Re: [AVT] Final edits to RTP spec and A/V profile

[Stephen Casner <casner@acm.org>]

On Thu, 16 Jan 2003, John Lazzaro wrote:

> Sfront currently ships with multi-unicast support, and the
> implementation is a hybrid of both approaches. A sender does
> sent the same RTP packet to each receiver, and in this way
> everything but the lowest layers of the stack believes a
> true multicast session is underway. Sfront needs to do this,
> because it lets a sender create and maintain one recovery
> journal system for all receivers, a big efficiency improvement
> over N recovery journal systems.
>
> However, a receiver doesn't send RTCP RR packets describing
> a sent stream to all other participants, only to the original
> sender. This is for efficiency ...

This is really not a multi-party RTP session then, because each pair
of participants sees only their two SSRC identifiers in packets
exchanged between them.  It is an "arranged coincidence" that a given
participant uses the same SSRC for all of its point-to-point unicast
RTP sessions.

How do you avoid collisions in the choice of SSRC identifiers?  Is the
choice of SSRC coordinated at the time of session establishment?  Or
does a new participant listen for packets from all of the other
participants before choosing its own SSRC?  (This can be tricky if
multiple participants join at once.)

                                                        -- Steve



_______________________________________________
Audio/Video Transport Working Group
avt@ietf.org
https://www1.ietf.org/mailman/listinfo/avt