- From: Harald Alvestrand <harald@alvestrand.no>
- Date: Thu, 19 Dec 2013 23:10:21 +0100
- To: public-webrtc@w3.org
On 12/19/2013 07:40 PM, Martin Thomson wrote: > On 19 December 2013 10:30, Justin Uberti <juberti@google.com> wrote: >> Agreed. I was thinking that the stream could be obtained from the track, >> which pushes this somewhat upstream to the MediaStream spec. > Yeah, that doesn't really work. Besides creating a circular > reference, a track can be part of multiple streams. Learning all of > them would suck a little. > Well, on the outgoing end, either you tell the PC explicitly what streams to teach the remote end about (AddStream), or you let the PC figure it out (AddTrack with implicit addition of all streams it's a member of). I don't see a third option (unless you're suggesting that the app needs to call *both* AddStream and AddTrack). I don't see a huge cost in transmitting the info on all the streams a track is a member of, and I don't see a huge cost at the receiving side in being able to know about some MediaStream objects that it otherwise has no use for. But this kind of illustrates my point from Seattle/Shenzen that when we change the PC model from stream to track, there is going to be a certain amount of effort involved in adjusting everything until it fits the new model. We did not get all the points clear in Seattle/Shenzen.
Received on Thursday, 19 December 2013 22:10:30 UTC