[p2p-sip] Revised Draft Charter for P2PSIP
Roy, Radhika R.
RADHIKA.R.ROY at saic.com
Fri Sep 22 10:38:46 EDT 2006
Let us see what comes out of this. I am stamping a time on it: September 22,
2006 - 10:38 AM EDT USA.
I have not seen even a single draft on NAT crossing in p2p-SIP other than
David gave some head start which is far away in solving the NAT crossing.
We will wait for the next 2 years to see the p2p-SIP NAT crossing solution.
ICE is still there even after 5 years.
BTW: NATs created their own problems for using them beyond what they have
been standardized by the IETF. Since NATs crossed their IETF standardized
boundary, they are also being solved by non-standard ways. We are getting
VoIP and other services through crossing of NATs all the time in
non-standardized ways. No one including Skype is coming for standards of NAT
crossing. Someday NATs will also have their natural death as IETF hoped for.
SIP standards have hinted even indirectly NAT-aware SIP proxy that can
modify SDP going beyond all standardized norms. Even it did not work.
Bottom line: NATs are hopeless in proving services like security and
configuration hiding. To get more address spaces, no one needs NATs anymore.
What is left for usage of NATs? So, the idea is to just cross it by
applications like SIP and p2p-SIP. To just cross the NATs and not to bother
about the security and other aspects, just keep them open as you want. If
this is the case, it is not a big deal as so many proprietary solutions are
used today. Is it is not good enough?
ICE: It wants a standardized way by IETF. So, it is concerned about
security, compatibility, interoperability, and billions of other features.
Here we go again: NATs can be crossed using some standard mechanisms someday
- may be 5 years from now. How about QOS aspects of NATs because they are
the bottleneck as they will be carrying out audio and video real-time media?
NSIS is working on it. It has no clue what kind of signaling messages will
be good to provide the indication that NATs will maintain QOS. Otherwise,
the whole network is OK to provide QOS, but NATs will not.
Conclusion: By the time we solve NAT problems, quantum computers will be a
reality, and IETF will have nothing, but NAT-standardization-supper power
From: Michael Slavitch [mailto:slavitch at gmail.com]
Sent: Fri 9/22/2006 9:45 AM
To: Roy, Radhika R.
Cc: David A. Bryan; p2p-sip at cs.columbia.edu
Subject: Re: [p2p-sip] Revised Draft Charter for P2PSIP
It is certainly not a nonsense problem outside of some rare circles as
nearly all consumer and business networks fall behind NATs. Without NAT
crossing P2PSIP is useless. I agree that in a deterministic system a perfect
solution is required.
That is why deterministic systems are always so brittle. The world is a
messy fractal place.
By definition P2PSIP will not be deterministic so a perfect solution is not
only unnecessary, it is not desirable if the cost of complexity exceeds the
The solutions that exist are statistical and fractal, not perfect and
deterministic. Skype is living proof of a broadly used implementation that
shows that the problem is solved in a matter that meets the requirements for
P2P systems. Since all P2P systems share a common fractal nature a fractal
solution is perfectly suited to the domain.
For all the criticism of ICE, and I've done my fair share, it is getting
simple enough to implement from the draft to be expressed in BNF notation.
It is now fairly easy to build ICE into a finite state machine, as ICE now
describes a completion mechanism where the protocol is complete. That
encourages interoperability from the start.
Any inadequacies that remain result from the fact that client-server SIP is
the only official user of the protocol. These are minor issues that can
only be addressed by MMUSIC once we have a P2PSIP WG that can request
changes. That is a diplomatic problem, not an engineering problem, if the
underlying protocol in ICE can meet base P2PSIP requirements. That is why
WG chairs are chosen carefully.
Regarding unsolved problems: If the underlying system is fractal then a
statistical solution with a high probability of success meets the
requirements for the system. This is true for all fractal systems. The
only issue is determining the probability of success of different solutions
and either choosing one that meets the circumstances at hand or modifying
another to make it an appropriate choice.
I further argue that the fractal nature of ICE is far more suited to the
fractal P2P world than it is to the assumed-deterministic IMS/CS model. It
may be not ever be solved perfectly enough for IMS but for self-organizing
P2PSIP it won't have to be perfect as the nature of a self-organizing
systems is to work around their own shortcomings. P2PSIP will be rugged
enough such that perfect solutions are not needed. If the other working
groups are failing it may be because the systems they are building towards
may be broken and brittle and not rugged enough to tolerate statistical
solutions. Given the talent at hand that will certainly not be the case with
More information about the P2p-sip