Re: [rtcweb] Plan A, respun

[Christer Holmberg <[e-mail]>]

Hi,

>>>> Let's be very, very precise about this.
>>>>
>>>> Demultiplexing is *always* performed based on SSRC.  It's just that 
>>>> correlation with signaling components (the m-line in this case) 
>>>> might use of PT in the absence of prior knowledge of SSRC.
>>>>
>>>> We didn't get this quite right in Plan A.
>>> Yeah, that's a valid point. It's not really being proposed as a demux point as much as a correlation token. Thanks for clearing that up -- we should definitely clear that up in the next version of the document.
>>
>> Rather than try to overload the PT, I would suggest we define an explicit correlation token that can be sent. Put something in RTCP SDES 
>> and/or RTP header extensions, and in the SDP, to signal what RTP flow corresponds to what m-line.
>
> That kind of defeats some of the purpose here, which includes working with existing deployed implementations.

Existing deployments will use different port numbers, so you would be able to separate media belonging to different m- lines (as different 5-tuples will be used), so you would not need to use the extension.

Regards,

Christer