- From: Stefan Håkansson LK <stefan.lk.hakansson@ericsson.com>
- Date: Thu, 30 Apr 2015 05:29:52 +0000
- To: Peter Thatcher <pthatcher@google.com>, "public-webrtc@w3.org" <public-webrtc@w3.org>
I like this approach. Stefan On 22/04/15 01:10, Peter Thatcher wrote: > While talking to Stefan about the possibility of per-RtpSender control > of VAD, and thinking about all the recent discussion around MSID, I > realized that there was a common need to add parameters to addTrack. > > I propose we add a dictionary parameter "RtpSenderInit", like we have > DataChannelInit that contains parameters for > PeerConnection.createDataChannel without making the parameter list a mess. > > IDL that would retain current semantics could look like this: > > > partial interface PeerConnection { > RtpSender addTrack(MediaStreamTrack track, RTCRtpSenderInit init); > } > > dictionary RTCRtpSenderInit { > sequence<MediaStream> streams; > } > > > A more advanced form, if we choose to add more parameters later, could > look like this: > > dictionary RTCRtpSenderInit { > // These are based on our discussion of MSID. > DOMString label; > sequence<DOMString> synchronizationGroups; > > // The initial RtpParameters to use, as if you call .setParameters > immediately after > // calling addTrack. > RtpParameters parameters; > } > > dictionary RtpParameters { > // This is based my discussion with Stefan. > boolean voiceActivityDetection; > } > > > > Let's continue the discussion around voiceActivityDetection, > synchronizationGroups, and label/ID on separate threads, but for this > thread, I'd like to ask the group: Does it look like a good idea to > follow the createDataChannel(..., DataChannelInit) pattern by having > addTrack(..., RtpSenderInit)? > > > I'm in favor of it because: > 1. We'd avoid with a mess of parameters in addTrack over time. > 2. It gives the app to specify initial RtpParameters (such as VAD, > potentially). > 3. It follows the same pattern we already have for createDataChannel, > which has been successful in adding many options without making the > parameter list a mess. > 4. It gives us more flexibility in our work around IDs and labels. > > Thanks, > Peter >
Received on Thursday, 30 April 2015 05:30:19 UTC