[AVT] RE: [Ietf-behave] RTP in draft-ietf-behave-nat-00

> If a SIP/SDP application receives instructions to send RTP to 
> an odd address (that is, port) when the SDP does not include 
> an a=rtcp attribute (RFC 3605) to specify the RTCP port 
> separately, then indeed, RFC 3550 does say that the RTP 
> should be sent to the even port that is one lower.  But, I 
> claim this SIP/SDP application is broken if it was expecting 
> the application to send RTP on the odd port and RTCP on the 
> next higher even port (or maybe the next lower even port, or 
> even the odd port two higher).  Do you disagree?

No, I do not disagree. In fact, I completely agree with you. 
The simple odd/even rule that we are adding does NOT address the port
contiguity rule (and we've added wording on that that you haven't 
seen yet for the next release).

This is one of the many reason why the next thing we do after
the NAT Behavior document (for UDP Unicast, then Multicast and
TCP) is the "Application" document which will explain all this.

I think it will be very important to have a review of
the upcoming Application document from AVT (at least for those
aspects related to RTP/RTCP. We will definitively seek your feedback.

Again, really what we were looking at addressing is the "substract
one to the odd port" statement (which one could read as applying
even if both the RTP and RTCP ports are seperately negotiated).
We are just ensuring that a NAT will not change the parity.