[xiph-rtp] Ident field and configuration - stream mapping

[Luca Barbato]

Ralph Giles wrote:

> 
> I'd decided that 16 bits was good enough, and just left the other 8 
> reserved as a hedge against needs. We could use them to define 
> a standard set of 'profile' codebooks, for example. What do you think
> about that?

I put 24bit since I couldn't find any other use for it and that "just" 
227days of music could be not enough in other worst cases.

Not sure about profile codebooks (as modeled as mpeg4 profiles) since 
one of the strengths of vorbis is the tunability.

> 
> Yes, Aaron suggested a very similar url contruction scheme to allow
> new sets of headers to be posted and retrieved during live encoding.
> I'm ok including this as a MAY in one of the out-of-band transfer
> documents.

Currently I didn't state that since is quite a special case (and could 
reside in the companion rfcs)

> 
> Speaking of, if you want to document retrieval over http separately,
> could you write that up? I'd like to describe both it and the separate
> multicast channel techniques asap because they're important for useful
> implementation. We could also mention them in passing in the payload 
> draft, which would give people the idea of what to do.

For the out of band protocol I'm more concerned about rtsp since I know 
an implementation that will require it (fenice-ng). I'm not sure who 
will use the http method since it could have huge latency issues.

I'd focus on syncing the theora-rtp with vorbis-rtp if the current draft 
is ok for everybody and keep extending the reference implementation to 
have real life tests.

lu

-- 

Luca Barbato

Gentoo/linux Developer		Gentoo/PPC Operational Leader
http://dev.gentoo.org/~lu_zero