[MMUSIC] Relation between SRCNAME and MSID

[Magnus Westerlund <[e-mail]>]

Hi,

I got to reading the MSID this morning and as it was some questions how
SRCNMANE (draft-westerlund-avtext-rtcp-sdes-srcname-02) relates to MSID
(draft-alvestrand-mmusic-msid-01) I will give my view on this.

So SRCNAME is a hierarchical structure who's root is a media stream
source, i.e. a camera or a microphone. Its primary usage is to bind
SSRCs that are all related to that media source together. A simple
example would be if one produced simulcast, i.e. alternative encodings,
of the same media source V1, you can have a source name that is V1.enc1
for the first encoding, V2.enc2 for the second one, and so on. When you
apply RTP retransmission or FEC you do that to a particular SSRC, which
then is one of the encodings. Or even more complex if an encoding is an
SVC MST stream, then you have several streams in different RTP sessions
for the different layers, and your FEC or retransmissions only applies
to one of these substreams. Thus you can have a SSRC that carries a
SRCNAME of V1.enc1.baselayer and use that label for both the media
stream and on the FEC SSRC to tie them together. So SRCNAME shares a
property with MSID that is that is ties SSRCs both within and across RTP
Sessions together.

So MSID, especially in its usage with WebRTC to create the WebRTC
MediaStream actually have an interesting relation to SRCNAMES. As MSID
are application level groupings of particular mediaStreamTracks, a MSID
can be used to tie in RTP media stream and its related SSRCs (the ones
carrying the FEC or the layers) into a MediaStream object. Lets try an
simple example to illustrate.

A WebRTC end-point has one video camera (V1) and one microphone (A1).
The video is to be provided at two different encodings V1.enc1 and
V1.enc2 in SRCNAME terms. But for a WebRTC application it can actually
select to deal with these two encodings as two different WebRTC
MediaStreams which I will give MSID values Var1 and Var2, where the
variants contain one video encoding and the audio

In SDP using a=ssrc to provide the SRCNAME it would look like this


m=video
a=ssrc: 1111 SRCNAME:V1.enc1
a=ssrc: 1111 msid:Var1 v1
a=ssrc: 2222 SRCNAME:V1.enc2
a=ssrc: 2222 msid:Var2 v1

a=audio
a=ssrc: 3333 SRCNAME:A1
a=ssrc: 3333 msid:Var1 a1
a=ssrc: 3333 msid:Var2 a1

In other words a given SSRC might actually be a MediaStreamTrack in
multiple MediaStreams and thus have multiple MSIDs. While a SSRC will
only have one SRCNAME, but that one may specify to a higher detail.

If we would have RTP retransmission in the above, a question around MSID
arises. Should all SSRCs that are part of representing be given a MSID
or would it be simpler to just label the media entry point to the
hierarchical relations SRCNAME is capable of expressing? And I need to
point out that MSID can't replace the SRCNAME as currently defined as it
doesn't make clear the hierarchical relation of the SSRCs.

If you are interested in discussion SRCNAME please show up in AVTEXT on
Wednesday at 1300-1430.

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]
----------------------------------------------------------------------