VoIP aside, there are roughly 5,000 papers that deal with P2P 
and for what it is worth, p2p traffic dominates the Internet.
Obviously for good reasons.

Thanks, Henry
________________________________________
From: p2p-sip-bounces at cs.columbia.edu 
[mailto:p2p-sip-bounces at cs.columbia.edu] On Behalf Of Adam Fisk
Sent: Monday, December 11, 2006 9:11 AM 
To: David Barrett
Cc: p2p-sip at cs.columbia.edu
Subject: Re: [p2p-sip] First Rule of Distributed Computing:

The example I encounter most frequently is search. Distributed search
is hard. It's hard to search a lot of computers without taking up a 
lot of bandwidth, not to mention the algorithms are just tough and
require massive coordination amongst many machines. The bandwidth
constraints lead to architectures that increase latency even if
they're scalable, like DHTs or distributed indexing with random 
walk-type approaches on top of them. The bandwidth constraints and
latency also put a constraint on message sizes, significantly limiting
your options for whatever protocol you're working with. Updates are
also hard. Instead of updating a server cluster, you have to update 
sometimes millions of machines.

With centralized search, these constraints largely dissappear. You get
low latency, global search, all with much less work. Things like
collaborative filtering and tagging also become possible, whereas they 
are extremely complicated at best with distributed systems. Sure
there's a single point of failure, but that's rarely an issue in
practice if you know what you're doing.

There have to be really, really good reasons to choose the first, 
distributed approach. In the case of the file sharing networks, that
reason was legal liability. In other cases, it might be cost.

That's just a rule I frequently have to remind myself of, and it's
saved me a great deal of time in the past. 

-Adam

On 12/10/06, David Barrett <dbarrett at quinthar.com> wrote:
Heh, this seems like very good advice, but I'm not sure I follow your 
meaning. Can you give us an example? 

-david

________________________________________
From: p2p-sip-bounces at cs.columbia.edu [mailto:
p2p-sip-bounces at cs.columbia.edu ] On Behalf Of Adam Fisk
Sent: Sunday, December 10, 2006 1:51 PM
To: p2p-sip at cs.columbia.edu
Subject: [p2p-sip] First Rule of Distributed Computing:

Don't distribute your objects. 

No specific point to be made here, but I find that rule useful in my
own coding and thinking, and it's surprisingly easy to forget. If you
can avoid distributing your objects, you're usually left with a much 
easier system to implement and maintain. Is that Radia Perlman? Not
sure.

-Adam


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

-------------- next part --------------
An HTML attachment was scrubbed...
URL: https://lists.cs.columbia.edu/pipermail/p2p-sip/attachments/20061212/e3030698/attachment.html

From hgs at cs.columbia.edu  Wed Dec 13 17:22:04 2006
From: hgs at cs.columbia.edu (Henning Schulzrinne)
Date: Wed, 13 Dec 2006 17:22:04 -0500
Subject: [p2p-sip] Test
Message-ID: <4E1831CC-11B8-440A-8345-7FA1D428CAF5@cs.columbia.edu>

Test only - please ignore.

From warlock at media.mit.edu  Fri Dec 15 00:50:06 2006
From: warlock at media.mit.edu (Jacky Mallett)
Date: Fri, 15 Dec 2006 00:50:06 -0500
Subject: [p2p-sip] First Rule of Distributed Computing:
In-Reply-To: <24CCCC428EFEA2469BF046DB3C7A8D221413FE@namail5.corp.adobe.com>
References: <24CCCC428EFEA2469BF046DB3C7A8D221413FE@namail5.corp.adobe.com>
Message-ID: <A4E88481-CCD2-4DE2-85E7-0B09C6609E0D@media.mit.edu>


On Dec 11, 2006, at 10:50 PM, Henry Sinnreich wrote:

> Adam Fisk wrote:
>
>> There have to be really, really good reasons to choose the first,  
>> distributed approach.
>
>
> Skype with its >100M users has proven in the market that P2P avoids  
> this complexity and cost (staffed operations 7x24, staffed custom  
> system integration, huge IT systems development, etc.) and most of  
> all, lack of flexibility can be mostly avoided and still provide  
> rich consumer presence, IM, VoIP, video, file transfer and now also  
> conferencing and various enterprise applications.

I've seen you say this a few times, so i'd like to challenge a couple  
of things about it. First of all - Skype may have 100M users in their  
(centralised) databases. The most i've ever seen online is about 8.5  
million.  When talking about scaling issues within peer-to-peer  
applications, it's the 8.5 million figure that's significant. And the  
proper comparison would be the number of concurrent users of the  
global phone system.

In addition to the centralised login database, it appears there are  
also at least 7 bootstrap nodes with statically assigned IP addresses.

http://skype.nexxen.cz/dokumentace/public/SkypeProtocol/html/ 
SkypeProtocol_4.htm

It is correct to say that the SIP protocol as implemented is the  
endpoint application. The distributed _system_ that uses it and other  
protocols to interconnect users, is not. It has network like  
properties all of its own.  Network in this sense is used to mean  
systems with organised communications between large numbers of nodes.

....jacky


>
> I don't have profit numbers for Skype (they had a good business  
> model based on prepaid float I believe before becoming eBay), but I  
> have not seen any profit statements from any VoIP company that I  
> know of, except when VoIP is bundled with broadband Internet or  
> with something else. Does the bundling hide the losses for VoIP  
> service?
>
> VoIP is just an application sitting in the endpoints and that's how  
> it should be on "the dumb network".
> Since when do we need the "Intelligent Network" on the Internet  
> (the IETF's task)?
> This is the rationale for developing standards for P2P SIP.
>
> Leaving VoIP aside, there are roughly 5,000 papers that deal with  
> P2P and for what it is worth, p2p traffic dominates the Internet.  
> Obviously for good reasons.
>
> Thanks, Henry
> ________________________________________
> From: p2p-sip-bounces at cs.columbia.edu [mailto:p2p-sip- 
> bounces at cs.columbia.edu] On Behalf Of Adam Fisk
> Sent: Monday, December 11, 2006 9:11 AM
> To: David Barrett
> Cc: p2p-sip at cs.columbia.edu
> Subject: Re: [p2p-sip] First Rule of Distributed Computing:
>
> The example I encounter most frequently is search.  Distributed  
> search is hard.  It's hard to search a lot of computers without  
> taking up a lot of bandwidth, not to mention the algorithms are  
> just tough and require massive coordination amongst many machines.   
> The bandwidth constraints lead to architectures that increase  
> latency even if they're scalable, like DHTs or distributed indexing  
> with random walk-type approaches on top of them.  The bandwidth  
> constraints and latency also put a constraint on message sizes,  
> significantly limiting your options for whatever protocol you're  
> working with.  Updates are also hard.  Instead of updating a server  
> cluster, you have to update sometimes millions of machines.
>
> With centralized search, these constraints largely dissappear.  You  
> get low latency, global search, all with much less work.  Things  
> like collaborative filtering and tagging also become possible,  
> whereas they are extremely complicated at best with distributed  
> systems.  Sure there's a single point of failure, but that's rarely  
> an issue in practice if you know what you're doing.
>
> There have to be really, really good reasons to choose the first,  
> distributed approach.  In the case of the file sharing networks,  
> that reason was legal liability.  In other cases, it might be cost.
>
> That's just a rule I frequently have to remind myself of, and it's  
> saved me a great deal of time in the past.
>
> -Adam
>
> On 12/10/06, David Barrett <dbarrett at quinthar.com> wrote:
> Heh, this seems like very good advice, but I'm not sure I follow  
> your meaning.  Can you give us an example?
>
> -david
>
> ________________________________________
> From: p2p-sip-bounces at cs.columbia.edu [mailto: p2p-sip- 
> bounces at cs.columbia.edu] On Behalf Of Adam Fisk
> Sent: Sunday, December 10, 2006 1:51 PM
> To: p2p-sip at cs.columbia.edu
> Subject: [p2p-sip] First Rule of Distributed Computing:
>
> Don't distribute your objects.
>
> No specific point to be made here, but I find that rule useful in  
> my own coding and thinking, and it's surprisingly easy to forget.   
> If you can avoid distributing your objects, you're usually left  
> with a much easier system to implement and maintain.  Is that Radia  
> Perlman?  Not sure.
>
> -Adam
>
>
> _______________________________________________
> p2p-sip mailing list
> p2p-sip at cs.columbia.edu
> https://lists.cs.columbia.edu/cucslists/listinfo/p2p-sip


From br at brianrosen.net  Fri Dec 15 09:43:05 2006
From: br at brianrosen.net (Brian Rosen)
Date: Fri, 15 Dec 2006 09:43:05 -0500
Subject: [p2p-sip] Concepts draft: Open Issue #3: Enrollment
	as	aPeer/Client Protocol operation?
In-Reply-To: <457D5CE4.1070708@ericsson.com>
Message-ID: <18c501c72057$5809cf70$640fa8c0@cis.neustar.com>

I don't think we need to standardize the method of obtaining credentials.

I believe we agreed we need to standardize at least one mandatory to
implement FORM of credentials, so we could use credentials obtained by an
unspecified mechanism

Brian

> -----Original Message-----
> From: p2p-sip-bounces at cs.columbia.edu [mailto:p2p-sip-
> bounces at cs.columbia.edu] On Behalf Of Jani Hautakorpi (JO/LMF)
> Sent: Monday, December 11, 2006 8:28 AM
> To: Spencer Dawkins
> Cc: P2P-SIP
> Subject: Re: [p2p-sip] Concepts draft: Open Issue #3: Enrollment as
> aPeer/Client Protocol operation?
> 
> Hi all,
> 
> Spencer Dawkins wrote:
> > I liked David's additional statement, which was
> >
> > "- if we DO standardize, it could be this is useful to others outside
> >    P2PSIP and might even be some work we push into another WG."
> >
> > so, taking David's comment and your comment above into account, my list
> ends
> > up as
> >
> > - we don't HAVE to specify the method of obtaining credentials now, and
> >
> > - if we DO standardize the method of obtaining credentials, it may not
> >   be part of the peer and/or client protocol,
> >
> > - and might developed in another working group, if the resulting
> >   standardized method would be useful outside P2PSIP.
> >
> > Does this work for people?
> 
> This makes sense and sounds good at least to me.
> 
> 
> --
> Jani
> _______________________________________________
> p2p-sip mailing list
> p2p-sip at cs.columbia.edu
> https://lists.cs.columbia.edu/cucslists/listinfo/p2p-sip


From hannes.tschofenig at siemens.com  Fri Dec 15 10:14:55 2006
From: hannes.tschofenig at siemens.com (Tschofenig, Hannes)
Date: Fri, 15 Dec 2006 16:14:55 +0100
Subject: [p2p-sip] Concepts draft: Open Issue #3:
	Enrollmentas	aPeer/Client Protocol operation?
In-Reply-To: <18c501c72057$5809cf70$640fa8c0@cis.neustar.com>
Message-ID: <6F0CB04509C11D46A54232E852E390AC0240F67F@MCHP7R6A.ww002.siemens.net>

Hi Brian, 

I personally think we need a way to obtain credentials -- not only for P2P SIP. Obviously, this would not be done in the P2P SIP working group. Hence, the approach suggested by you seems to be reasonable. 

Previous attemps to get credential enrollment work started in the IETF failed (e.g., EasyCert BoF). 
I contribute my thoughts to a working group called ENROLL. Here is an example of the type of documents I wrote: http://www.tschofenig.com/drafts/draft-tschofenig-enroll-next-steps-00.txt. The group also failed. 

Ciao
Hannes

-----Urspr?ngliche Nachricht-----
Von: p2p-sip-bounces at cs.columbia.edu [mailto:p2p-sip-bounces at cs.columbia.edu] Im Auftrag von Brian Rosen
Gesendet: Freitag, 15. Dezember 2006 15:43
An: 'Jani Hautakorpi (JO/LMF)'; 'Spencer Dawkins'
Cc: 'P2P-SIP'
Betreff: Re: [p2p-sip] Concepts draft: Open Issue #3: Enrollmentas aPeer/Client Protocol operation?

I don't think we need to standardize the method of obtaining credentials.

I believe we agreed we need to standardize at least one mandatory to
implement FORM of credentials, so we could use credentials obtained by an
unspecified mechanism

Brian

> -----Original Message-----
> From: p2p-sip-bounces at cs.columbia.edu [mailto:p2p-sip-
> bounces at cs.columbia.edu] On Behalf Of Jani Hautakorpi (JO/LMF)
> Sent: Monday, December 11, 2006 8:28 AM
> To: Spencer Dawkins
> Cc: P2P-SIP
> Subject: Re: [p2p-sip] Concepts draft: Open Issue #3: Enrollment as
> aPeer/Client Protocol operation?
> 
> Hi all,
> 
> Spencer Dawkins wrote:
> > I liked David's additional statement, which was
> >
> > "- if we DO standardize, it could be this is useful to others outside
> >    P2PSIP and might even be some work we push into another WG."
> >
> > so, taking David's comment and your comment above into account, my list
> ends
> > up as
> >
> > - we don't HAVE to specify the method of obtaining credentials now, and
> >
> > - if we DO standardize the method of obtaining credentials, it may not
> >   be part of the peer and/or client protocol,
> >
> > - and might developed in another working group, if the resulting
> >   standardized method would be useful outside P2PSIP.
> >
> > Does this work for people?
> 
> This makes sense and sounds good at least to me.
> 
> 
> --
> Jani
> _______________________________________________
> p2p-sip mailing list
> p2p-sip at cs.columbia.edu
> https://lists.cs.columbia.edu/cucslists/listinfo/p2p-sip

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


From br at brianrosen.net  Fri Dec 15 10:37:19 2006
From: br at brianrosen.net (Brian Rosen)
Date: Fri, 15 Dec 2006 10:37:19 -0500
Subject: [p2p-sip] First Rule of Distributed Computing:
In-Reply-To: <A4E88481-CCD2-4DE2-85E7-0B09C6609E0D@media.mit.edu>
Message-ID: <191b01c7205e$ef195d20$640fa8c0@cis.neustar.com>

Jacky

> I've seen you say this a few times, so i'd like to challenge a couple
> of things about it. First of all - Skype may have 100M users in their
> (centralised) databases. The most i've ever seen online is about 8.5
> million.  When talking about scaling issues within peer-to-peer
> applications, it's the 8.5 million figure that's significant. And the
> proper comparison would be the number of concurrent users of the
> global phone system.
I'm not sure what "concurrent users" means.  The 8.5M number is the number
of subscribers on-line, not the number talking.  It's equivalent to a
wireless phone connected to a network and able to, or actively placing or
receiving a call.  I haven't seen numbers like that for wireless systems.
What you see is "subscribers", which is the big number for Skype.  It's
clearly true that the ratio of subscribers to devices on line and available
for or in a call is much higher for wireless systems than it is for Skype.

Also, the numbers are for a single carrier, not the entire PSTN.  Carriers
have interconnects, as will P2P-SIP domains.

> 
> In addition to the centralised login database, it appears there are
> also at least 7 bootstrap nodes with statically assigned IP addresses.
> 
> http://skype.nexxen.cz/dokumentace/public/SkypeProtocol/html/
> SkypeProtocol_4.htm
Yes, the bootstrap uses fixed addresses.  Once you have contacted a peer,
you don't need them any more.  We will need some kind of a bootstrap
mechanism, and I think while we won't prohibit a fixed set of IP addresses,
I think we're probably going to have other mechanisms as well.

> 
> It is correct to say that the SIP protocol as implemented is the
> endpoint application. The distributed _system_ that uses it and other
> protocols to interconnect users, is not. It has network like
> properties all of its own.  Network in this sense is used to mean
> systems with organised communications between large numbers of nodes.
Well, at least most of the proposals for the client protocol have that being
SIP.  There are some people who think that the peer protocol is some kind of
extension to SIP.  We'll see.

Brian


From dan at ag-projects.com  Fri Dec 15 10:58:57 2006
From: dan at ag-projects.com (Dan Pascu)
Date: Fri, 15 Dec 2006 17:58:57 +0200
Subject: [p2p-sip] First Rule of Distributed Computing:
In-Reply-To: <191b01c7205e$ef195d20$640fa8c0@cis.neustar.com>
References: <191b01c7205e$ef195d20$640fa8c0@cis.neustar.com>
Message-ID: <200612151758.58022.dan@ag-projects.com>

On Friday 15 December 2006 17:37, Brian Rosen wrote:
> > In addition to the centralised login database, it appears there are
> > also at least 7 bootstrap nodes with statically assigned IP
> > addresses.
> >
> > http://skype.nexxen.cz/dokumentace/public/SkypeProtocol/html/
> > SkypeProtocol_4.htm
>
> Yes, the bootstrap uses fixed addresses.  Once you have contacted a
> peer, you don't need them any more.  We will need some kind of a
> bootstrap mechanism, and I think while we won't prohibit a fixed set of
> IP addresses, I think we're probably going to have other mechanisms as
> well.

I think that a more effective and also simpler mechanism for bootstrapping 
(than the list of fixed IP addresses) would be to have a domain name as 
the bootstrap point. This domain name would have a specific DNS SRV 
record which would finally point to a set of A records holding the IP 
addresses that will be used for bootstrapping.

While it has the same result, it is simpler to operate because if the 
fixed IP addresses change, you do not have to update the IP address list 
in each peer. All you have to do is update the list of A records to which 
the SRV record points and any peer will find the new IP's immediately 
using a simple DNS lookup.

-- 
Dan

From hsinnrei at adobe.com  Fri Dec 15 11:59:46 2006
From: hsinnrei at adobe.com (Henry Sinnreich)
Date: Fri, 15 Dec 2006 08:59:46 -0800
Subject: [p2p-sip] Concepts draft: Open Issue
	#3:Enrollmentas	aPeer/Client Protocol operation?
In-Reply-To: <6F0CB04509C11D46A54232E852E390AC0240F67F@MCHP7R6A.ww002.siemens.net>
Message-ID: <24CCCC428EFEA2469BF046DB3C7A8D22141419@namail5.corp.adobe.com>

Hannes,

> Previous attempts to get credential enrollment work started in the IETF failed

I believe attempts to standardize enrollment will continue to fail, since just as any other applications it is subject to competition and it should stay this way.

Enrollment has nothing to do with interoperability over the Internet and is better discussed at trade shows or in industry consortiums IMHO.

Thanks, Henry

-----Original Message-----
From: p2p-sip-bounces at cs.columbia.edu [mailto:p2p-sip-bounces at cs.columbia.edu] On Behalf Of Tschofenig, Hannes
Sent: Friday, December 15, 2006 9:15 AM
To: Brian Rosen; Jani Hautakorpi (JO/LMF); Spencer Dawkins
Cc: P2P-SIP
Subject: Re: [p2p-sip] Concepts draft: Open Issue #3:Enrollmentas aPeer/Client Protocol operation?

Hi Brian, 

I personally think we need a way to obtain credentials -- not only for P2P SIP. Obviously, this would not be done in the P2P SIP working group. Hence, the approach suggested by you seems to be reasonable. 

Previous attemps to get credential enrollment work started in the IETF failed (e.g., EasyCert BoF). 
I contribute my thoughts to a working group called ENROLL. Here is an example of the type of documents I wrote: http://www.tschofenig.com/drafts/draft-tschofenig-enroll-next-steps-00.txt. The group also failed. 

Ciao
Hannes

-----Urspr?ngliche Nachricht-----
Von: p2p-sip-bounces at cs.columbia.edu [mailto:p2p-sip-bounces at cs.columbia.edu] Im Auftrag von Brian Rosen
Gesendet: Freitag, 15. Dezember 2006 15:43
An: 'Jani Hautakorpi (JO/LMF)'; 'Spencer Dawkins'
Cc: 'P2P-SIP'
Betreff: Re: [p2p-sip] Concepts draft: Open Issue #3: Enrollmentas aPeer/Client Protocol operation?

I don't think we need to standardize the method of obtaining credentials.

I believe we agreed we need to standardize at least one mandatory to
implement FORM of credentials, so we could use credentials obtained by an
unspecified mechanism

Brian

> -----Original Message-----
> From: p2p-sip-bounces at cs.columbia.edu [mailto:p2p-sip-
> bounces at cs.columbia.edu] On Behalf Of Jani Hautakorpi (JO/LMF)
> Sent: Monday, December 11, 2006 8:28 AM
> To: Spencer Dawkins
> Cc: P2P-SIP
> Subject: Re: [p2p-sip] Concepts draft: Open Issue #3: Enrollment as
> aPeer/Client Protocol operation?
> 
> Hi all,
> 
> Spencer Dawkins wrote:
> > I liked David's additional statement, which was
> >
> > "- if we DO standardize, it could be this is useful to others outside
> >    P2PSIP and might even be some work we push into another WG."
> >
> > so, taking David's comment and your comment above into account, my list
> ends
> > up as
> >
> > - we don't HAVE to specify the method of obtaining credentials now, and
> >
> > - if we DO standardize the method of obtaining credentials, it may not
> >   be part of the peer and/or client protocol,
> >
> > - and might developed in another working group, if the resulting
> >   standardized method would be useful outside P2PSIP.
> >
> > Does this work for people?
> 
> This makes sense and sounds good at least to me.
> 
> 
> --
> Jani
> _______________________________________________
> p2p-sip mailing list
> p2p-sip at cs.columbia.edu
> https://lists.cs.columbia.edu/cucslists/listinfo/p2p-sip

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

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


From hannes.tschofenig at siemens.com  Fri Dec 15 16:36:10 2006
From: hannes.tschofenig at siemens.com (Tschofenig, Hannes)
Date: Fri, 15 Dec 2006 22:36:10 +0100
Subject: [p2p-sip] Concepts draft: Open Issue
	#3:Enrollmentas	aPeer/Client Protocol operation?
In-Reply-To: <24CCCC428EFEA2469BF046DB3C7A8D22141419@namail5.corp.adobe.com>
Message-ID: <6F0CB04509C11D46A54232E852E390AC0240F741@MCHP7R6A.ww002.siemens.net>

Hi Henry, 

Interesting view.

I am sure we have different things in mind when we use the term enrollment. 

Let me give you a specific example: I have a long-term password and I would like to use different protocols that use public key based techniques to secure the protocol operation. 

Now imagine that there is a protocol that allows me to perform a password-based authentication exchange once (which might require more roundtrips to be secure) and then a short-lived public/private key pair is created that can later be used for usage with other protocols. 

We already have some protocols. In the SIP area we have SIP CERT. In the 3GPP we have the Generic Bootstrapping Architecture. Kerberos could be seen as one such protocol. These protocols have different properties and their advantages and disadvantages. 

Does this sound like a reasonable thing?  

Ciao
Hannes
 
-----Urspr?ngliche Nachricht-----
Von: Henry Sinnreich [mailto:hsinnrei at adobe.com] 
Gesendet: Freitag, 15. Dezember 2006 18:00
An: Tschofenig, Hannes; Brian Rosen; Jani Hautakorpi (JO/LMF); Spencer Dawkins
Cc: P2P-SIP
Betreff: RE: [p2p-sip] Concepts draft: Open Issue #3:Enrollmentas aPeer/Client Protocol operation?

Hannes,

> Previous attempts to get credential enrollment work started in the IETF failed

I believe attempts to standardize enrollment will continue to fail, since just as any other applications it is subject to competition and it should stay this way.

Enrollment has nothing to do with interoperability over the Internet and is better discussed at trade shows or in industry consortiums IMHO.

Thanks, Henry

-----Original Message-----
From: p2p-sip-bounces at cs.columbia.edu [mailto:p2p-sip-bounces at cs.columbia.edu] On Behalf Of Tschofenig, Hannes
Sent: Friday, December 15, 2006 9:15 AM
To: Brian Rosen; Jani Hautakorpi (JO/LMF); Spencer Dawkins
Cc: P2P-SIP
Subject: Re: [p2p-sip] Concepts draft: Open Issue #3:Enrollmentas aPeer/Client Protocol operation?

Hi Brian, 

I personally think we need a way to obtain credentials -- not only for P2P SIP. Obviously, this would not be done in the P2P SIP working group. Hence, the approach suggested by you seems to be reasonable. 

Previous attemps to get credential enrollment work started in the IETF failed (e.g., EasyCert BoF). 
I contribute my thoughts to a working group called ENROLL. Here is an example of the type of documents I wrote: http://www.tschofenig.com/drafts/draft-tschofenig-enroll-next-steps-00.txt. The group also failed. 

Ciao
Hannes

-----Urspr?ngliche Nachricht-----
Von: p2p-sip-bounces at cs.columbia.edu [mailto:p2p-sip-bounces at cs.columbia.edu] Im Auftrag von Brian Rosen
Gesendet: Freitag, 15. Dezember 2006 15:43
An: 'Jani Hautakorpi (JO/LMF)'; 'Spencer Dawkins'
Cc: 'P2P-SIP'
Betreff: Re: [p2p-sip] Concepts draft: Open Issue #3: Enrollmentas aPeer/Client Protocol operation?

I don't think we need to standardize the method of obtaining credentials.

I believe we agreed we need to standardize at least one mandatory to
implement FORM of credentials, so we could use credentials obtained by an
unspecified mechanism

Brian

> -----Original Message-----
> From: p2p-sip-bounces at cs.columbia.edu [mailto:p2p-sip-
> bounces at cs.columbia.edu] On Behalf Of Jani Hautakorpi (JO/LMF)
> Sent: Monday, December 11, 2006 8:28 AM
> To: Spencer Dawkins
> Cc: P2P-SIP
> Subject: Re: [p2p-sip] Concepts draft: Open Issue #3: Enrollment as
> aPeer/Client Protocol operation?
> 
> Hi all,
> 
> Spencer Dawkins wrote:
> > I liked David's additional statement, which was
> >
> > "- if we DO standardize, it could be this is useful to others outside
> >    P2PSIP and might even be some work we push into another WG."
> >
> > so, taking David's comment and your comment above into account, my list
> ends
> > up as
> >
> > - we don't HAVE to specify the method of obtaining credentials now, and
> >
> > - if we DO standardize the method of obtaining credentials, it may not
> >   be part of the peer and/or client protocol,
> >
> > - and might developed in another working group, if the resulting
> >   standardized method would be useful outside P2PSIP.
> >
> > Does this work for people?
> 
> This makes sense and sounds good at least to me.
> 
> 
> --
> Jani
> _______________________________________________
> p2p-sip mailing list
> p2p-sip at cs.columbia.edu
> https://lists.cs.columbia.edu/cucslists/listinfo/p2p-sip

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

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


From warlock at media.mit.edu  Sat Dec 16 15:31:18 2006
From: warlock at media.mit.edu (Jacky Mallett)
Date: Sat, 16 Dec 2006 15:31:18 -0500
Subject: [p2p-sip] First Rule of Distributed Computing:
In-Reply-To: <191b01c7205e$ef195d20$640fa8c0@cis.neustar.com>
References: <191b01c7205e$ef195d20$640fa8c0@cis.neustar.com>
Message-ID: <DD67E852-000F-4F3E-9B31-01937C07B6F2@media.mit.edu>

> I'm not sure what "concurrent users" means.  The 8.5M number is the  
> number
> of subscribers on-line, not the number talking.  It's equivalent to a
> wireless phone connected to a network and able to, or actively  
> placing or
> receiving a call.  I haven't seen numbers like that for wireless  
> systems.
> What you see is "subscribers", which is the big number for Skype.   
> It's
> clearly true that the ratio of subscribers to devices on line and  
> available
> for or in a call is much higher for wireless systems than it is for  
> Skype.

that's what i meant. Skype essentially provides backbone network  
support to locate its users and allow them to circumvent NAT  
restrictions when necessary, and then the call traffic itself is  
mostly peer-to-peer. It doesn't seem to be widely appreciated that  
there are also limits on the size of the super node layer that is  
used to achieve this,  which implies that there are also limits on  
the number of online users that can be supported with their current  
architecture.

There is a similar disconnect in the phone network between the number  
of people with phone lines, and the number of people who could  
simultaneously make a phone call - i don't know the wireless figures  
either, but iirc the old PSTN engineering guideline was something  
like 400:1.


>
> Also, the numbers are for a single carrier, not the entire PSTN.   
> Carriers
> have interconnects, as will P2P-SIP domains.

Yes, in fact that's the only way we'll be able to scale the solution.  
Skype, of course, is notoriously a closed and encrypted system with  
no interconnects. Since some people keep using it as the reference  
case, i felt it might be useful to be a little more precise about how  
it actually works.

>
>>
>> It is correct to say that the SIP protocol as implemented is the
>> endpoint application. The distributed _system_ that uses it and other
>> protocols to interconnect users, is not. It has network like
>> properties all of its own.  Network in this sense is used to mean
>> systems with organised communications between large numbers of nodes.
> Well, at least most of the proposals for the client protocol have  
> that being
> SIP.  There are some people who think that the peer protocol is  
> some kind of
> extension to SIP.  We'll see.

Indeed.

...Jacky
>
> Brian


From br at brianrosen.net  Sat Dec 16 16:25:25 2006
From: br at brianrosen.net (Brian Rosen)
Date: Sat, 16 Dec 2006 16:25:25 -0500
Subject: [p2p-sip] First Rule of Distributed Computing:
In-Reply-To: <DD67E852-000F-4F3E-9B31-01937C07B6F2@media.mit.edu>
Message-ID: <02cc01c72158$b5c936d0$640fa8c0@cis.neustar.com>

> that's what i meant. Skype essentially provides backbone network
> support to locate its users and allow them to circumvent NAT
> restrictions when necessary, and then the call traffic itself is
> mostly peer-to-peer. It doesn't seem to be widely appreciated that
> there are also limits on the size of the super node layer that is
> used to achieve this,  which implies that there are also limits on
> the number of online users that can be supported with their current
> architecture.
I didn't think there was any effective limit on the size of the supernode
layer.  What kind of limit would there be?  It seems to be a pretty
classical P2P system which normally doesn't have such limits, at least until
the systems get larger than the kinds of systems we're talking about.

I've often wondered if there is an effective limit on the number of clients
behind a NAT related to the number of supernodes not behind the NAT.  I
suspect that if there are too many clients per supernode, it won't work
well.

> 
> There is a similar disconnect in the phone network between the number
> of people with phone lines, and the number of people who could
> simultaneously make a phone call - i don't know the wireless figures
> either, but iirc the old PSTN engineering guideline was something
> like 400:1.

Here the Internet has a significant advantage.  Because at least voice
communications is so much smaller than pipe size, I think it would actually
be possible for every logged in Skype user to be talking at the same time.
Unlike most TDM switches, with respect to the kinds of bandwidth needed for
voice and/or text, Skype should be non-blocking.

Brian


From warlock at media.mit.edu  Sat Dec 16 17:16:48 2006
From: warlock at media.mit.edu (Jacky Mallett)
Date: Sat, 16 Dec 2006 17:16:48 -0500
Subject: [p2p-sip] First Rule of Distributed Computing:
In-Reply-To: <02cc01c72158$b5c936d0$640fa8c0@cis.neustar.com>
References: <02cc01c72158$b5c936d0$640fa8c0@cis.neustar.com>
Message-ID: <EDC24A00-55A3-44B2-8F42-FC21D6DC8967@media.mit.edu>



> I didn't think there was any effective limit on the size of the  
> supernode
> layer.  What kind of limit would there be?  It seems to be a pretty
> classical P2P system which normally doesn't have such limits, at  
> least until
> the systems get larger than the kinds of systems we're talking about.

the limit on the size of the supernode layer is the limit on the  
number of connections each supernode can support with every other  
supernode (assuming a completely meshed network.) Otherwise, for a  
partially meshed network, it's a combination of that and the traffic  
within that level, bearing in mind that the partial mesh reduces the  
amount of connections each super node has to support, but increases  
the traffic within the supernode layer because not all traffic can  
then be sent directly supernode to supernode any more. Although most  
descriptions of Skype present the supernode layer as a complete mesh,  
it has to be partial,  otherwise the number of connections per  
supernode would be ridiculous.

Similarly there's a limit on the number of clients each supernode can  
support, since they're effectively in a client-server relationship  
with each other.

When the first p2p file sharing networks came in, they essentially  
functioned as a partial mesh network themselves, so they were in  
essence the super-node layer, and they eventually bogged down under  
the limits of scale under this model. Introducing the super-node  
layer doesn't solve that problem so much as postpone it, the same  
problems of excessive hop related administrative traffic eventually  
re-appears at that level as it grows.

>
> I've often wondered if there is an effective limit on the number of  
> clients
> behind a NAT related to the number of supernodes not behind the  
> NAT.  I
> suspect that if there are too many clients per supernode, it won't  
> work
> well.

that's the other problem. Which is what will make the upcoming surge  
in wifi-pda users quite interesting to watch.

> Here the Internet has a significant advantage.  Because at least voice
> communications is so much smaller than pipe size, I think it would  
> actually
> be possible for every logged in Skype user to be talking at the  
> same time.
> Unlike most TDM switches, with respect to the kinds of bandwidth  
> needed for
> voice and/or text, Skype should be non-blocking.

And this is where it will get very interesting. Yes, the fibre optic  
lines have dramatically increased the size of the pipe. And yes, the  
amount of traffic that a mesh/partial network can support is  
dramatically larger than that of any hierarchically organised  
network. O(Lsqrt(N)), vs O(L), where L is the connectivity supported  
by respectively the nodes in the mesh, and the hierarchical servers  
in the respective network topologies. However, the amount of traffic  
generated by all Skype users talking to each other simultaneously is O 
(N/2) which is eventually greater than O(Lsqrt(N)) for large numbers  
of N and reasonable assumptions for L.

The less theoretical observation that should also be made, is that  
today's internet is not engineered for continuous two way traffic, it  
is, by and large engineered with the assumption that traffic is  
highly asymmetric at the end points, which is already presenting its  
own problems. And let's not even think about what any significant  
amount of real time video traffic could do to this.

Some of these problems can be solved by technology as the size of the  
pipes increases, some of them are limits on any distributed system  
and will have to be engineered.

...Jacky


>
> Brian


From baumgart at tm.uka.de  Mon Dec 18 10:29:04 2006
From: baumgart at tm.uka.de (Ingmar Baumgart)
Date: Mon, 18 Dec 2006 16:29:04 +0100
Subject: [p2p-sip] OverSim: A Flexible Overlay Network Simulation Framework
Message-ID: <4586B3C0.4080603@tm.uka.de>

We are pleased to announce the first release of OverSim, a flexible
overlay network simulation framework based on OMNeT++. It was designed
to fulfill a number of requirements that have been partially neglected
by existing peer-to-peer simulation frameworks.

These requirements include the availability of a GUI for debugging
purposes, the possibility to exchange models of the underlaying network
with different levels of detail and the possibility to reuse protocol
implementations for both simulation and real world networks.

Some features of the first release:

- Includes implementations of the peer-to-peer protocols Chord and GIA.

- Interactive GUI

- Flexible and modular design (e.g. use of the common API for
  structured p2p overlays)

- Scalability: Simulation of networks with up to 100,000 nodes

For screenshots and further information please take a look at the
project web site:

http://www.oversim.org/

OverSim is developed at the Institute of Telematics, Universitaet
Karlsruhe (TH) within the scope of the ScaleNet project funded by the
German Federal Ministry of Education and Research. The simulator is
actively developed and open to contributions.

Best Regards,
  Ingmar

-- 
Dipl.-Inform. Ingmar Baumgart
Institut f?r Telematik, Universit?t Karlsruhe (TH)
Zirkel 2, 76128 Karlsruhe, Germany
Phone: +49 721 608 8281, Fax: +49 721 608 6789

From philip_matthews at magma.ca  Mon Dec 18 10:53:56 2006
From: philip_matthews at magma.ca (Philip Matthews)
Date: Mon, 18 Dec 2006 10:53:56 -0500
Subject: [p2p-sip] Concepts draft: Open Issue #3: Enrollment as a
	Peer/Client Protocol operation?
In-Reply-To: <6F0CB04509C11D46A54232E852E390AC0240F67F@MCHP7R6A.ww002.siemens.net>
References: <6F0CB04509C11D46A54232E852E390AC0240F67F@MCHP7R6A.ww002.siemens.net>
Message-ID: <6F7D605A-475B-4D52-B76F-5ADD9E2F4B4B@magma.ca>

Hannes:

Despite your frustration at trying to get the IETF to standardize a  
method of obtaining credentials, it sounds like you still support the  
idea that standardizing this is out-of-scope for P2PSIP (at least for  
now).
Am I correct?

- Philip

On 15-Dec-06, at 10:14 , Tschofenig, Hannes wrote:

> Hi Brian,
>
> I personally think we need a way to obtain credentials -- not only  
> for P2P SIP. Obviously, this would not be done in the P2P SIP  
> working group. Hence, the approach suggested by you seems to be  
> reasonable.
>
> Previous attemps to get credential enrollment work started in the  
> IETF failed (e.g., EasyCert BoF).
> I contribute my thoughts to a working group called ENROLL. Here is  
> an example of the type of documents I wrote: http:// 
> www.tschofenig.com/drafts/draft-tschofenig-enroll-next- 
> steps-00.txt. The group also failed.
>
> Ciao
> Hannes
>
> -----Urspr?ngliche Nachricht-----
> Von: p2p-sip-bounces at cs.columbia.edu [mailto:p2p-sip- 
> bounces at cs.columbia.edu] Im Auftrag von Brian Rosen
> Gesendet: Freitag, 15. Dezember 2006 15:43
> An: 'Jani Hautakorpi (JO/LMF)'; 'Spencer Dawkins'
> Cc: 'P2P-SIP'
> Betreff: Re: [p2p-sip] Concepts draft: Open Issue #3: Enrollmentas  
> aPeer/Client Protocol operation?
>
> I don't think we need to standardize the method of obtaining  
> credentials.
>
> I believe we agreed we need to standardize at least one mandatory to
> implement FORM of credentials, so we could use credentials obtained  
> by an
> unspecified mechanism
>
> Brian
>
>> -----Original Message-----
>> From: p2p-sip-bounces at cs.columbia.edu [mailto:p2p-sip-
>> bounces at cs.columbia.edu] On Behalf Of Jani Hautakorpi (JO/LMF)
>> Sent: Monday, December 11, 2006 8:28 AM
>> To: Spencer Dawkins
>> Cc: P2P-SIP
>> Subject: Re: [p2p-sip] Concepts draft: Open Issue #3: Enrollment as
>> aPeer/Client Protocol operation?
>>
>> Hi all,
>>
>> Spencer Dawkins wrote:
>>> I liked David's additional statement, which was
>>>
>>> "- if we DO standardize, it could be this is useful to others  
>>> outside
>>>    P2PSIP and might even be some work we push into another WG."
>>>
>>> so, taking David's comment and your comment above into account,  
>>> my list
>> ends
>>> up as
>>>
>>> - we don't HAVE to specify the method of obtaining credentials  
>>> now, and
>>>
>>> - if we DO standardize the method of obtaining credentials, it  
>>> may not
>>>   be part of the peer and/or client protocol,
>>>
>>> - and might developed in another working group, if the resulting
>>>   standardized method would be useful outside P2PSIP.
>>>
>>> Does this work for people?
>>
>> This makes sense and sounds good at least to me.
>>
>>
>> --
>> Jani
>> _______________________________________________
>> p2p-sip mailing list
>> p2p-sip at cs.columbia.edu
>> https://lists.cs.columbia.edu/cucslists/listinfo/p2p-sip
>
> _______________________________________________
> p2p-sip mailing list
> p2p-sip at cs.columbia.edu
> https://lists.cs.columbia.edu/cucslists/listinfo/p2p-sip
>
> _______________________________________________
> p2p-sip mailing list
> p2p-sip at cs.columbia.edu
> https://lists.cs.columbia.edu/cucslists/listinfo/p2p-sip
>



From philip_matthews at magma.ca  Wed Dec 20 14:30:59 2006
From: philip_matthews at magma.ca (Philip Matthews)
Date: Wed, 20 Dec 2006 14:30:59 -0500
Subject: [p2p-sip] Concepts Draft: Open Issue #4: Can clients offer services?
Message-ID: <2043E08B-FCB3-424A-B23C-B06BA0F65532@magma.ca>

Open Issue #4: Can clients offer services? Or are services only  
offered by peers?

Philip's Strawman Proposal: Only peers can offer services. Clients  
can use services, but  do not offer them.

There are two reasons for this proposal:
One is that, to offer a service, a client would presumably need some  
sort of ID in order to identify the client offering the service, and  
the Concepts Draft currently states clients do not have IDs.
The second is that this is consistent with the view that the client  
protocol could be just vanilla SIP. It is not clear how to offer  
services if the client protocol is SIP.

The authors of the Concepts Draft have not been able to resolve this  
amongst ourselves, so we are throwing the issue out for general  
debate. (Err ... to be honest, we haven't tried very hard either, as  
people are busy with other things).

- Philip



From hsinnrei at adobe.com  Wed Dec 20 15:10:44 2006
From: hsinnrei at adobe.com (Henry Sinnreich)
Date: Wed, 20 Dec 2006 12:10:44 -0800
Subject: [p2p-sip] Concepts Draft: Open Issue #4: Can clients offer
	services?
In-Reply-To: <2043E08B-FCB3-424A-B23C-B06BA0F65532@magma.ca>
Message-ID: <24CCCC428EFEA2469BF046DB3C7A8D2214144D@namail5.corp.adobe.com>

This sounds good enough for me, unless someone comes up with a
compelling usage scenario to the contrary.

Happy holidays and a successful new year!

Henry

-----Original Message-----
From: p2p-sip-bounces at cs.columbia.edu
[mailto:p2p-sip-bounces at cs.columbia.edu] On Behalf Of Philip Matthews
Sent: Wednesday, December 20, 2006 1:31 PM
To: P2P-SIP
Cc: David A. Bryan
Subject: [p2p-sip] Concepts Draft: Open Issue #4: Can clients offer
services?

Open Issue #4: Can clients offer services? Or are services only  
offered by peers?

Philip's Strawman Proposal: Only peers can offer services. Clients  
can use services, but  do not offer them.

There are two reasons for this proposal:
One is that, to offer a service, a client would presumably need some  
sort of ID in order to identify the client offering the service, and  
the Concepts Draft currently states clients do not have IDs.
The second is that this is consistent with the view that the client  
protocol could be just vanilla SIP. It is not clear how to offer  
services if the client protocol is SIP.

The authors of the Concepts Draft have not been able to resolve this  
amongst ourselves, so we are throwing the issue out for general  
debate. (Err ... to be honest, we haven't tried very hard either, as  
people are busy with other things).

- Philip


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


From spencer at mcsr-labs.org  Wed Dec 20 16:56:12 2006
From: spencer at mcsr-labs.org (Spencer Dawkins)
Date: Wed, 20 Dec 2006 15:56:12 -0600
Subject: [p2p-sip] Concepts Draft: Open Issue #4: Can clients
	offerservices?
References: <24CCCC428EFEA2469BF046DB3C7A8D2214144D@namail5.corp.adobe.com>
Message-ID: <044601c72481$aaa5eb50$6a01a8c0@china.huawei.com>

I agree with Philip's proposal. I'm not doing well on understanding why a 
client offering a service *could not* be a peer instead of being a client - 
in addition to Philip's two reasons stated below, which both make sense to 
me.

Like Henry, I'd appreciate counterexamples in the form of use cases.

Thanks,

Spencer

From: "Henry Sinnreich" <hsinnrei at adobe.com>

> This sounds good enough for me, unless someone comes up with a
> compelling usage scenario to the contrary.
>
> Happy holidays and a successful new year!
>
> Henry
>
> -----Original Message-----
> From: p2p-sip-bounces at cs.columbia.edu
> [mailto:p2p-sip-bounces at cs.columbia.edu] On Behalf Of Philip Matthews
> Sent: Wednesday, December 20, 2006 1:31 PM
> To: P2P-SIP
> Cc: David A. Bryan
> Subject: [p2p-sip] Concepts Draft: Open Issue #4: Can clients offer
> services?
>
> Open Issue #4: Can clients offer services? Or are services only
> offered by peers?
>
> Philip's Strawman Proposal: Only peers can offer services. Clients
> can use services, but  do not offer them.
>
> There are two reasons for this proposal:
> One is that, to offer a service, a client would presumably need some
> sort of ID in order to identify the client offering the service, and
> the Concepts Draft currently states clients do not have IDs.
> The second is that this is consistent with the view that the client
> protocol could be just vanilla SIP. It is not clear how to offer
> services if the client protocol is SIP. 



From br at brianrosen.net  Wed Dec 20 17:26:09 2006
From: br at brianrosen.net (Brian Rosen)
Date: Wed, 20 Dec 2006 17:26:09 -0500
Subject: [p2p-sip] Concepts Draft: Open Issue #4: Can
	clientsofferservices?
In-Reply-To: <044601c72481$aaa5eb50$6a01a8c0@china.huawei.com>
Message-ID: <0a6e01c72485$db613660$640fa8c0@cis.neustar.com>

I admit to not having enough experience here to really stick my neck out too
far, but suppose that the only role of a "peer" was to relay messages to
clients (it could be a client itself), e.g. Nat traversal relays.

Then a client is really all there are, and all services come from clients.

Brian

> -----Original Message-----
> From: p2p-sip-bounces at cs.columbia.edu [mailto:p2p-sip-
> bounces at cs.columbia.edu] On Behalf Of Spencer Dawkins
> Sent: Wednesday, December 20, 2006 4:56 PM
> To: P2P-SIP
> Subject: Re: [p2p-sip] Concepts Draft: Open Issue #4: Can
> clientsofferservices?
> 
> I agree with Philip's proposal. I'm not doing well on understanding why a
> client offering a service *could not* be a peer instead of being a client
> -
> in addition to Philip's two reasons stated below, which both make sense to
> me.
> 
> Like Henry, I'd appreciate counterexamples in the form of use cases.
> 
> Thanks,
> 
> Spencer
> 
> From: "Henry Sinnreich" <hsinnrei at adobe.com>
> 
> > This sounds good enough for me, unless someone comes up with a
> > compelling usage scenario to the contrary.
> >
> > Happy holidays and a successful new year!
> >
> > Henry
> >
> > -----Original Message-----
> > From: p2p-sip-bounces at cs.columbia.edu
> > [mailto:p2p-sip-bounces at cs.columbia.edu] On Behalf Of Philip Matthews
> > Sent: Wednesday, December 20, 2006 1:31 PM
> > To: P2P-SIP
> > Cc: David A. Bryan
> > Subject: [p2p-sip] Concepts Draft: Open Issue #4: Can clients offer
> > services?
> >
> > Open Issue #4: Can clients offer services? Or are services only
> > offered by peers?
> >
> > Philip's Strawman Proposal: Only peers can offer services. Clients
> > can use services, but  do not offer them.
> >
> > There are two reasons for this proposal:
> > One is that, to offer a service, a client would presumably need some
> > sort of ID in order to identify the client offering the service, and
> > the Concepts Draft currently states clients do not have IDs.
> > The second is that this is consistent with the view that the client
> > protocol could be just vanilla SIP. It is not clear how to offer
> > services if the client protocol is SIP.
> 
> 
> _______________________________________________
> p2p-sip mailing list
> p2p-sip at cs.columbia.edu
> https://lists.cs.columbia.edu/cucslists/listinfo/p2p-sip


From spencer at mcsr-labs.org  Wed Dec 20 17:59:35 2006
From: spencer at mcsr-labs.org (Spencer Dawkins)
Date: Wed, 20 Dec 2006 16:59:35 -0600
Subject: [p2p-sip] Concepts Draft: Open Issue #4: Can
	clientsofferservices?
References: <0a6e01c72485$db613660$640fa8c0@cis.neustar.com>
Message-ID: <046501c7248a$84fcdf40$6a01a8c0@china.huawei.com>

This would be a reasonable view of the world, but it doesn't seem to be the 
view of the world described in the Concepts draft.

Section 3.2 has a reference architecture diagram that pretty clearly shows 
combination peer/proxies, etc. and does not show any services on a client.

Not that this is a working group draft, or even a working group yet, but if 
other people have been waiting to look at the proposed architecture more 
closely, this would be a lovely time for that to happen...

Thanks,

Spencer

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


>I admit to not having enough experience here to really stick my neck out 
>too
> far, but suppose that the only role of a "peer" was to relay messages to
> clients (it could be a client itself), e.g. Nat traversal relays.
>
> Then a client is really all there are, and all services come from clients.
>
> Brian
>
>> -----Original Message-----
>> From: p2p-sip-bounces at cs.columbia.edu [mailto:p2p-sip-
>> bounces at cs.columbia.edu] On Behalf Of Spencer Dawkins
>> Sent: Wednesday, December 20, 2006 4:56 PM
>> To: P2P-SIP
>> Subject: Re: [p2p-sip] Concepts Draft: Open Issue #4: Can
>> clientsofferservices?
>>
>> I agree with Philip's proposal. I'm not doing well on understanding why a
>> client offering a service *could not* be a peer instead of being a client
>> -
>> in addition to Philip's two reasons stated below, which both make sense 
>> to
>> me.
>>
>> Like Henry, I'd appreciate counterexamples in the form of use cases.
>>
>> Thanks,
>>
>> Spencer
>>
>> From: "Henry Sinnreich" <hsinnrei at adobe.com>
>>
>> > This sounds good enough for me, unless someone comes up with a
>> > compelling usage scenario to the contrary.
>> >
>> > Happy holidays and a successful new year!
>> >
>> > Henry
>> >
>> > -----Original Message-----
>> > From: p2p-sip-bounces at cs.columbia.edu
>> > [mailto:p2p-sip-bounces at cs.columbia.edu] On Behalf Of Philip Matthews
>> > Sent: Wednesday, December 20, 2006 1:31 PM
>> > To: P2P-SIP
>> > Cc: David A. Bryan
>> > Subject: [p2p-sip] Concepts Draft: Open Issue #4: Can clients offer
>> > services?
>> >
>> > Open Issue #4: Can clients offer services? Or are services only
>> > offered by peers?
>> >
>> > Philip's Strawman Proposal: Only peers can offer services. Clients
>> > can use services, but  do not offer them.
>> >
>> > There are two reasons for this proposal:
>> > One is that, to offer a service, a client would presumably need some
>> > sort of ID in order to identify the client offering the service, and
>> > the Concepts Draft currently states clients do not have IDs.
>> > The second is that this is consistent with the view that the client
>> > protocol could be just vanilla SIP. It is not clear how to offer
>> > services if the client protocol is SIP.
>>
>>
>> _______________________________________________
>> p2p-sip mailing list
>> p2p-sip at cs.columbia.edu
>> https://lists.cs.columbia.edu/cucslists/listinfo/p2p-sip
>
> 



From enrico.marocco at telecomitalia.it  Wed Dec 20 19:52:24 2006
From: enrico.marocco at telecomitalia.it (Enrico Marocco)
Date: Thu, 21 Dec 2006 01:52:24 +0100
Subject: [p2p-sip] Concepts Draft: Open Issue #4: Can clients offer
 services?
In-Reply-To: <2043E08B-FCB3-424A-B23C-B06BA0F65532@magma.ca>
References: <2043E08B-FCB3-424A-B23C-B06BA0F65532@magma.ca>
Message-ID: <4589DAC8.2070900@telecomitalia.it>

Philip Matthews wrote:
> Open Issue #4: Can clients offer services? Or are services only  
> offered by peers?
> 
> Philip's Strawman Proposal: Only peers can offer services. Clients  
> can use services, but  do not offer them.
> 
> There are two reasons for this proposal:
> One is that, to offer a service, a client would presumably need some  
> sort of ID in order to identify the client offering the service, and  
> the Concepts Draft currently states clients do not have IDs.
> The second is that this is consistent with the view that the client  
> protocol could be just vanilla SIP. It is not clear how to offer  
> services if the client protocol is SIP.

I think it would be a big mistake to take such a decision now.  First of
all, we don't really know what a service is (anybody subscribed to the
sipping mailing list knows what I mean); it could be a conference server
identified by a common SIP URL, or a STUN or TEREDO relay, identified by
a URI following a non-yet-defined schema.  Then we neither actually know
what a peer is; possibly a device behind a NAT could be a perfect
voicemail server (i.e. could host a service) while being unable to
participate in the DHT maintainance because of network limitations.

Such an issue is too technical for being in the scope of the concept draft.

> The authors of the Concepts Draft have not been able to resolve this  
> amongst ourselves, so we are throwing the issue out for general  
> debate. (Err ... to be honest, we haven't tried very hard either, as  
> people are busy with other things).

I don't see the need to address such an issue in the concept draft;
again, it is a merely technical decision and perhaps it is worth
deferring it to the deployer's will.

-- 
Ciao,
Enrico

From jiang.x.f at huawei.com  Wed Dec 20 21:12:29 2006
From: jiang.x.f at huawei.com (JiangXingFeng)
Date: Thu, 21 Dec 2006 10:12:29 +0800
Subject: [p2p-sip] Concepts Draft: Open Issue #4: Can clients offer
 services?
In-Reply-To: <2043E08B-FCB3-424A-B23C-B06BA0F65532@magma.ca>
Message-ID: <000601c724a5$77ed4270$7205a40a@china.huawei.com>

Hi, all:
IMHO, I think client and service are two independent concepts. First, a
client is a client because it has some limitations to participate in overlay
and maintain overly. The reasons include small storage, low computing
capability, the node being behind NAT, and the like. Second, services could
be thought of as a function which could help others. In the P2PSIP case,
what clients do is to publish the service and do not maintain overlay and
what peers do is to not only publish service but also participate in and
maintain the overlay. If others want to get a specific service, they could
search through the overlay made up of peers and find out who could provide
the service. Maybe service providers are clients. 
So, in my view, it is no need to put a limitation on the client. Comments
are welcome. 
Regards!
--
Jiang XingFeng
> -----Original Message-----
> From: p2p-sip-bounces at cs.columbia.edu
> [mailto:p2p-sip-bounces at cs.columbia.edu] On Behalf Of Philip Matthews
> Sent: Thursday, December 21, 2006 3:31 AM
> To: P2P-SIP
> Cc: David A. Bryan
> Subject: [p2p-sip] Concepts Draft: Open Issue #4: Can clients offer
services?
> 
> Open Issue #4: Can clients offer services? Or are services only
> offered by peers?
> 
> Philip's Strawman Proposal: Only peers can offer services. Clients
> can use services, but  do not offer them.
> 
> There are two reasons for this proposal:
> One is that, to offer a service, a client would presumably need some
> sort of ID in order to identify the client offering the service, and
> the Concepts Draft currently states clients do not have IDs.
> The second is that this is consistent with the view that the client
> protocol could be just vanilla SIP. It is not clear how to offer
> services if the client protocol is SIP.
> 
> The authors of the Concepts Draft have not been able to resolve this
> amongst ourselves, so we are throwing the issue out for general
> debate. (Err ... to be honest, we haven't tried very hard either, as
> people are busy with other things).
> 
> - Philip
> 
> 
> _______________________________________________
> p2p-sip mailing list
> p2p-sip at cs.columbia.edu
> https://lists.cs.columbia.edu/cucslists/listinfo/p2p-sip



From RADHIKA.R.ROY at saic.com  Wed Dec 20 22:01:05 2006
From: RADHIKA.R.ROY at saic.com (Roy, Radhika R.)
Date: Wed, 20 Dec 2006 22:01:05 -0500
Subject: [p2p-sip] Concepts Draft: Open Issue #4: Can clients offer
	services?
References: <000601c724a5$77ed4270$7205a40a@china.huawei.com>
Message-ID: <62D92A9A02BCC845B202323D49A48426E1CF6C@0591-ITS-EXMP02.us.saic.com>

Does it mean that clients can will a "subset" of functionalities of those of peers, while peers will have the "full-set" of functionalities in P2P-SIP? 
 
If not, what does differentiate between peers and clients in P2P-SIP?
 
Best regards,
Radhika
 
Note: I did not use the term "service" as we have not yet defined what a service means. Or, it may so happen, like SIPPING WG as Encrico mentioned, we may NOT even need to define the term "service" in the first place.

________________________________

From: p2p-sip-bounces at cs.columbia.edu on behalf of JiangXingFeng
Sent: Wed 12/20/2006 9:12 PM
To: 'Philip Matthews'
Cc: 'P2P-SIP'
Subject: Re: [p2p-sip] Concepts Draft: Open Issue #4: Can clients offer services?


  

Hi, all:
IMHO, I think client and service are two independent concepts. First, a
client is a client because it has some limitations to participate in overlay
and maintain overly. The reasons include small storage, low computing
capability, the node being behind NAT, and the like. Second, services could
be thought of as a function which could help others. In the P2PSIP case,
what clients do is to publish the service and do not maintain overlay and
what peers do is to not only publish service but also participate in and
maintain the overlay. If others want to get a specific service, they could
search through the overlay made up of peers and find out who could provide
the service. Maybe service providers are clients.
So, in my view, it is no need to put a limitation on the client. Comments
are welcome.
Regards!
--
Jiang XingFeng
> -----Original Message-----
> From: p2p-sip-bounces at cs.columbia.edu
> [mailto:p2p-sip-bounces at cs.columbia.edu] On Behalf Of Philip Matthews
> Sent: Thursday, December 21, 2006 3:31 AM
> To: P2P-SIP
> Cc: David A. Bryan
> Subject: [p2p-sip] Concepts Draft: Open Issue #4: Can clients offer
services?
>
> Open Issue #4: Can clients offer services? Or are services only
> offered by peers?
>
> Philip's Strawman Proposal: Only peers can offer services. Clients
> can use services, but  do not offer them.
>
> There are two reasons for this proposal:
> One is that, to offer a service, a client would presumably need some
> sort of ID in order to identify the client offering the service, and
> the Concepts Draft currently states clients do not have IDs.
> The second is that this is consistent with the view that the client
> protocol could be just vanilla SIP. It is not clear how to offer
> services if the client protocol is SIP.
>
> The authors of the Concepts Draft have not been able to resolve this
> amongst ourselves, so we are throwing the issue out for general
> debate. (Err ... to be honest, we haven't tried very hard either, as
> people are busy with other things).
>
> - Philip
>
>
> _______________________________________________
> p2p-sip mailing list
> p2p-sip at cs.columbia.edu
> https://lists.cs.columbia.edu/cucslists/listinfo/p2p-sip


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


-------------- next part --------------
An HTML attachment was scrubbed...
URL: https://lists.cs.columbia.edu/pipermail/p2p-sip/attachments/20061220/15c76364/attachment-0001.html

From jiang.x.f at huawei.com  Wed Dec 20 22:24:01 2006
From: jiang.x.f at huawei.com (JiangXingFeng)
Date: Thu, 21 Dec 2006 11:24:01 +0800
Subject: [p2p-sip] Concepts Draft: Open Issue #4: Can clients offer
 services?
In-Reply-To: <62D92A9A02BCC845B202323D49A48426E1CF6C@0591-ITS-EXMP02.us.saic.com>
Message-ID: <000301c724af$76106090$7205a40a@china.huawei.com>

Hi? Roy:

   Please see inline.

Regards!

--
Jiang XingFeng



  _____  

From: Roy, Radhika R. [mailto:RADHIKA.R.ROY at saic.com] 
Sent: Thursday, December 21, 2006 11:01 AM
To: JiangXingFeng; Philip Matthews
Cc: P2P-SIP
Subject: RE: [p2p-sip] Concepts Draft: Open Issue #4: Can clients offer
services?

 

Does it mean that clients can will a "subset" of functionalities of those of
peers, while peers will have the "full-set" of functionalities in P2P-SIP? 

Yes.  The functionalities what we are discussing here are related to P2P
overlay, not including others, such as stun, media relay. 

 

If not, what does differentiate between peers and clients in P2P-SIP?

Peers could provide routing and storage service which clients could not
provide. I think it is the essential difference between peers and clients. 

 

Best regards,

Radhika

 

Note: I did not use the term "service" as we have not yet defined what a
service means. Or, it may so happen, like SIPPING WG as Encrico mentioned,
we may NOT even need to define the term "service" in the first place.

 

  _____  

From: p2p-sip-bounces at cs.columbia.edu on behalf of JiangXingFeng
Sent: Wed 12/20/2006 9:12 PM
To: 'Philip Matthews'
Cc: 'P2P-SIP'
Subject: Re: [p2p-sip] Concepts Draft: Open Issue #4: Can clients offer
services?

  

Hi, all:
IMHO, I think client and service are two independent concepts. First, a
client is a client because it has some limitations to participate in overlay
and maintain overly. The reasons include small storage, low computing
capability, the node being behind NAT, and the like. Second, services could
be thought of as a function which could help others. In the P2PSIP case,
what clients do is to publish the service and do not maintain overlay and
what peers do is to not only publish service but also participate in and
maintain the overlay. If others want to get a specific service, they could
search through the overlay made up of peers and find out who could provide
the service. Maybe service providers are clients.
So, in my view, it is no need to put a limitation on the client. Comments
are welcome.
Regards!
--
Jiang XingFeng
> -----Original Message-----
> From: p2p-sip-bounces at cs.columbia.edu
> [mailto:p2p-sip-bounces at cs.columbia.edu] On Behalf Of Philip Matthews
> Sent: Thursday, December 21, 2006 3:31 AM
> To: P2P-SIP
> Cc: David A. Bryan
> Subject: [p2p-sip] Concepts Draft: Open Issue #4: Can clients offer
services?
>
> Open Issue #4: Can clients offer services? Or are services only
> offered by peers?
>
> Philip's Strawman Proposal: Only peers can offer services. Clients
> can use services, but  do not offer them.
>
> There are two reasons for this proposal:
> One is that, to offer a service, a client would presumably need some
> sort of ID in order to identify the client offering the service, and
> the Concepts Draft currently states clients do not have IDs.
> The second is that this is consistent with the view that the client
> protocol could be just vanilla SIP. It is not clear how to offer
> services if the client protocol is SIP.
>
> The authors of the Concepts Draft have not been able to resolve this
> amongst ourselves, so we are throwing the issue out for general
> debate. (Err ... to be honest, we haven't tried very hard either, as
> people are busy with other things).
>
> - Philip
>
>
> _______________________________________________
> p2p-sip mailing list
> p2p-sip at cs.columbia.edu
> https://lists.cs.columbia.edu/cucslists/listinfo/p2p-sip


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

-------------- next part --------------
An HTML attachment was scrubbed...
URL: https://lists.cs.columbia.edu/pipermail/p2p-sip/attachments/20061220/1978cc26/attachment.html

From philip_matthews at magma.ca  Thu Dec 21 14:32:17 2006
From: philip_matthews at magma.ca (Philip Matthews)
Date: Thu, 21 Dec 2006 14:32:17 -0500
Subject: [p2p-sip] Concepts Draft: Open Issue #4: Can
	clientsofferservices?
In-Reply-To: <0a6e01c72485$db613660$640fa8c0@cis.neustar.com>
References: <0a6e01c72485$db613660$640fa8c0@cis.neustar.com>
Message-ID: <8197ADBD-5EAD-4B6B-AD83-D582D6B106B0@magma.ca>


On 20-Dec-06, at 17:26 , Brian Rosen wrote:

> I admit to not having enough experience here to really stick my  
> neck out too
> far, but suppose that the only role of a "peer" was to relay  
> messages to
> clients (it could be a client itself), e.g. Nat traversal relays.
>
> Then a client is really all there are, and all services come from  
> clients.
>

Interesting.
As I see it,  what you are calling a "client" is what the Concepts  
draft calls a "peer" !!
With a small difference that some of these clients/peers help with  
forwarding messages to other clients/peers (assuming we adopt this  
proposal for the resolution of Open Issue #2, which most people seem  
to support).

- Philip



From philip_matthews at magma.ca  Thu Dec 21 14:39:06 2006
From: philip_matthews at magma.ca (Philip Matthews)
Date: Thu, 21 Dec 2006 14:39:06 -0500
Subject: [p2p-sip] Concepts Draft: Open Issue #4: Can
	clientsofferservices?
In-Reply-To: <046501c7248a$84fcdf40$6a01a8c0@china.huawei.com>
References: <0a6e01c72485$db613660$640fa8c0@cis.neustar.com>
	<046501c7248a$84fcdf40$6a01a8c0@china.huawei.com>
Message-ID: <60A0B289-D5D3-4F8E-8075-CD1A8AF96AC4@magma.ca>


On 20-Dec-06, at 17:59 , Spencer Dawkins wrote:

> This would be a reasonable view of the world, but it doesn't seem  
> to be the
> view of the world described in the Concepts draft.
>
> Section 3.2 has a reference architecture diagram that pretty  
> clearly shows
> combination peer/proxies, etc. and does not show any services on a  
> client.
>
> Not that this is a working group draft, or even a working group  
> yet, but if
> other people have been waiting to look at the proposed architecture  
> more
> closely, this would be a lovely time for that to happen...
>
> Thanks,
>
> Spencer

I would like to second Spencer's sentiment.

We don't have an official WG yet (though the last think I heard was  
that things are looking very good that there will be by March), so  
there is no such thing as official WG documents yet, but my co- 
authors and I are trying hard to reflect the ideas of the group in  
the Concepts draft. So if you have a different view of what a P2PSIP  
network looks like, speak up. If the rest of the group agrees with  
you, then we will update the Concepts draft to match.

It is precisely for this reason that we are bringing the open issues  
(as we see them) to the mailing list before doing the next revision.

- Philip

From br at brianrosen.net  Thu Dec 21 14:54:30 2006
From: br at brianrosen.net (Brian Rosen)
Date: Thu, 21 Dec 2006 14:54:30 -0500
Subject: [p2p-sip] Concepts Draft: Open Issue #4:
	Canclientsofferservices?
In-Reply-To: <60A0B289-D5D3-4F8E-8075-CD1A8AF96AC4@magma.ca>
Message-ID: <0cce01c72539$d69a79c0$640fa8c0@cis.neustar.com>

Okay, so we just used this argument with the "DNS is all we need" folks, and
now I'm on the other side a little.  This view would NOT conflict with the
charter, I would note.

I have been bothered by the special status we give to peers over clients.
We need to have two kinds of entities for one very good reason: clients
can't talk to other clients (or peers) directly, primarily because of NAT
issues.

However, we have taken that very real problem, and, for some reason I don't
understand, decided that SINCE you can't exchange messages directly, THEN
peers should have other distinguishing characteristics from clients.  That's
the part I don't get.  Why is that?  Why should being behind a NAT
automatically make you a dumb client instead of a smart peer?  Why can't
clients store data for example?  In this case, why can't they provide
services?

What is the requirement, or the practical reality, that makes the difference
between a client and a peer more than just a NAT traversal problem?

Brian

Who still admits to not being an expert here and thus gets to ask the stupid
questions that sometimes generates "aha's" from those that are.




> -----Original Message-----
> From: p2p-sip-bounces at cs.columbia.edu [mailto:p2p-sip-
> bounces at cs.columbia.edu] On Behalf Of Philip Matthews
> Sent: Thursday, December 21, 2006 2:39 PM
> To: P2P-SIP
> Subject: Re: [p2p-sip] Concepts Draft: Open Issue #4:
> Canclientsofferservices?
> 
> 
> On 20-Dec-06, at 17:59 , Spencer Dawkins wrote:
> 
> > This would be a reasonable view of the world, but it doesn't seem
> > to be the
> > view of the world described in the Concepts draft.
> >
> > Section 3.2 has a reference architecture diagram that pretty
> > clearly shows
> > combination peer/proxies, etc. and does not show any services on a
> > client.
> >
> > Not that this is a working group draft, or even a working group
> > yet, but if
> > other people have been waiting to look at the proposed architecture
> > more
> > closely, this would be a lovely time for that to happen...
> >
> > Thanks,
> >
> > Spencer
> 
> I would like to second Spencer's sentiment.
> 
> We don't have an official WG yet (though the last think I heard was
> that things are looking very good that there will be by March), so
> there is no such thing as official WG documents yet, but my co-
> authors and I are trying hard to reflect the ideas of the group in
> the Concepts draft. So if you have a different view of what a P2PSIP
> network looks like, speak up. If the rest of the group agrees with
> you, then we will update the Concepts draft to match.
> 
> It is precisely for this reason that we are bringing the open issues
> (as we see them) to the mailing list before doing the next revision.
> 
> - Philip
> _______________________________________________
> p2p-sip mailing list
> p2p-sip at cs.columbia.edu
> https://lists.cs.columbia.edu/cucslists/listinfo/p2p-sip


From philip_matthews at magma.ca  Thu Dec 21 17:40:11 2006
From: philip_matthews at magma.ca (Philip Matthews)
Date: Thu, 21 Dec 2006 17:40:11 -0500
Subject: [p2p-sip] Concepts Draft: Open Issue #4: Can clients offer
	services?
In-Reply-To: <2043E08B-FCB3-424A-B23C-B06BA0F65532@magma.ca>
References: <2043E08B-FCB3-424A-B23C-B06BA0F65532@magma.ca>
Message-ID: <56B6DA37-D085-44F2-8173-0A321D422196@magma.ca>



On 20-Dec-06, at 14:30 , Philip Matthews wrote:

> Open Issue #4: Can clients offer services? Or are services only  
> offered by peers?
>
> Philip's Strawman Proposal: Only peers can offer services. Clients  
> can use services, but  do not offer them.
>
> There are two reasons for this proposal:
> One is that, to offer a service, a client would presumably need  
> some sort of ID in order to identify the client offering the  
> service, and the Concepts Draft currently states clients do not  
> have IDs.
> The second is that this is consistent with the view that the client  
> protocol could be just vanilla SIP. It is not clear how to offer  
> services if the client protocol is SIP.
>
> The authors of the Concepts Draft have not been able to resolve  
> this amongst ourselves, so we are throwing the issue out for  
> general debate. (Err ... to be honest, we haven't tried very hard  
> either, as people are busy with other things).
>
> - Philip
>

Folks:

I have thought about this issue some more, and would like to change  
some of the things I said above.

I now think it is possible to treat service registration in a way  
similar to user registration. In particular, if clients offered  
services, this could be done using the same techniques as user  
registration.

IF this technique is acceptable,  then services can be offered using  
standard SIP mechanisms, and this can be done without requiring a  
client to have an ID. This applies at least to services that are  
invokable using a SIP Invite.


Below is an example of how a C/S SIP UA might offer voicemail  
services to a P2PSIP Overlay:
1. The C/S SIP UA (call it X) registers the URI   
"sip:voicemail.p2psipNetwork.org" with a peer (call it Y) that acts  
as a SIP proxy and registrar. Here "p2psipNetwork.org" is the Overlay  
Identifier, and "voicemail" indicates the service.  (This is one way  
we might name services in a P2PSIP overlay.)
2. Peer Y publishes this name into the Overlay using the peer  
protocol. If X does not have a public IP address, then X is reachable  
only by sending a message through the Overlay to Y, which then  
proxies it on to X, so peer Y needs to change the contact URI for  
this registration to be itself (or add itself as an alternative  
contact for this service).
3. A peer Z which wishes to use the service does a lookup on  
"sip:voicemail.p2psipNetwork.org" and gets back a bunch of contact URIs.
4. Peer Z decides that it wishes to use the service offered by X.
5. Peer Z sends an Invite to X. This Invite is actually sent to Y,  
since Y changed the contact URI in the registrar to point at itself.
6. Peer Y receives the Invite and proxies it on to X. The call is  
established, and X and Z negotiate the details of the voicemail  
service that X will offer Z. Z asks X to handle voicemail on behalf  
of "alice at p2psipNetwork.org"
7. Peer X registers "alice at p2psipNetwork.org" with Y (with a Q value  
of .5, so this contact is less preferred than other contacts for  
Alice). Y publishes this contact into the Overlay.

To summarize, I think that it is possible for us to allow clients and  
C/S SIP entities to offer at least certain services, using the same  
mechanisms as used for user registration.

Comments?

- Philip



From philip_matthews at magma.ca  Thu Dec 21 17:43:53 2006
From: philip_matthews at magma.ca (Philip Matthews)
Date: Thu, 21 Dec 2006 17:43:53 -0500
Subject: [p2p-sip] Concepts Draft: Open Issue #4: Can clients offer
	services?
In-Reply-To: <4589DAC8.2070900@telecomitalia.it>
References: <2043E08B-FCB3-424A-B23C-B06BA0F65532@magma.ca>
	<4589DAC8.2070900@telecomitalia.it>
Message-ID: <0F1E475D-9B7F-4823-BFD3-177CEC9A0516@magma.ca>


On 20-Dec-06, at 19:52 , Enrico Marocco wrote:

> Philip Matthews wrote:
>> Open Issue #4: Can clients offer services? Or are services only
>> offered by peers?
>>
>> Philip's Strawman Proposal: Only peers can offer services. Clients
>> can use services, but  do not offer them.
>>
>> There are two reasons for this proposal:
>> One is that, to offer a service, a client would presumably need some
>> sort of ID in order to identify the client offering the service, and
>> the Concepts Draft currently states clients do not have IDs.
>> The second is that this is consistent with the view that the client
>> protocol could be just vanilla SIP. It is not clear how to offer
>> services if the client protocol is SIP.
>
> I think it would be a big mistake to take such a decision now.   
> First of
> all, we don't really know what a service is (anybody subscribed to the
> sipping mailing list knows what I mean); it could be a conference  
> server
> identified by a common SIP URL, or a STUN or TEREDO relay,  
> identified by
> a URI following a non-yet-defined schema.  Then we neither actually  
> know
> what a peer is; possibly a device behind a NAT could be a perfect
> voicemail server (i.e. could host a service) while being unable to
> participate in the DHT maintainance because of network limitations.
>
> Such an issue is too technical for being in the scope of the  
> concept draft.
>
>> The authors of the Concepts Draft have not been able to resolve this
>> amongst ourselves, so we are throwing the issue out for general
>> debate. (Err ... to be honest, we haven't tried very hard either, as
>> people are busy with other things).
>
> I don't see the need to address such an issue in the concept draft;
> again, it is a merely technical decision and perhaps it is worth
> deferring it to the deployer's will.


Enrico:

If we do service registration in the same way as we do user  
registration (see the e-mail I just sent to the list), then I agree  
with you that we can defer this question.

If we want to use a different mechanism (e.g. a mechanism that can  
not be implemented using standard SIP and/or requires clients to have  
unique client IDs), then I think this question will interact with  
other design decisions and we need to consider it now.

- Philip


From enrico.marocco at telecomitalia.it  Thu Dec 21 18:27:30 2006
From: enrico.marocco at telecomitalia.it (Enrico Marocco)
Date: Fri, 22 Dec 2006 00:27:30 +0100
Subject: [p2p-sip] Concepts Draft: Open Issue #4: Can clients offer
 services?
In-Reply-To: <0F1E475D-9B7F-4823-BFD3-177CEC9A0516@magma.ca>
References: <2043E08B-FCB3-424A-B23C-B06BA0F65532@magma.ca>
	<4589DAC8.2070900@telecomitalia.it>
	<0F1E475D-9B7F-4823-BFD3-177CEC9A0516@magma.ca>
Message-ID: <458B1862.1070101@telecomitalia.it>

Philip Matthews wrote:
> If we do service registration in the same way as we do user registration
> (see the e-mail I just sent to the list), then I agree with you that we
> can defer this question.
> 
> If we want to use a different mechanism (e.g. a mechanism that can not
> be implemented using standard SIP and/or requires clients to have unique
> client IDs), then I think this question will interact with other design
> decisions and we need to consider it now.

I get your point, but I still think that forcing such kind of decision
now would be a mistake.  Talking about services, we should now only
define the _concept_; how it will be registered, stored and accessed
within the overlay is technical matter and should be separate from the
concept.

I'd propose to address this issue starting with some general text, such
as: "Sevices are identified by specific URIs and possibly by other
means, can be accessed by clients and peers and can be offered by peers
or by clients and peers."  Then, if when working on protocols
specifications we realize that such a definition is too vague, we'll
still be able to replace it with a more restrictive one.  Doing the
contrary is dangerous because it would discourage the proposal of
possibly valid solutions just because they are not compliant with a too
strict concept draft.

My two cents.

-- 
Ciao,
Enrico

From RADHIKA.R.ROY at saic.com  Thu Dec 21 18:47:04 2006
From: RADHIKA.R.ROY at saic.com (Roy, Radhika R.)
Date: Thu, 21 Dec 2006 18:47:04 -0500
Subject: [p2p-sip] Concepts Draft: Open Issue #4: Can clients
	offerservices?
References: <2043E08B-FCB3-424A-B23C-B06BA0F65532@magma.ca>
	<56B6DA37-D085-44F2-8173-0A321D422196@magma.ca>
Message-ID: <62D92A9A02BCC845B202323D49A48426E1CF6D@0591-ITS-EXMP02.us.saic.com>

Philip:
 
URL does not represents any services. A URL is a url. That URL is a resource used like an address.
 
So, just using "voicemail" in the URL does not mean that the "voice services" are available in this address.
 
In SIP, there is no standard for any services. It just sets up a session: p2p or mpt.
 
I do not think that you can standardize any services in P2P-SIP as well.
 
Best regards,
Radhika

________________________________

From: p2p-sip-bounces at cs.columbia.edu on behalf of Philip Matthews
Sent: Thu 12/21/2006 5:40 PM
To: P2P-SIP
Subject: Re: [p2p-sip] Concepts Draft: Open Issue #4: Can clients offerservices?





On 20-Dec-06, at 14:30 , Philip Matthews wrote:

> Open Issue #4: Can clients offer services? Or are services only 
> offered by peers?
>
> Philip's Strawman Proposal: Only peers can offer services. Clients 
> can use services, but  do not offer them.
>
> There are two reasons for this proposal:
> One is that, to offer a service, a client would presumably need 
> some sort of ID in order to identify the client offering the 
> service, and the Concepts Draft currently states clients do not 
> have IDs.
> The second is that this is consistent with the view that the client 
> protocol could be just vanilla SIP. It is not clear how to offer 
> services if the client protocol is SIP.
>
> The authors of the Concepts Draft have not been able to resolve 
> this amongst ourselves, so we are throwing the issue out for 
> general debate. (Err ... to be honest, we haven't tried very hard 
> either, as people are busy with other things).
>
> - Philip
>

Folks:

I have thought about this issue some more, and would like to change 
some of the things I said above.

I now think it is possible to treat service registration in a way 
similar to user registration. In particular, if clients offered 
services, this could be done using the same techniques as user 
registration.

IF this technique is acceptable,  then services can be offered using 
standard SIP mechanisms, and this can be done without requiring a 
client to have an ID. This applies at least to services that are 
invokable using a SIP Invite.


Below is an example of how a C/S SIP UA might offer voicemail 
services to a P2PSIP Overlay:
1. The C/S SIP UA (call it X) registers the URI  
"sip:voicemail.p2psipNetwork.org" with a peer (call it Y) that acts 
as a SIP proxy and registrar. Here "p2psipNetwork.org" is the Overlay 
Identifier, and "voicemail" indicates the service.  (This is one way 
we might name services in a P2PSIP overlay.)
2. Peer Y publishes this name into the Overlay using the peer 
protocol. If X does not have a public IP address, then X is reachable 
only by sending a message through the Overlay to Y, which then 
proxies it on to X, so peer Y needs to change the contact URI for 
this registration to be itself (or add itself as an alternative 
contact for this service).
3. A peer Z which wishes to use the service does a lookup on 
"sip:voicemail.p2psipNetwork.org" and gets back a bunch of contact URIs.
4. Peer Z decides that it wishes to use the service offered by X.
5. Peer Z sends an Invite to X. This Invite is actually sent to Y, 
since Y changed the contact URI in the registrar to point at itself.
6. Peer Y receives the Invite and proxies it on to X. The call is 
established, and X and Z negotiate the details of the voicemail 
service that X will offer Z. Z asks X to handle voicemail on behalf 
of "alice at p2psipNetwork.org"
7. Peer X registers "alice at p2psipNetwork.org" with Y (with a Q value 
of .5, so this contact is less preferred than other contacts for 
Alice). Y publishes this contact into the Overlay.

To summarize, I think that it is possible for us to allow clients and 
C/S SIP entities to offer at least certain services, using the same 
mechanisms as used for user registration.

Comments?

- Philip


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


-------------- next part --------------
An HTML attachment was scrubbed...
URL: https://lists.cs.columbia.edu/pipermail/p2p-sip/attachments/20061221/aa30dd07/attachment-0001.html

From RADHIKA.R.ROY at saic.com  Thu Dec 21 18:56:55 2006
From: RADHIKA.R.ROY at saic.com (Roy, Radhika R.)
Date: Thu, 21 Dec 2006 18:56:55 -0500
Subject: [p2p-sip] Concepts Draft: Open Issue #4: Can clients offer
	services?
References: <2043E08B-FCB3-424A-B23C-B06BA0F65532@magma.ca><4589DAC8.2070900@telecomitalia.it><0F1E475D-9B7F-4823-BFD3-177CEC9A0516@magma.ca>
	<458B1862.1070101@telecomitalia.it>
Message-ID: <62D92A9A02BCC845B202323D49A48426E1CF6E@0591-ITS-EXMP02.us.saic.com>

Enrico:
 
You can see a separate email sent to Philip why you cannot specify services using URLs. There are other standards in IETF for services registration. Let us use those standards.
 
It seems to be easy to say like this - using URL for services registration. SIPPING WG also debated this for sometime to clarify this mis-concepts. But in practice, a lot of works have been done in IETF in this direction for services registration and how to use and get services. 
 
Standards are already there in the IETF and we need to see those standards first before re-inventing the "services" wheel! It will simply be wastage of our energy and time at this point of time for P2P-SIP as people debated for years for creation of services registration standards.
 
Best regards,
Radhika

________________________________

From: p2p-sip-bounces at cs.columbia.edu on behalf of Enrico Marocco
Sent: Thu 12/21/2006 6:27 PM
To: Philip Matthews
Cc: P2P-SIP; David A. Bryan
Subject: Re: [p2p-sip] Concepts Draft: Open Issue #4: Can clients offer services?



Philip Matthews wrote:
> If we do service registration in the same way as we do user registration
> (see the e-mail I just sent to the list), then I agree with you that we
> can defer this question.
>
> If we want to use a different mechanism (e.g. a mechanism that can not
> be implemented using standard SIP and/or requires clients to have unique
> client IDs), then I think this question will interact with other design
> decisions and we need to consider it now.

I get your point, but I still think that forcing such kind of decision
now would be a mistake.  Talking about services, we should now only
define the _concept_; how it will be registered, stored and accessed
within the overlay is technical matter and should be separate from the
concept.

I'd propose to address this issue starting with some general text, such
as: "Sevices are identified by specific URIs and possibly by other
means, can be accessed by clients and peers and can be offered by peers
or by clients and peers."  Then, if when working on protocols
specifications we realize that such a definition is too vague, we'll
still be able to replace it with a more restrictive one.  Doing the
contrary is dangerous because it would discourage the proposal of
possibly valid solutions just because they are not compliant with a too
strict concept draft.

My two cents.

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


-------------- next part --------------
An HTML attachment was scrubbed...
URL: https://lists.cs.columbia.edu/pipermail/p2p-sip/attachments/20061221/411036d3/attachment.html

From philip_matthews at magma.ca  Thu Dec 21 19:29:23 2006
From: philip_matthews at magma.ca (Philip Matthews)
Date: Thu, 21 Dec 2006 19:29:23 -0500
Subject: [p2p-sip] Concepts Draft: Open Issue #4:
	Canclientsofferservices?
In-Reply-To: <0cce01c72539$d69a79c0$640fa8c0@cis.neustar.com>
References: <0cce01c72539$d69a79c0$640fa8c0@cis.neustar.com>
Message-ID: <9F2C8AEE-C4D3-40C5-8B81-458CDADAED1D@magma.ca>


On 21-Dec-06, at 14:54 , Brian Rosen wrote:

> Okay, so we just used this argument with the "DNS is all we need"  
> folks, and
> now I'm on the other side a little.  This view would NOT conflict  
> with the
> charter, I would note.
>
> I have been bothered by the special status we give to peers over  
> clients.
> We need to have two kinds of entities for one very good reason:  
> clients
> can't talk to other clients (or peers) directly, primarily because  
> of NAT
> issues.
>
> However, we have taken that very real problem, and, for some reason  
> I don't
> understand, decided that SINCE you can't exchange messages  
> directly, THEN
> peers should have other distinguishing characteristics from  
> clients.  That's
> the part I don't get.  Why is that?  Why should being behind a NAT
> automatically make you a dumb client instead of a smart peer?  Why  
> can't
> clients store data for example?  In this case, why can't they provide
> services?
>
> What is the requirement, or the practical reality, that makes the  
> difference
> between a client and a peer more than just a NAT traversal problem?
>
> Brian
>
> Who still admits to not being an expert here and thus gets to ask  
> the stupid
> questions that sometimes generates "aha's" from those that are.
>

Brian:

According to the Concepts draft, peers CAN be located behind NATs,  
and the difference between clients and peers is simply that peers  
support storage and routing, while clients do not.

However, recently I have started to wonder if, rather than the black- 
and-white peer vs. client distinction, we don't want instead the  
concept of more- vs. less-capable peers. That is, some peers have  
more memory and can store more registrations (think softphone on a  
computer with a large disk), while other peers have less memory and  
can store fewer registrations (think hardphone with limited memory  
and no disk). Similarly, some peers have higher bandwidth  
connections  and can forward more messages on behave of others (think  
PC with gigabit ethernet connection), while others have less  
bandwidth and can forward fewer (think cell phone with limited  
battery lifetime).

Clients then become an extreme point in this space: a peer that  
cannot store any registrations and cannot forward any messages.

- Philip

PS. Just so you know, I was not the person who asked for the concept  
of clients, and I too am still trying to understand the use-cases for  
them.


From philip_matthews at magma.ca  Thu Dec 21 19:51:04 2006
From: philip_matthews at magma.ca (Philip Matthews)
Date: Thu, 21 Dec 2006 19:51:04 -0500
Subject: [p2p-sip] Concepts Draft: Open Issue #4: Can clients offer
	services?
In-Reply-To: <458B1862.1070101@telecomitalia.it>
References: <2043E08B-FCB3-424A-B23C-B06BA0F65532@magma.ca>
	<4589DAC8.2070900@telecomitalia.it>
	<0F1E475D-9B7F-4823-BFD3-177CEC9A0516@magma.ca>
	<458B1862.1070101@telecomitalia.it>
Message-ID: <BDBB298D-CC86-4262-AD69-979E2A4A306B@magma.ca>


On 21-Dec-06, at 18:27 , Enrico Marocco wrote:

> Philip Matthews wrote:
>> If we do service registration in the same way as we do user  
>> registration
>> (see the e-mail I just sent to the list), then I agree with you  
>> that we
>> can defer this question.
>>
>> If we want to use a different mechanism (e.g. a mechanism that can  
>> not
>> be implemented using standard SIP and/or requires clients to have  
>> unique
>> client IDs), then I think this question will interact with other  
>> design
>> decisions and we need to consider it now.
>
> I get your point, but I still think that forcing such kind of decision
> now would be a mistake.  Talking about services, we should now only
> define the _concept_; how it will be registered, stored and accessed
> within the overlay is technical matter and should be separate from the
> concept.
>
> I'd propose to address this issue starting with some general text,  
> such
> as: "Sevices are identified by specific URIs and possibly by other
> means, can be accessed by clients and peers and can be offered by  
> peers
> or by clients and peers."  Then, if when working on protocols
> specifications we realize that such a definition is too vague, we'll
> still be able to replace it with a more restrictive one.  Doing the
> contrary is dangerous because it would discourage the proposal of
> possibly valid solutions just because they are not compliant with a  
> too
> strict concept draft.
>
> My two cents.
>
> -- 
> Ciao,
> Enrico

You are right. We could limit the discussion of service lookup in the  
next revision of the draft to something as general as this.

It would be nice to add an example of service lookup in section 3.4,  
and I suggest the example be the lookup of a STUN Server service, but  
we could do this without going much beyond what you said above.

Anyone else have comments on Enrico's proposal?

- Philip

PS. We still need some definition of a "service". That is going to be  
my next open issue ...


From philip_matthews at magma.ca  Thu Dec 21 19:59:39 2006
From: philip_matthews at magma.ca (Philip Matthews)
Date: Thu, 21 Dec 2006 19:59:39 -0500
Subject: [p2p-sip] Concepts Draft: Open Issue #4: Can clients offer
	services?
In-Reply-To: <62D92A9A02BCC845B202323D49A48426E1CF6E@0591-ITS-EXMP02.us.saic.com>
References: <2043E08B-FCB3-424A-B23C-B06BA0F65532@magma.ca><4589DAC8.2070900@telecomitalia.it><0F1E475D-9B7F-4823-BFD3-177CEC9A0516@magma.ca>
	<458B1862.1070101@telecomitalia.it>
	<62D92A9A02BCC845B202323D49A48426E1CF6E@0591-ITS-EXMP02.us.saic.com>
Message-ID: <873F591A-ACF6-470D-988E-D3E4F2A7F537@magma.ca>


On 21-Dec-06, at 18:56 , Roy, Radhika R. wrote:

> Philip:
>
> URL does not represents any services. A URL is a url. That URL is a  
> resource used like an address.
>
> So, just using "voicemail" in the URL does not mean that the "voice  
> services" are available in this address.
>
> In SIP, there is no standard for any services. It just sets up a  
> session: p2p or mpt.
>
> I do not think that you can standardize any services in P2P-SIP as  
> well.
>
> Best regards,
> Radhika

and

> Enrico:
>
> You can see a separate email sent to Philip why you cannot specify  
> services using URLs. There are other standards in IETF for services  
> registration. Let us use those standards.
>
> It seems to be easy to say like this - using URL for services  
> registration. SIPPING WG also debated this for sometime to clarify  
> this mis-concepts. But in practice, a lot of works have been done  
> in IETF in this direction for services registration and how to use  
> and get services.
>
> Standards are already there in the IETF and we need to see those  
> standards first before re-inventing the "services" wheel! It will  
> simply be wastage of our energy and time at this point of time for  
> P2P-SIP as people debated for years for creation of services  
> registration standards.
>
> Best regards,
> Radhika

Radhika:

I am not familiar with these discussions -- perhaps you could  
summarize them for us?

- Philip

-------------- next part --------------
An HTML attachment was scrubbed...
URL: https://lists.cs.columbia.edu/pipermail/p2p-sip/attachments/20061221/2a6f1f8e/attachment.html

From edpimentl at gmail.com  Thu Dec 21 20:06:08 2006
From: edpimentl at gmail.com (EdPimentl)
Date: Thu, 21 Dec 2006 20:06:08 -0500
Subject: [p2p-sip] Concepts Draft: Open Issue #4: Can clients offer
	services?
In-Reply-To: <BDBB298D-CC86-4262-AD69-979E2A4A306B@magma.ca>
References: <2043E08B-FCB3-424A-B23C-B06BA0F65532@magma.ca>
	<4589DAC8.2070900@telecomitalia.it>
	<0F1E475D-9B7F-4823-BFD3-177CEC9A0516@magma.ca>
	<458B1862.1070101@telecomitalia.it>
	<BDBB298D-CC86-4262-AD69-979E2A4A306B@magma.ca>
Message-ID: <9dc4a1670612211706q503f5eact22bc637639699da0@mail.gmail.com>

These are all excellent discussions on services..
I like to submit to look and borrow from the guiding principles supporting
XML WebServices SOBA (Service OrientedBusiness Applications)
Where these are automagically orchestrated and choreograph and essentially
programatically established a business contract using the above dejure
standards..

Sincerely and Best Regards,
-Ed

Mail:       edpimentl[at]gmail[dot]com
VoIP:      edpimentl [Skype | GoogleTalk]
IM:          edpimentl [ AOL | Jabber | Yahoo | MSN ]

Mobile Content Management
http://mobilecentral.ws

Mobile Social Networks(Private Label)
http://GooGaYa.com

Business At the Speed of Thought
http://biscentral.biz

Mobile Payment
http://AgilePAY.ws
-------------- next part --------------
An HTML attachment was scrubbed...
URL: https://lists.cs.columbia.edu/pipermail/p2p-sip/attachments/20061221/51e2aee0/attachment-0001.html

From RADHIKA.R.ROY at saic.com  Thu Dec 21 22:42:49 2006
From: RADHIKA.R.ROY at saic.com (Roy, Radhika R.)
Date: Thu, 21 Dec 2006 22:42:49 -0500
Subject: [p2p-sip] Concepts Draft: Open Issue #4: Can clients offer
	services?
References: <2043E08B-FCB3-424A-B23C-B06BA0F65532@magma.ca><4589DAC8.2070900@telecomitalia.it><0F1E475D-9B7F-4823-BFD3-177CEC9A0516@magma.ca>
	<458B1862.1070101@telecomitalia.it>
	<62D92A9A02BCC845B202323D49A48426E1CF6E@0591-ITS-EXMP02.us.saic.com>
	<873F591A-ACF6-470D-988E-D3E4F2A7F537@magma.ca>
Message-ID: <62D92A9A02BCC845B202323D49A48426E1CF6F@0591-ITS-EXMP02.us.saic.com>

Philip:
 
That's it what has been explained in my emails.
 
SLP is the protocol in the IETF for services registration, services query, and services to be invoked. Other than this, there is no other standards for services in the IETF.
 
By the way, SLP is not p2p protocol per se.
 
Best regards,
Radhika
 
PS.
 
1. If people think, p2p-SIP protocol is a service, then it is a different story. 
 
2. Some people also think routing services, QOS services, NAT/FW crossing services, VoIP services, security services, voice mail services, ptp/mpt conferencing services, and others as a part of services. In this case, we will be mixing between "apples" and "oranges." That is, we confusing the protocol features and (application) services.
 
3. Some people want to register voice mail services. SLP is there to do so. That's it.
 
4. People need to see URL standards what they are. In URL, there is no connotation of services. Period.
 
5. There is no standard in SIP for services. Period. SIP just provides standardization how to set up sessions. All the features that are invoked in SIP during the session are the requirements to set up the session between the parties satisfying the needs of setting up a session successfully after negotiations that all parties agree on.
 
6. People should examine item 5 very carefully going through all SIP standards/RFCs and must NOT confuse between session setups after negotiations to have all the features to satisfy the needs, protocol features, and services like voice mail.
 
7. There is a very fine line between items 5 and 6, for example, how voice mail service/server, media servers for media bridging, and other servers are brought as a part of the session establishment. It is not a "crude" thing like register a voice mail service using URL as an indicator of the service and then just use those services as a part of the session. If people think so, they are highly confused after so many years SIP standardization in the IETF.
 
8. Like SIP, p2p-SIP also follow the same standards. They will have no scope for standardization of services like voice servers, etc.
 
9. What p2p-SIP needs to focus on is to provide all features and functionalities in setting up the sessions (ptp and mpt) in just peer-to-peer mode and, like SIP, not to standardize any "services." (If people think that SIP is providing services by invoking voice mail servers/services, then again they are HIGHLY confused.)
 
10. For example, SIP session has a need for audio and video codec and even announcement (for example, pre-paid call where the caller needs to be made aware of the payment or for other reasons) as a part of the session. It is true that an announcement application server is needed to provide announce. Here is a fine line. In the session establishment layer, there is no knowledge that there is an announcement application server that provides the announcement services. The entity that has the knowledge of this service may be a SIP proxy/call control/application server. Here is the catch: How does the SIP proxy/call control/application server will have this knowledge of this announcement application server? In any case, it is not the SIP protocol. For example, how a DNS service is invoked for resolving an IP address from a SIP URL? It is not a SIP protocol.

________________________________

From: Philip Matthews [mailto:philip_matthews at magma.ca]
Sent: Thu 12/21/2006 7:59 PM
To: Roy, Radhika R.
Cc: P2P-SIP
Subject: Re: [p2p-sip] Concepts Draft: Open Issue #4: Can clients offer services?



On 21-Dec-06, at 18:56 , Roy, Radhika R. wrote:


	Philip:

	 

	URL does not represents any services. A URL is a url. That URL is a resource used like an address.

	 

	So, just using "voicemail" in the URL does not mean that the "voice services" are available in this address.

	 

	In SIP, there is no standard for any services. It just sets up a session: p2p or mpt.

	 

	I do not think that you can standardize any services in P2P-SIP as well.

	 

	Best regards,
	Radhika


and


	Enrico:
	 
	You can see a separate email sent to Philip why you cannot specify services using URLs. There are other standards in IETF for services registration. Let us use those standards.
	 
	It seems to be easy to say like this - using URL for services registration. SIPPING WG also debated this for sometime to clarify this mis-concepts. But in practice, a lot of works have been done in IETF in this direction for services registration and how to use and get services. 
	 
	Standards are already there in the IETF and we need to see those standards first before re-inventing the "services" wheel! It will simply be wastage of our energy and time at this point of time for P2P-SIP as people debated for years for creation of services registration standards.
	
	 
	Best regards,
	Radhika


Radhika:

I am not familiar with these discussions -- perhaps you could summarize them for us?

- Philip

-------------- next part --------------
An HTML attachment was scrubbed...
URL: https://lists.cs.columbia.edu/pipermail/p2p-sip/attachments/20061221/8206ca6f/attachment.html

From RADHIKA.R.ROY at saic.com  Thu Dec 21 22:46:08 2006
From: RADHIKA.R.ROY at saic.com (Roy, Radhika R.)
Date: Thu, 21 Dec 2006 22:46:08 -0500
Subject: [p2p-sip] Concepts Draft: Open Issue #4: Can clients
	offerservices?
References: <2043E08B-FCB3-424A-B23C-B06BA0F65532@magma.ca><4589DAC8.2070900@telecomitalia.it><0F1E475D-9B7F-4823-BFD3-177CEC9A0516@magma.ca><458B1862.1070101@telecomitalia.it><BDBB298D-CC86-4262-AD69-979E2A4A306B@magma.ca>
	<9dc4a1670612211706q503f5eact22bc637639699da0@mail.gmail.com>
Message-ID: <62D92A9A02BCC845B202323D49A48426E1CF70@0591-ITS-EXMP02.us.saic.com>

Ed:
 
What is the point from standard point of view?
 
Do you want LAN services? IP services? QOS services? Security services? Web services? Routing services? Bandwidth services? Voicemail services? Multicast services? Multimedia conferencing services? Email services? Brokering services? Session establishment services? HTML-XML conversion services? .....?????
 
Best regards,
Radhika

________________________________

From: p2p-sip-bounces at cs.columbia.edu on behalf of EdPimentl
Sent: Thu 12/21/2006 8:06 PM
To: Philip Matthews
Cc: P2P-SIP
Subject: Re: [p2p-sip] Concepts Draft: Open Issue #4: Can clients offerservices?


These are all excellent discussions on services..
I like to submit to look and borrow from the guiding principles supporting
XML WebServices SOBA (Service OrientedBusiness Applications)
Where these are automagically orchestrated and choreograph and essentially programatically established a business contract using the above dejure standards.. 

Sincerely and Best Regards,
-Ed

Mail:       edpimentl[at]gmail[dot]com
VoIP:      edpimentl [Skype | GoogleTalk]
IM:          edpimentl [ AOL | Jabber | Yahoo | MSN ]

Mobile Content Management 
http://mobilecentral.ws <http://mobilecentral.ws/> 

Mobile Social Networks(Private Label)
http://GooGaYa.com <http://googaya.com/> 

Business At the Speed of Thought
http://biscentral.biz <http://biscentral.biz/> 

Mobile Payment
http://AgilePAY.ws <http://agilepay.ws/>  
-------------- next part --------------
An HTML attachment was scrubbed...
URL: https://lists.cs.columbia.edu/pipermail/p2p-sip/attachments/20061221/d1984131/attachment.html

From hsinnrei at adobe.com  Fri Dec 22 12:01:40 2006
From: hsinnrei at adobe.com (Henry Sinnreich)
Date: Fri, 22 Dec 2006 09:01:40 -0800
Subject: [p2p-sip] Concepts Draft: Open Issue #4: Can clients offer
	services?
In-Reply-To: <9dc4a1670612211706q503f5eact22bc637639699da0@mail.gmail.com>
Message-ID: <24CCCC428EFEA2469BF046DB3C7A8D22141456@namail5.corp.adobe.com>

Ed,

 

These are excellent technologies, but a distraction from the present
charter. We must walk first before we run.

 

Besides, we badly need to revive the old good habit in the IETF of
having running code before working at a standard and debating all kinds
of ideas in this list.

 

Not to mention open source code as well.

 

We can keep SOBA on the back of our minds until someone is willing to
spend their money and/or their time on a prototype before coming to the
WG (which it is not even yet).

 

Please y'all help us stay focused and get something accomplished. In
defining _Basic_ P2P SIP.

 

As for defining URLs for P2P SIP standards, can we please defer this
until the participants on this list can make a P2P SIP call?

 

That is, unless you want to keep using Skype forever :-)

 

Happy Holidays!

 

Henry

 

________________________________

From: EdPimentl [mailto:edpimentl at gmail.com] 
Sent: Thursday, December 21, 2006 5:06 PM
To: Philip Matthews
Cc: Enrico Marocco; P2P-SIP; Henry Sinnreich
Subject: Re: [p2p-sip] Concepts Draft: Open Issue #4: Can clients offer
services?

 

These are all excellent discussions on services..
I like to submit to look and borrow from the guiding principles
supporting
XML WebServices SOBA (Service OrientedBusiness Applications)
Where these are automagically orchestrated and choreograph and
essentially programatically established a business contract using the
above dejure standards.. 

Sincerely and Best Regards,
-Ed

Mail:       edpimentl[at]gmail[dot]com
VoIP:      edpimentl [Skype | GoogleTalk]
IM:          edpimentl [ AOL | Jabber | Yahoo | MSN ]

Mobile Content Management 
http://mobilecentral.ws

Mobile Social Networks(Private Label)
http://GooGaYa.com

Business At the Speed of Thought
http://biscentral.biz

Mobile Payment
http://AgilePAY.ws 

-------------- next part --------------
An HTML attachment was scrubbed...
URL: https://lists.cs.columbia.edu/pipermail/p2p-sip/attachments/20061222/0774155f/attachment-0001.html

From edpimentl at gmail.com  Fri Dec 22 12:48:01 2006
From: edpimentl at gmail.com (EdPimentl)
Date: Fri, 22 Dec 2006 12:48:01 -0500
Subject: [p2p-sip] Concepts Draft: Open Issue #4: Can clients offer
	services?
In-Reply-To: <24CCCC428EFEA2469BF046DB3C7A8D22141456@namail5.corp.adobe.com>
References: <9dc4a1670612211706q503f5eact22bc637639699da0@mail.gmail.com>
	<24CCCC428EFEA2469BF046DB3C7A8D22141456@namail5.corp.adobe.com>
Message-ID: <9dc4a1670612220948x4d4144ffi16be7a0c965aa116@mail.gmail.com>

Hello Henry, Et Al,

I am in vehement agreement with your insightful feedback.
To that end I am willing to devote a resource to flesh out open-source code
which embodies the IETF current draft...
Then other like mind and join in....

For clarity I am in favor of a "kinder, gentler, 1000 point of light" open
p2p-sip infrastructure.

Best to all,
-E

On 12/22/06, Henry Sinnreich <hsinnrei at adobe.com> wrote:
>
>  Ed,
>
>
>
> These are excellent technologies, but a distraction from the present
> charter. We must walk first before we run.
>
>
>
> Besides, we badly need to revive the old good habit in the IETF of having
> running code before working at a standard and debating all kinds of ideas in
> this list.
>
>
>
> Not to mention open source code as well.
>
>
>
> We can keep SOBA on the back of our minds until someone is willing to
> spend their money and/or their time on a prototype before coming to the WG
> (which it is not even yet).
>
>
>
> Please y'all help us stay focused and get something accomplished. In
> defining _*Basic*_ P2P SIP.
>
>
>
> As for defining URLs for P2P SIP standards, can we please defer this until
> the participants on this list can make a P2P SIP call?
>
>
>
> That is, unless you want to keep using Skype forever J
>
>
>
> Happy Holidays!
>
>
>
> Henry
>
>
>  ------------------------------
>
> *From:* EdPimentl [mailto:edpimentl at gmail.com]
> *Sent:* Thursday, December 21, 2006 5:06 PM
> *To:* Philip Matthews
> *Cc:* Enrico Marocco; P2P-SIP; Henry Sinnreich
> *Subject:* Re: [p2p-sip] Concepts Draft: Open Issue #4: Can clients offer
> services?
>
>
>
> These are all excellent discussions on services..
> I like to submit to look and borrow from the guiding principles supporting
> XML WebServices SOBA (Service OrientedBusiness Applications)
> Where these are automagically orchestrated and choreograph and essentially
> programatically established a business contract using the above dejure
> standards..
>
> Sincerely and Best Regards,
> -Ed
>
> Mail:       edpimentl[at]gmail[dot]com
> VoIP:      edpimentl [Skype | GoogleTalk]
> IM:          edpimentl [ AOL | Jabber | Yahoo | MSN ]
>
> Mobile Content Management
> http://mobilecentral.ws
>
> Mobile Social Networks(Private Label)
> http://GooGaYa.com
>
> Business At the Speed of Thought
> http://biscentral.biz
>
> Mobile Payment
> http://AgilePAY.ws
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: https://lists.cs.columbia.edu/pipermail/p2p-sip/attachments/20061222/f709db77/attachment.html

From RADHIKA.R.ROY at saic.com  Sat Dec 23 12:41:35 2006
From: RADHIKA.R.ROY at saic.com (Roy, Radhika R.)
Date: Sat, 23 Dec 2006 12:41:35 -0500
Subject: [p2p-sip] Concepts Draft: Open Issue #4: Can clients
	offerservices?
References: <2043E08B-FCB3-424A-B23C-B06BA0F65532@magma.ca><4589DAC8.2070900@telecomitalia.it><0F1E475D-9B7F-4823-BFD3-177CEC9A0516@magma.ca><458B1862.1070101@telecomitalia.it><BDBB298D-CC86-4262-AD69-979E2A4A306B@magma.ca><9dc4a1670612211706q503f5eact22bc637639699da0@mail.gmail.com><62D92A9A02BCC845B202323D49A48426E1CF70@0591-ITS-EXMP02.us.saic.com>
	<9dc4a1670612220849t39674d6obbd3aad36f9ae63a@mail.gmail.com>
Message-ID: <62D92A9A02BCC845B202323D49A48426E1CF74@0591-ITS-EXMP02.us.saic.com>

Ed:
 
I appreciate your email.
 
These kinds of things were discussed before. It is nothing new.
 
These are NOT standards to use URLs with service connotations. These are used defying standards to do some applications here and there having some like-minded people.
 
If billions of people start using with billions and billions of services with those services names specifying URLs, nothing will left in this universe that will not be termed as "services."
 
That is why, we go for standardization bringing sanity throughout the whole world.
 
If people feel so strongly for using URLs specifying services, why do they need to go via back door "violating" standards?
 
Rather, they will be welcomed to make their case and create standards like this. It is so simple to do so, but there must not be violation of standards.
 
Best regards,
Radhika
 
PS: Please analyze the XML script that has been used as an example by you.
 
1. The word SMS is a service known as "standard." This SMS has been used to qualify the URL. So, it is perfectly alright because it is not the URL itself that specifies services.
 
2. If you want to build features as shown in you example, it will NOT violate URL standard so far I think.
 
3. For all other features that you have described in the example, we have to examine in terms of the scope of the standards. For sending SMS to Henry means invocation of the SMS application service probably using the SMS application server. Whether we are allowed to invoke other services "directly" in p2p-SIP layer is debatable as a matter of standardization. I would say no. For example, in SIP, the invocation of SMS application is not a part of the session establishment. A call control SIP proxy might have some extra intelligent what to do next if the Henry's phone is busy. 
 
3a. Here is the most important fine line: To invoke SMS is called another service initiation and it is not done in the SIP layer. That is, in service creation/execution environment, an intelligent proxy (that has the knowledge beyond SIP layer) may be able to invoke SMS application. If so, it is called service creation/execution capability of the proxy server which is NOT in the SIP layer.
 
3b. Now, you have got the whole new services paradigm - Services Creation/Execution Environment. In fact, SIP has created a whole application services industry where all those application servers are being invoked as a side effects of the SIP session establishment: Parlay, CORBA, Open Source, OMA, just to mention a few. This is not in the SIP layer.
 
4. Conclusion: The example that you have provided it should not be for the p2p-SIP layer. It will be a perfect example for p2p-SIP services creation/execution environments that reside above the p2p-SIP Layer.

________________________________

From: EdPimentl [mailto:edpimentl at gmail.com]
Sent: Fri 12/22/2006 11:49 AM
To: Roy, Radhika R.
Cc: Philip Matthews; P2P-SIP
Subject: Re: [p2p-sip] Concepts Draft: Open Issue #4: Can clients offerservices?


Hello Roy,

I read your previous post and agree with many of your comments.
We respectfully differ on your URL is just a url.... since today a URL  can be equivalent to
the hard coded/brittle closely couple CORBA like "SERVICES" offer by the unwilling (telcos)partners. 

Regarding the standard point is that we can "consume" and "produce" Network Services, Applications (Note VoIP is an Application over IP) Services, Payment Services...
In fact the recent enormous investment in Web2.0 and MobileWeb type companies are companies
that "orchestrate, choreograph, compose service which originate in another company/enterprise/client service catalog...
ParlayX is based upon that very same premise... and I am not advocating using it in a P2P-SIP but pointing out that P2P-SIP can benefit/borrow the guiding principles of such standards. 

Consider this... SMS on Busy P2P-SIP Subscriber
<p2p-sip_svc>
   <register>
   <disconnected causeCode="CAUSE_BUSY"
     destination=tel:5165551212"                      // comment IE Henry's number :-) 
    connection="terminating" block="false">
 <getOriginatingAddress address="orig"/>
<sendSMS to="sip://henry at adobe.com" 
   <content=" ' call attempt by '+%orig;+'on 5165551212'"/>
</disconnected>
</register>
<p2p-sip_svc>

This is only and example and not meant to define or dicatate any real way to do it... 

Therefore I believe that adopting XML ParlayX, XML WebServices SOA principles on how
 services are produce and consume can be a benefit... and advantage then trying to create the wheel. The initial services can be Location Based, Context Based, Ambient Based or Payment based. 
Here is another example http://agilepay.ws/ws  of exactly what I am talking about..
Regards,
-E




On 12/21/06, Roy, Radhika R. <RADHIKA.R.ROY at saic.com> wrote: 

	Ed:
	 
	What is the point from standard point of view?
	 
	Do you want LAN services? IP services? QOS services? Security services? Web services? Routing services? Bandwidth services? Voicemail services? Multicast services? Multimedia conferencing services? Email services? Brokering services? Session establishment services? HTML-XML conversion services? .....????? 
	 
	Best regards,
	Radhika

________________________________

	From: p2p-sip-bounces at cs.columbia.edu on behalf of EdPimentl 
	Sent: Thu 12/21/2006 8:06 PM
	To: Philip Matthews
	Cc: P2P-SIP
	Subject: Re: [p2p-sip] Concepts Draft: Open Issue #4: Can clients offerservices?
	
	
	
	These are all excellent discussions on services..
	I like to submit to look and borrow from the guiding principles supporting
	XML WebServices SOBA (Service OrientedBusiness Applications)
	Where these are automagically orchestrated and choreograph and essentially programatically established a business contract using the above dejure standards.. 
	
	Sincerely and Best Regards,
	-Ed
	
	Mail:       edpimentl[at]gmail[dot]com
	VoIP:      edpimentl [Skype | GoogleTalk]
	IM:          edpimentl [ AOL | Jabber | Yahoo | MSN ]
	
	Mobile Content Management 
	http://mobilecentral.ws <http://mobilecentral.ws/> 
	
	Mobile Social Networks(Private Label)
	http://GooGaYa.com <http://googaya.com/> 
	
	Business At the Speed of Thought
	http://biscentral.biz <http://biscentral.biz/> 
	
	Mobile Payment
	http://AgilePAY.ws <http://agilepay.ws/>  




-- 
-Ed

Mail:       edpimentl[at]gmail[dot]com
Direct:     678.522.8511
VoIP:      edpimentl [Skype | GoogleTalk]
IM:          edpimentl [ AOL | Jabber | Yahoo | MSN ] 

Mobile Content Management
http://mobilecentral.ws <http://mobilecentral.ws/> 

Mobile Social Networks(Private Label)
http://GooGaYa.com <http://googaya.com/> 

Mobile Payment
http://AgilePAY.ws <http://agilepay.ws/>  
-------------- next part --------------
An HTML attachment was scrubbed...
URL: https://lists.cs.columbia.edu/pipermail/p2p-sip/attachments/20061223/e0de5e1c/attachment-0001.html

From edpimentl at gmail.com  Sat Dec 23 15:07:36 2006
From: edpimentl at gmail.com (EdPimentl)
Date: Sat, 23 Dec 2006 15:07:36 -0500
Subject: [p2p-sip] Concepts Draft: Open Issue #4: Can clients
	offerservices?
In-Reply-To: <62D92A9A02BCC845B202323D49A48426E1CF74@0591-ITS-EXMP02.us.saic.com>
References: <2043E08B-FCB3-424A-B23C-B06BA0F65532@magma.ca>
	<4589DAC8.2070900@telecomitalia.it>
	<0F1E475D-9B7F-4823-BFD3-177CEC9A0516@magma.ca>
	<458B1862.1070101@telecomitalia.it>
	<BDBB298D-CC86-4262-AD69-979E2A4A306B@magma.ca>
	<9dc4a1670612211706q503f5eact22bc637639699da0@mail.gmail.com>
	<62D92A9A02BCC845B202323D49A48426E1CF70@0591-ITS-EXMP02.us.saic.com>
	<9dc4a1670612220849t39674d6obbd3aad36f9ae63a@mail.gmail.com>
	<62D92A9A02BCC845B202323D49A48426E1CF74@0591-ITS-EXMP02.us.saic.com>
Message-ID: <9dc4a1670612231207s4c3eff3ela9195f82176e49ff@mail.gmail.com>

Hello Roy,

Hope this message finds you well.
To follow Henry's suggestion, let us table discussing P2P client services in
this forum after WG is established and we have contributed open-source code
as the foundation for the WG.

However I would be glad to continue this healthy discussion on a blog or
privately if you wish.... We are actually extremely involved in  the "real"
orchestration and choreographing (as well as Amazon,  Google and
Micro$oft)of SOA WebServices... and if interested talk about some more...
:-)
Sincerely and best regards,
-E
Edpimentl : Skype | Google | MSN | Yahoo
http://edpimentel.com

On 12/23/06, Roy, Radhika R. <RADHIKA.R.ROY at saic.com> wrote:
>
>  Ed:
>
> I appreciate your email.
>
> These kinds of things were discussed before. It is nothing new.
>
> These are NOT standards to use URLs with service connotations. These are
> used defying standards to do some applications here and there having some
> like-minded people.
>
> If billions of people start using with billions and billions of services
> with those services names specifying URLs, nothing will left in this
> universe that will not be termed as "services."
>
> That is why, we go for standardization bringing sanity throughout the
> whole world.
>
> If people feel so strongly for using URLs specifying services, why do they
> need to go via back door "violating" standards?
>
> Rather, they will be welcomed to make their case and create standards like
> this. It is so simple to do so, but there must not be violation of
> standards.
>
> Best regards,
> Radhika
>
> PS: Please analyze the XML script that has been used as an example by you.
>
> 1. The word SMS is a service known as "standard." This SMS has been used
> to qualify the URL. So, it is perfectly alright because it is not the URL
> itself that specifies services.
>
> 2. If you want to build features as shown in you example, it will NOT
> violate URL standard so far I think.
>
> 3. For all other features that you have described in the example, we have
> to examine in terms of the scope of the standards. For sending SMS to Henry
> means invocation of the SMS application service probably using the SMS
> application server. Whether we are allowed to invoke other services
> "directly" in p2p-SIP layer is debatable as a matter of standardization. I
> would say no. For example, in SIP, the invocation of SMS application is not
> a part of the session establishment. A call control SIP proxy might have
> some extra intelligent what to do next if the Henry's phone is busy.
>
> 3a. Here is the most important fine line: To invoke SMS is called another
> service initiation and it is not done in the SIP layer. That is, in service
> creation/execution environment, an intelligent proxy (that has the knowledge
> beyond SIP layer) may be able to invoke SMS application. If so, it is called
> service creation/execution capability of the proxy server which is NOT in
> the SIP layer.
>
> 3b. Now, you have got the whole new services paradigm - Services
> Creation/Execution Environment. In fact, SIP has created a whole application
> services industry where all those application servers are being invoked as a
> side effects of the SIP session establishment: Parlay, CORBA, Open Source,
> OMA, just to mention a few. This is not in the SIP layer.
>
> 4. Conclusion: The example that you have provided it should not be for the
> p2p-SIP layer. It will be a perfect example for p2p-SIP services
> creation/execution environments that reside above the p2p-SIP Layer.
>
> ------------------------------
> *From:* EdPimentl [mailto:edpimentl at gmail.com]
> *Sent:* Fri 12/22/2006 11:49 AM
> *To:* Roy, Radhika R.
> *Cc:* Philip Matthews; P2P-SIP
> *Subject:* Re: [p2p-sip] Concepts Draft: Open Issue #4: Can clients
> offerservices?
>
> Hello Roy,
>
> I read your previous post and agree with many of your comments.
> We respectfully differ on your URL is just a url.... since today a URL
> can be equivalent to
> the hard coded/brittle closely couple CORBA like "SERVICES" offer by the
> unwilling (telcos)partners.
>
> Regarding the standard point is that we can "consume" and "produce"
> Network Services, Applications (Note VoIP is an Application over IP)
> Services, Payment Services...
> In fact the recent enormous investment in Web2.0 and MobileWeb type
> companies are companies
> that "orchestrate, choreograph, compose service which originate in another
> company/enterprise/client service catalog...
> ParlayX is based upon that very same premise... and I am not advocating
> using it in a P2P-SIP but pointing out that P2P-SIP can benefit/borrow the
> guiding principles of such standards.
>
> Consider this... SMS on Busy P2P-SIP Subscriber
> <p2p-sip_svc>
>    <register>
>    <disconnected causeCode="CAUSE_BUSY"
>      destination=tel:5165551212"                      // comment IE
> Henry's number :-)
>     connection="terminating" block="false">
>  <getOriginatingAddress address="orig"/>
> <sendSMS to="sip://henry at adobe.com"
>    <content=" ' call attempt by '+%orig;+'on 5165551212'"/>
> </disconnected>
> </register>
> <p2p-sip_svc>
>
> This is only and example and not meant to define or dicatate any real way
> to do it...
>
> Therefore I believe that adopting XML ParlayX, XML WebServices SOA
> principles on how
>  services are produce and consume can be a benefit... and advantage then
> trying to create the wheel. The initial services can be Location Based,
> Context Based, Ambient Based or Payment based.
> Here is another example http://agilepay.ws/ws  of exactly what I am
> talking about..
> Regards,
> -E
>
>
>
> On 12/21/06, Roy, Radhika R. <RADHIKA.R.ROY at saic.com> wrote:
> >
> >  Ed:
> >
> > What is the point from standard point of view?
> >
> > Do you want LAN services? IP services? QOS services? Security services?
> > Web services? Routing services? Bandwidth services? Voicemail services?
> > Multicast services? Multimedia conferencing services? Email services?
> > Brokering services? Session establishment services? HTML-XML conversion
> > services? .....?????
> >
> > Best regards,
> > Radhika
> >
> > ------------------------------
> > *From:* p2p-sip-bounces at cs.columbia.edu on behalf of EdPimentl
> > *Sent:* Thu 12/21/2006 8:06 PM
> > *To:* Philip Matthews
> > *Cc:* P2P-SIP
> > *Subject:* Re: [p2p-sip] Concepts Draft: Open Issue #4: Can clients
> > offerservices?
> >
> >  These are all excellent discussions on services..
> > I like to submit to look and borrow from the guiding principles
> > supporting
> > XML WebServices SOBA (Service OrientedBusiness Applications)
> > Where these are automagically orchestrated and choreograph and
> > essentially programatically established a business contract using the above
> > dejure standards..
> >
> > Sincerely and Best Regards,
> > -Ed
> >
> > Mail:       edpimentl[at]gmail[dot]com
> > VoIP:      edpimentl [Skype | GoogleTalk]
> > IM:          edpimentl [ AOL | Jabber | Yahoo | MSN ]
> >
> > Mobile Content Management
> > http://mobilecentral.ws
> >
> > Mobile Social Networks(Private Label)
> > http://GooGaYa.com <http://googaya.com/>
> >
> > Business At the Speed of Thought
> > http://biscentral.biz
> >
> > Mobile Payment
> > http://AgilePAY.ws <http://agilepay.ws/>
> >
>
>
>
> --
> -Ed
>
> Mail:       edpimentl[at]gmail[dot]com
> VoIP:      edpimentl [Skype | GoogleTalk]
> IM:          edpimentl [ AOL | Jabber | Yahoo | MSN ]
>
> Mobile Content Management
> http://mobilecentral.ws
>
> Mobile Social Networks(Private Label)
> http://GooGaYa.com <http://googaya.com/>
>

   Business At the Speed of Thought
   http://biscentral.biz

Mobile Payment
> http://AgilePAY.ws <http://agilepay.ws/>
>



-- 
-Ed

Mail:       edpimentl[at]gmail[dot]com
Direct:     678.522.8511
VoIP:      edpimentl [Skype | GoogleTalk]
IM:          edpimentl [ AOL | Jabber | Yahoo | MSN ]

Mobile Content Management
http://mobilecentral.ws

Mobile Social Networks(Private Label)
http://GooGaYa.com

Mobile Payment
http://AgilePAY.ws
-------------- next part --------------
An HTML attachment was scrubbed...
URL: https://lists.cs.columbia.edu/pipermail/p2p-sip/attachments/20061223/f741c088/attachment.html

From sukerry at 126.com  Sat Dec 30 07:50:38 2006
From: sukerry at 126.com (sukerry)
Date: Sat, 30 Dec 2006 20:50:38 +0800
Subject: [p2p-sip] Is there any open project about p2psip
Message-ID: <4596609F.113A0A.02158>

Hello, every one,

Is there any open projects related with p2p sip , I would like to join it,
If you know , please let me know, Thanks in advance!

Thanks
Kerry
-------------- next part --------------
An HTML attachment was scrubbed...
URL: https://lists.cs.columbia.edu/pipermail/p2p-sip/attachments/20061230/5a0e6401/attachment.html

From enrico.marocco at telecomitalia.it  Sat Dec 30 08:06:02 2006
From: enrico.marocco at telecomitalia.it (Enrico Marocco)
Date: Sat, 30 Dec 2006 14:06:02 +0100
Subject: [p2p-sip] Is there any open project about p2psip
In-Reply-To: <4596609F.113A0A.02158@dns2.cselt.it>
References: <4596609F.113A0A.02158@dns2.cselt.it>
Message-ID: <4596643A.5000008@telecomitalia.it>

sukerry wrote:
> Is there any open projects related with p2p sip , I would like to join it,
> If you know , please let me know, Thanks in advance!

Some people in this group are used to playing on the sipdht project
(http://sipdht.sf.net);  give a look and feel free to ask any question
on its mailing list, jabber chatroom or SIP conference (which actually
aren't very crowded, expecially during these holidays).

Please note that it isn't an implementation of the p2psip protocol,
mainly because it doesn't exist yet ;-)

-- 
Ciao,
Enrico

From edpimentl at gmail.com  Sat Dec 30 10:53:21 2006
From: edpimentl at gmail.com (EdPimentl)
Date: Sat, 30 Dec 2006 10:53:21 -0500
Subject: [p2p-sip] Is there any open project about p2psip
In-Reply-To: <4596643A.5000008@telecomitalia.it>
References: <4596609F.113A0A.02158@dns2.cselt.it>
	<4596643A.5000008@telecomitalia.it>
Message-ID: <9dc4a1670612300753i46c95a4fs1e4255b28e4fb593@mail.gmail.com>

Hello Enrico,

We are interested in building a JRuby/Ruby implementation of latest IETF
P2P-SIP Draft...

-E

On 12/30/06, Enrico Marocco <enrico.marocco at telecomitalia.it> wrote:
>
> sukerry wrote:
> > Is there any open projects related with p2p sip , I would like to join
> it,
> > If you know , please let me know, Thanks in advance!
>
> Some people in this group are used to playing on the sipdht project
> (http://sipdht.sf.net);  give a look and feel free to ask any question
> on its mailing list, jabber chatroom or SIP conference (which actually
> aren't very crowded, expecially during these holidays).
>
> Please note that it isn't an implementation of the p2psip protocol,
> mainly because it doesn't exist yet ;-)
>
> --
> Ciao,
> Enrico
> _______________________________________________
> p2p-sip mailing list
> p2p-sip at cs.columbia.edu
> https://lists.cs.columbia.edu/cucslists/listinfo/p2p-sip
>



-- 
-Ed

Mail:       edpimentl[at]gmail[dot]com
VoIP:      edpimentl [Skype | GoogleTalk]
IM:          edpimentl [ AOL | Jabber | Yahoo | MSN ]

Mobile Content Management
http://mobilecentral.ws

Mobile Social Networks(Private Label)
http://GooGaYa.com

Mobile Payment
http://AgilePAY.ws
-------------- next part --------------
An HTML attachment was scrubbed...
URL: https://lists.cs.columbia.edu/pipermail/p2p-sip/attachments/20061230/2e3d5240/attachment-0001.html

From enrico.marocco at telecomitalia.it  Sat Dec 30 11:23:47 2006
From: enrico.marocco at telecomitalia.it (Enrico Marocco)
Date: Sat, 30 Dec 2006 17:23:47 +0100
Subject: [p2p-sip] Is there any open project about p2psip
In-Reply-To: <9dc4a1670612300753i46c95a4fs1e4255b28e4fb593@mail.gmail.com>
References: <4596609F.113A0A.02158@dns2.cselt.it>
	<4596643A.5000008@telecomitalia.it>
	<9dc4a1670612300753i46c95a4fs1e4255b28e4fb593@mail.gmail.com>
Message-ID: <45969293.8010804@telecomitalia.it>

EdPimentl wrote:
> We are interested in building a JRuby/Ruby implementation of latest IETF 
> P2P-SIP Draft...

If your implementation is meant for commercial purposes, I suggest that 
you wait some time.  The group is not likely to come out with a protocol 
for a fairly long time and, actually, there isn't an official working 
group either.

On the other hand, if you want to propose a particular solution, a 
working implementation will help much.

-- 
Ciao,
Enrico

