Re: [MMUSIC] How does mmt grouping work with multiple streams of the same type?

[Harald Alvestrand <[e-mail]>]

On 10/11/2012 03:12 PM, Stach, Thomas wrote:
> Christer, Harald, Jonathan,
>
> I have the impression that the draft-holmberg-mmusic-sdp-mmt-negotiation is based on Jonathan's proposal
> to have an explicit desrciption for the bundled streams as proposed in
> http://www.ietf.org/mail-archive/web/mmusic/current/msg09454.html
>
> Among other points the following is stated:
> * The m=bundle is a complete description of an entirely separate m-line; if it's used, the other m-lines in the group are rejected and the information they carry is ignored.
> I think this a desirable property
>
> Now, consider a case were you have 4 streams, i.e. 4 m-lines (e.g. 2 audio and 2 video).
> How would that be mapped to the m=anymedia line?
> How does mmt grouping work with multiple streams of the same type?
>
>  From the currently proposed syntax for the m=anymedia line I cannot deduce that multiple audio or video streams are offered.
> I would have to look into the "legacy" m-lines to recognize this.
> In case that e.g. the video stream use different codecs, I could also learn that only from the legacy m-lines.

In the case where it's desirable to carry multiple video streams in one 
RTP session, and have different descriptions in SDP for the different 
video streams, there has to be some signalling that tells the recipient 
which description applies to which video stream.

The video streams will have different SSRCSs. Therefore, the only 
reasonable answer in the MMT grouping case seems to be SSRC-associated 
signalling.

In the other variants, it's actually unclear to me how the video streams 
described by the different m= lines would be detected by the recipients; 
I'd been assuming SSRC-associated signalling for other reasons, so only 
when this discussion resurfaced, I started thinking about this again, 
and discovered that I didn't have an answer.

That is, in the description

a=group:bundle foo bar baz
m=video 4270
a=mid:foo
a=property-x:1
m=video
a=mid:bar
a=property-x:2
m=video
a=mid:baz
a=property-x:3

one would expect a single RTP stream to be set up; when the recipient 
gets 3 video streams with SSRCS 234, 235 and 236, how does he know which 
stream should have which value for "property-x"?
One can solve it with SSRC-associated signalling in this case too:

a=ssrc:234 property-x:1

but then you're back to doing SSRC-associated signalling, and if you're 
already doing that, we're not very different from the MMT description 
anyway.

That's how I think of it at the moment, anyway - YMMV.