[Nice] Gracefull fallback, renegotiations?

[Rémi Denis-Courmont]

On Monday 07 July 2008 18:52:00 ext mikhail.zabaluev at nokia.com, you wrote:
> >I'm thinking of caracterising it in terms of remote af/protocol/ip/port
> >(which should uniquely identify it?)

If we were to follow the current ICE-TCP, I suspect not, but it is dubious 
whether the document will remain as it is (if it will remain at all).

> Will TURN figure there somehow? It's kind of scary if you mix
> TURN-encapsulated packets with others in the same socket.

TURN packets are exchanged solely with the TURN relay, so we _can_ demultiplex 
based on the remote port/IP. I would agree that using a separate socket is 
cleaner, but I don't know if it is OK to depend on SO_REUSEADDR.

Note that TURN superseeds STUN - in principle, you would do binding discovery 
with the TURN server, and therefore, you MUST use the same local UDP port to 
do TURN than the rest. Of course, this is only an optimization, so we have 
the option to not do this.

-- 
Rémi Denis-Courmont
Maemo Software, Nokia Devices R&D