Re: [rtcweb] path forward on bundle

[Christer Holmberg <[e-mail]>]

Hi,

>I'm trying to figure out how to move forward on bundle stuff. Let me start by summarizing what I think the three proposals are then we can try and sort out pros/cons of each. I think we agree that the goals here is to have something that gets the SDP so 
>that we can negotiate using less ports. 
>
>I see it as we have three proposed solutions that I will call "a=bundle same port", "a=bundle different port", and  "m=bundle". I'll try and describe these below to make sure we are on same page of what the three are then we can try and figurer out 
>pros/cons of each and what to do. I'll give examples of the SDP offer for each but I tried and make everything simple, I'm going to ignore the RTCP mux and such but I think you can see how that would get added to all the examples. 
>
>
>First lets set the baseline of an offer that does not want to multiplex the audio and video on same port 
 >        v=0
 >        o=alice 2890844526 2890844526 IN IP4 host.atlanta.example.com
 >        s=
 >        c=IN IP4 host.atlanta.example.com
 >        t=0 0
 >        m=audio 49170 RTP/AVP 0 
 >        a=mid:foo
 >        a=rtpmap:0 PCMU/8000
 >        m=video 49172 RTP/AVP 31 
 >        a=mid:bar
 >       a=rtpmap:31 H261/90000

Correct.


>Proposal "a=bundle same port"
>        v=0
>         o=alice 2890844526 2890844526 IN IP4 host.atlanta.example.com
>         s=
>         c=IN IP4 host.atlanta.example.com
>         t=0 0
>         a=group:BUNDLE foo bar
>         m=audio 49170 RTP/AVP 0 
>         a=mid:foo
>         a=rtpmap:0 PCMU/8000
>         m=video 49170 RTP/AVP 31 
>        a=mid:bar
>         a=rtpmap:31 H261/90000
>
> In the above example, note that all the m lines for a mid identified in the group:BUNLDE have the same port (49170)

Correct. This is what is currently described in draft-bundle.



>Proposal "a=bundle different port"
>         v=0
>         o=alice 2890844526 2890844526 IN IP4 host.atlanta.example.com
>         s=
>         c=IN IP4 host.atlanta.example.com
>         t=0 0
>         a=group:BUNDLE foo bar
>         m=audio 49170 RTP/AVP 0 
>         a=mid:foo
>         a=rtpmap:0 PCMU/8000
>         m=video 49172 RTP/AVP 31 
>         a=mid:bar
>         a=rtpmap:31 H261/90000

> In the above example, note that the port numbers are the same as baseline - if you device receiving this offer does not support group:BUNLDE, then this is identical to the baseline. So one m line has a port of 49170 and the other has a different port. 
> Other than that, it is the same as the "a=bundle same port". The semantics of group:BUNLDE would be defined to be that if the SDP receiver supports BUNDLE and wants to use it, then it places a group:BUNDLE line in the answer and it send all the media 
> to the mids in the in BUNDLE group to the port number identified for the m-line corresponding to the first mid in the BUNDLE group. In this example, the first mid on the a=group:BUNDLE lines is foo, which has a m line with a port of 49170, so both the 
> H261 and PCMU (mids foo and bar) would go to port 49170 if the device creating the answer wanted to use BUNDLE. We don't have a draft yet that says exactly this thought it is very close to the one-rtp draft ( and for that matter  close to bundle draft )

This is more or less Harald's original proposal.



>Proposal "m=bundle"
>         v=0
>         o=alice 2890844526 2890844526 IN IP4 host.atlanta.example.com
>         s=
>         c=IN IP4 host.atlanta.example.com
>         t=0 0
>         a=group:BUNDLE foo bar baz 
>         m=audio 49170 RTP/AVP 0 
>         a=mid:foo
>         a=rtpmap:0 PCMU/8000
>         m=video 49172 RTP/AVP 31 
>         a=mid:bar
>         a=rtpmap:31 H261/90000
>         m=bundle 10000 RTP/AVP 0 8 97 31 32
>         a=mid:baz
>         a=full-rtpmap:0 audio/PCMU/8000
>         a=full-rtpmap:31 video/H261/90000
>
>
> Note in the above example the SDP above the m=bundle line is pretty much the same as the baseline + the group:BUNDLE line. The m=bundle is new media type that indicates many things are bundled together on port 10000. A device that understood 
> bundle would know to ignore the m lines corresponding to foo and bar mids and would just use the stuff from the bas mid. 

Correct. Later today, we are planning to submit an individual draft (ie *not* a new version of draft-bundle) on this alternative, so that people have some text to look at.

Regards,

Christer