Re: Minimizing the protocol (Re: Data API)

On 3/1/2012 4:22 AM, Harald Alvestrand wrote:
> On 03/01/2012 12:26 AM, Cullen Jennings wrote:
>> I'd rather see it done in SDP as you were proposing - otherwise we end
>> up reinventing the whole thing for no good reasons. And I think the
>> SDP can associated the label with the correct SCTP channel. It will be
>> trivial and not require any code that we did not have to already write
>> for the RTP.
> Since I'm continually imagining a setup with the two clients in a single
> office in northern Finland and the server that relays SDP in a
> datacenter in Japan, I am hesitant.

Agreed

> Of course, there's always the option of running the SDP on channel zero
> of the SCTP channel, a la what I've heard people claim that Facetime
> does, which would make *that* one a non-concern - but that scenario
> doesn't work for the cases where the relaying server for some reason
> needs to inspect and/or modify the SDP (of which I think I've heard
> some, but I'm not sure if any of those scenarios need data).

I think many applications would be smart to immediately open a 
"signaling" dataChannel and use it for renegotiations - except it may 
have to fall back to server-relay to handle IP address changes and 
re-ICE-connection!  Probably smart to send the renegotiation message by 
*both* direct dataChannel and indirect server paths.  But this is in the 
app's domain.

-- 
Randell Jesup
randell-ietf@jesup.org

Received on Thursday, 1 March 2012 13:52:34 UTC