[p2p-hackers] Is P2P SIP Poised to Out-Hype Skype? (Voxilla)

[David Barrett]

I wrote some comments on Eric's blog (sipthat.com) detailing why I disagree:

	http://sipthat.com/archives/000351.html

Aswath has another relevant article:

	http://www.mocaedu.com/mt/archives/000168.html

Basically, I think P2P is useful to reduce legal vulnerability, obtain 
high security, minimize bottlenecks, and distribute computation.  None 
of these applies to SIP proxies, which are essentially directory-based 
"rendezvous" services.

That said, all of these certainly apply to the actual media components 
(audio, video, etc).  But these are generally RTP.  The SIP part is very 
low cost, low CPU, low bandwidth, and benefits from centralization.

My original comments to Eric's article follow:

-----Original Message-----
From: David Barrett [mailto:dbarrett at quinthar.com]
Sent: Tuesday, August 02, 2005 3:11 AM
To: erik at sipthat.com
Subject: Re: P2P VoIP using SIP and Open Standards

Hi, sorry for the late reply to your July 25th posting about P2P SIP, but
you posed the question:


>> Everyone is looking for a new solution for P2P VoIP, I think P2P
>> SIP is the answer, do you?


My answer is no, I don't believe P2P SIP is the key to P2P VoIP. More
specifically, I don't think an IETF standard for decentralized SIP proxies
will gain traction because it solves problems that nobody has, while
complicating the problems that we do.

Now, I recognize VoIP will grow, and the actual media components will
migrate to P2P. But I do not believe the SIP proxy component of the global
architecture gains anything by being decentralized in this way.

My reasoning is based on the observation that P2P is primarily useful in two
situations:

a) When a centralized solution is too costly
b) When a centralized solution is legally vulnerable

Consider the case of Napster. A kid in a college dorm room was able to host
a global rendezvous service (akin to a SIP proxy) for millions of
simultaneous users, for free. Obviously, he didn't suffer from (a).
But (b) was what took him down, and thus gave rise to Gnutella (et al).

But had (b) not occurred (ie, had the courts ruled in favor of Napster), I
offer that Gnutella simply would not exist because it offers precisely zero
incremental value. It is slower, less reliable, and less comprehensive than
Napster could have been. Had Napster been allowed to grow, it could have
offered better services, evolved faster, and been superior in every
measurable way.

SIP proxies are today like Napster was then. But VoIP isn't illegal, and
thus a SIP proxy is not a legal vulnerability. Thus I see no reason to
believe that a decentralized alternative to a SIP proxy would warrant the
resulting complexity cost inherent in any "pure"-P2P solution.

All that said, I do believe that P2P is the future of VoIP, and near-free
services will be the business model. I merely think that the SIP proxy part
of the equation is best left centralized.

-david
Posted by: David Barrett at August 6, 2005 01:01 AM

So again, I'm still not sold on the benefits of P2P SIP. I think it's
cool, don't get me wrong. And the idea of setting up an instant P2P
network in the middle of a desert or in a disaster area is appealing,
certainly. But you have to admit those are rather extreme circumstances.
For the remaining 99.999% of actual real world users, this is not the case.

Furthermore, I agree that the servers at Free World Dialup and other
free SIP services cost money. But not much, else they wouldn't be free.
For under $100/mo you can get a dedicated server right on the backbone.
The cost of deploying and maintaining hardware is virtually negligible
these days.

The real costs come in managing the network, and this is where
centralized services come in handy.

So P2P SIP is cool, certainly. And it might be handy the extreme
situations you offer. But by the time the intrinsic decentralized
problems have been hammered out, a centralized alternative will already
dominate the landscape. And when it comes to competition, the
decentralized solution offers no significant real world advantage (cost,
reliability, performance) to anyone (user, administrator), while
offering significant drawbacks in terms of complexity and the overall
uncontrolled nature of P2P.

Would you disagree? I mean, except for disaster areas and deserts, and
except for saving $100/mo for a million users, what's the benefit?

Furthermore, do you acknowledge the disadvantages in terms of
reliability and performance (ie, global distributed search versus
database lookup) of a P2P solution? How would you argue the benefits
outweigh the detriments?

-david