[Standards] Jingle and RTCP

Robert McQueen robert.mcqueen at collabora.co.uk
Wed Mar 28 14:00:52 UTC 2007


Peter Saint-Andre wrote:
> I don't think so. We have a note in XEP-0166 about RTCP but it needs to
> be defined more fully. IIRC, in our very old discussions about Jingle we
> thought that you would have a separate content type for RTCP vs. RTP,
> but we have not revisited the issue in quite a while. We'll need to look
> at the link that Aki posted and figure out what some of the best
> practices among people who have implemented RTCP more widely than we have.

I wouldn't advocate a seperate stream type. This is the intention of
"components" to a stream, so a RTP stream could also have an RTCP stream
along with it. I think ICE defines that if it's being used for RTP,
component 0 is the RTP stream, and if component 1 is present, it's the
RTCP. ICE optimises its state machine for components (doesn't make
connectity checks for the other ports in the candidates until one has
shown to work, I think). Provided we have a way of doing/expressing
components in the other transports (the UDP port +1 for example), this
approach will work fine for traditional RTP+RTCP streams, and keep us
SIP-wire-compatible.

> Peter

Regards,
Rob



More information about the Standards mailing list