[xiph-rtp] Re: [Speex-dev] new RTP development list

[Michael Smith]

On Thursday 21 October 2004 04:51, Aaron Colwell wrote:
> On Wed, Oct 20, 2004 at 08:30:03PM +0200, Tor-Einar Jarnbjo wrote:
> > Onsdag, 20 oktober 2004, skrev Aaron Colwell <acolwell at real.com>:
> > >Why are chained streams not going to be supported? I looked at the wiki
> > >pages and I don't agree with the arguments for dropping chaining
> >
> > support.
> >
> > Well, one major obstacle is that common control protocols for RTP
> > sessions (RTSP and SDP) are not directly supporting all the chaining
> > mechanisms. E.g., you cannot simply change the sample rate or number
> > of channels within a session as these are specified in the session's
> > SDP descriptor.
>
> Why not just pick a master sample rate with sufficient resolution to
> acommodate all the chains. If you still want to maintain sample accurate
> timestamps then you can put a field in the payload that contains the part
> that got rounded off when converting to the master sample rate. This allows
> the RTP timestamp sample rate to stay constant while the actual sample rate
> of the stream contents can vary. This mechanism can also apply to Theora
> streams where the frame rate changes.

Agreed. This seems to be the way the IETF is leaning at the moment (though 
it's largely undecided)
See, for example (this conveniently dropped into my inbox today, I haven't yet 
read all of it...):

http://www.ietf.org/internet-drafts/draft-ietf-avt-variable-rate-audio-00.txt

That draft is for audio, but as you say, this could also be useful for Theora; 
perhaps we should consider extending this draft for video.

Mike