> -----Original Message-----
> From: Philip Matthews [mailto:[EMAIL PROTECTED]
> Subject: Re: [Sip] RE: [BEHAVE] Re: ICE deployment data before LC for RFC
> 
> David Barrett's algorithm, if I understand it correctly, would work in
> this situation. (As far as I can tell, David's is a simplified version
> of ICE).

That's correct.  The major differences (so far as I know) between what I'm
doing and what ICE does include:

- Peer's needn't exchange SDP before attempting connection establishment
(rather, the rendezvous server just tells both sides the others' LAN/NAT
addresses and they try to connect immediately). 

- Paths are evaluated in parallel rather than sequentially (ie, I attempt to
contact you direct via the LAN and indirectly via your NAT at the same time,
and then pick the faster of the two)

- There's no support for TURN servers as my application doesn't require them
(though it would be extremely straightforward to add).

- Clients attempt NAT classification as a *purely advisory* function to
support debugging and feed into certain non-critical pairwise optimizations.
(Said another way, even if the classification is totally wrong, it'll still
work.)

- Clients use adaptive keepalives to a STUN server to keep their NAT/FW
"hole" open.  By "adaptive" I mean I use probing to determine highest safe
keepalive frequency of the NAT/FW.  (Incidentally, there's no requirement
that any two clients use the same STUN server, and indeed I intend to move
to enable a "distributed STUN service" using the peers themselves.)

Basically, I do the minimum necessary for my requirements, which I respect
are much easier than SIP's.  However, know that despite my system being much
easier, it has been *extremely difficult and time consuming* to get working,
despite dozens of iterations drawing upon a network of hundreds of thousands
of machines on which to test.  Thus I would suggest that if my solution took
as long as it did, and went through as many variations as it did, the
SIP/ICE implementers should be very wary of hardening anything too early
before it hits the real world.  Because the real world is really, really
strange and unpredictable, I can guarantee you that.

Phillip's asked me to put together a bit more precise overview of the system
I use and I'll try to get to that soon.

-david




_______________________________________________
Sip mailing list  https://www1.ietf.org/mailman/listinfo/sip
This list is for NEW development of the core SIP Protocol
Use [EMAIL PROTECTED] for questions on current sip
Use [EMAIL PROTECTED] for new developments on the application of sip

Reply via email to