Re: [MMUSIC] H.264 SVC and BUNDLE

[Christer Holmberg <[e-mail]>]

Hi,

One question is of course whether we are going to move ahead with the m=bundle proposal. In that case, I guess the m=video lines would look like they do today, and the question is how we describe the multiplexed SVC within the m=bundle line.

Regards,

Christer




From: [e-mail] [mailto:[e-mail]] On Behalf Of Harald Alvestrand
Sent: 31. elokuuta 2012 9:13
To: Bernard Aboba
Cc: [e-mail]
Subject: Re: [MMUSIC] H.264 SVC and BUNDLE

On 08/31/2012 03:30 AM, Bernard Aboba wrote:
Harald said:

I lack the background to think about this... in existing equipment, if you want to do SST, but would be willing to do MST if the other end did not support SST, how would your offer look (without BUNDLE)?

My head hurts a little when trying to guess, but from your question it sounds like you've done it, so it's better if you just paste an example.

(BTW, this is really an MMUSIC discussion)

[BA] RFC 6190 gives several examples of SST and MST offers:
7.3.1<http://tools.ietf.org/html/rfc6190#section-7.3.1>.  Example for Offering a Single SVC Session
7.3.2<http://tools.ietf.org/html/rfc6190#section-7.3.2>.  Example for Offering a Single SVC Session Using scalable-layer-Id
7.3.3<http://tools.ietf.org/html/rfc6190#section-7.3.3>.  Example for Offering Multiple Sessions in MST
7.3.4<http://tools.ietf.org/html/rfc6190#section-7.3.4>.  Example for Offering Multiple Sessions in MST Including Operation with Answerer Using scalable-laye
r-id



7.3.5<http://tools.ietf.org/html/rfc6190#section-7.3.5>.  Example for Negotiating an SVC Stream with a Constrained Base Layer in SST
I believe that the answer to your question is closest to example 7.3.5 (SST) with a bit of 7.3.4 (MST) thrown in. Notice that Payload types 97 and 99 are similar but not the same (97 is SST, 99 is MST):
Aha - so you'd send 3 video M-lines, an SST-only application would zero out the second and third, while an MST-only application would zero out the first, and one capable of doing both would accept all 3? and one could look at the payload type (96/97 vs 98 or 99) to decide which one a particular video stream belonged to?

It makes no sense (to my mind at least) to multiplex L1 and L2 together, but one could multiplex SST with L1 and save one port in the "both are supported" case:

a=group:BUNDLE SST L1

I think this would create sessions that would not create an unsolvable problem; if all were accepted, the decoder would have to look at the payload type of the first packet on an SSRC to figure out whether a particular SSRC was using SST or MST for that particular payload.


Offer:

      a=group:DDP L1 L2

      m=video 20000 RTP/AVP 97 96

      a=rtpmap:96 H264-SVC/90000

      a=fmtp:96 profile-level-id=53001e; packetization-mode=0;

      a=rtpmap:97 H264-SVC/90000

      a=fmtp:97 profile-level-id=53001f; packetization-mode=1;

      a=mid:SST

      m=video 20002 RTP/AVP 98

      a=rtpmap:98 H264/90000

      a=fmtp:98 profile-level-id=4de00a; packetization-mode=0;

       mst-mode=NI-T;

      a=mid:L1

      m=video 20004 RTP/AVP 99

      a=rtpmap:99 H264-SVC/90000

      a=fmtp:99 profile-level-id=53001F; packetization-mode=1;

       mst-mode=NI-TC; sprop-operation-point-info=<2,0,1,0,53000c,

      3200,352,288,384,512>,<3,1,2,0,53001F,6400,704,576,768,1024>;

      a=mid:L2

      a=depend:99 lay L1:98

On 08/30/2012 11:19 PM, Bernard Aboba wrote:
Harald said:
> My assumption has always been that if you were operating on a network
> that could only apply diffserv to separate 5-tuples, you would allocate
> the SSRCs comprising a multilayer encoding to different 5-tuples.

[BA] RFC 6190 Section 1 says:

   This memo defines two basic modes for transmission of SVC data,

   single-session transmission (SST) and multi-session transmission

   (MST).  In SST, a single RTP session is used for the transmission of

   all scalability layers comprising an SVC bitstream; in MST, the

   scalability layers are transported on different RTP sessions.

[BA] So it is possible either to use multiple 5-tuples (MST) or a single 5-tuple (SST).  However, my understanding
is that the SST approach is more popular than MST.

> This choice seems to be a choice that the application writer should take
> before deciding whether or not to use BUNDLE, so it's a little hard to
> see what the relationship to BUNDLE is.

[BA] The open question has been how to indicate the desire for both layered coding and BUNDLE in SDP,
given the backward compatibility issues we have talked about.  For example, if it is desired to do SST and
BUNDLE, do you signal MST and BUNDLE on different ports in the offer and then if SST and BUNDLE are
supported by the answer, have the answer respond with BUNDLE as well as dependency groups using
the same port?  That would seem to imply an assumption that every MST implementation can also do SST.
Or do do you signal SST and BUNDLE on the same port, and then if the response indicates that the SDP
was rejected, formulate an alternative offer (MST with BUNDLE?  H.264/AVC with BUNDLE?
SST without BUNDLE??)