Re: [AVTCORE] SSRC multiplexing of RTP

[Magnus Westerlund <[e-mail]>]

Harald Alvestrand skrev 2011-03-15 07:24:
> On 03/14/11 18:14, [e-mail] wrote:
>> On Monday, March 14 2011, "Magnus Westerlund" wrote to "Harald Alvestrand, [e-mail]" saying:
>>
>>>> - if it is a new stream, how to tell that it's a new stream
>>>> - if it is an old stream, renumbered, how to tell what the old stream was
>>> I don't think the specs are that unclear on what is what. You do need
>>> the CNAME to determine which case it is. Sure it can be clarified. I
>>> mostly think the issue is with the state of the implementations. They
>>> haven't considered the implications and need for multiple SSRC support.
>> There are valid cases where multiple independent streams will have the same
>> CNAME -- if they're multiple sources from the same participant.  Multiple
>> video streams in the Telepresence/CLUE environment are probably the current
>> most prominent example of this, but generally, this will be the case if you
>> have multiple streams that all need to be synchronized together.
> Hm. CNAME seems like it's not a stream identifier.
> 
> RFC 3550 section 6.5.1 contains lots of language saying that it is the 
> user@host, which would mean that if I send two video streams and two 
> audio streams, I have four streams with the same CNAME.

Well, generating good quality cnames are one question and that is solved by:
https://datatracker.ietf.org/doc/draft-ietf-avt-rtp-cnames/

The other question is how you use the CNAME to bind together streams. I
do agree that we do have cases today where you can't both get synch and
SSRC multiplexing of certain functionalities. It is definitely something
that should be included in the
https://datatracker.ietf.org/doc/draft-perkins-rtcweb-rtp-usage/

> 
> No matter whether these are in one RTP session or multiple RTP sessions, 
> this seems problematic - in the case of layered encodings using multiple 
> RTP sessions, it seems to me that if I send two video streams over four 
> sessions, I will have a hard time matching each base layer with its 
> corresponding higher layer if CNAME is my only correlator.

Well, you can use the same SSRC over multiple RTP streams to achieve
quicker and clearer synchronization. I still need to read SVC payload
format, but I know that Session multiplexing of RTP retransmissions (RFC
4588) does use that method for correlation.

I am not certain that we haven't forgotten any case, but when
introducing certain type of RTP functionalities you also at the same
time restrict the possibilities in how these mechanisms can be used.

I think AVT groups should work with CLUE and RTCWEB here to address the
issues found due to the use cases these WGs have.

Cheers

Magnus Westerlund

----------------------------------------------------------------------
Multimedia Technologies, Ericsson Research EAB/TVM
----------------------------------------------------------------------
Ericsson AB                | Phone  [phone]
Färögatan 6                | Mobile [phone]
SE-164 80 Stockholm, Sweden| mailto: [e-mail]
----------------------------------------------------------------------