Re: [rtcweb] New Version Notification for draft-uberti-rtcweb-plan-00.txt

[Martin Thomson <[e-mail]>]

On 12 May 2013 22:22, Justin Uberti <[e-mail]> wrote:
> On Sat, May 11, 2013 at 5:20 AM, Stefan Håkansson LK
> <[e-mail]> wrote:
>>
>> As I read draft-uberti-rtcweb-plan-00, no API changes are needed or
>> proposed, and the call flows could stay the same (but as I recollect the
>> discussion at the Boston interim they should be changed for other
>> reasons).
>>
> The sole API change that is proposed is the new .content property, to allow
> a MST to have its a=content property set, and as such be placed on its own
> m= line. The rest works using the existing APIs; the necessary a=remote-ssrc
> lines can be generated (to enable or disable a stream) based on whether a
> received MST is hooked up to an output or not.

[...]
On 11 May 2013 07:14, Eric Rescorla <[e-mail]> wrote:
> I believe that both Plan A and Plan B require at least one change to
> the API: the ability to indicate that a stream is important and thus
> requires backward-compatible treatment. In Plan A this means
> not marking it bundle-only. In Plan B it means putting it on its own
> m-line.

The .content property in Plan B goes some way to addressing this need,
but it does nothing to provide control over which of the tracks are
selected from the set that are on the same m-line in the
backwards-compatibility mode.

Neither option makes addressing simulcast streams easy for this case.