Re: [MMUSIC] Proposal for what bundle should say about demux

[Paul Kyzivat <[e-mail]>]

On 5/23/13 4:57 PM, Martin Thomson wrote:
> On 23 May 2013 11:39, Paul Kyzivat <[e-mail]> wrote:
>> Generally, once one packet has been bound to a processing pipeline then the
>> SSRC from that can be remembered, and used to shortcut the mapping for
>> future packets that have the same SSRC.
>
> This is actually the key.  The pipeline should be attached to the
> SSRC.  How that attachment is initially formed is what I think that
> Cullen is talking about.
>

Well, I said *generally*, not *always*.

For instance, we have a clue case where we can have two clue switched 
captures, one representing the video for loudest speaker, and one for 
the 2nd loudest speaker. A single SSRC can move from one of those 
captures to the other. Its the capture (identified by an rtp header 
extension) that selects the processing pipeline.

Or, you can view it as layered. Where the ssrc maps to an early stage of 
the pipeline (selected by ssrc) that does the codec processing, and then 
a different stage of the pipeline (selected by capture id from the 
header extension) that switches the decoded output to a particular 
rendering device.

	Thanks,
	Paul