- From: Peter Thatcher <pthatcher@google.com>
- Date: Tue, 21 Jan 2014 10:58:40 -0800
- To: Bernard Aboba <Bernard.Aboba@microsoft.com>
- Cc: Stefan HÃ¥kansson LK <stefan.lk.hakansson@ericsson.com>, "public-orca@w3c.org" <public-orca@w3c.org>, "Harald Alvestrand (hta@google.com)" <hta@google.com>, "Adalberto Foresti (MS OPEN TECH)" <aforesti@microsoft.com>
On Tue, Jan 21, 2014 at 10:51 AM, Bernard Aboba <Bernard.Aboba@microsoft.com> wrote: > Stefan said: > >> I think it is a bad idea to not signal and just stop sending RTP - the >> receiver would then have to guess if the sender stopped sending temporarily, >> or if there are a lot of packet losses, or if there is a speech pause (and >> CN should be generated) etc. > > [BA] I would agree that a packet may need to be sent as the result of a stop(), pause() or resume(). > However, it looks to me like the action may depend on the specific configuration and codec. > > An example of this is SST SVC where in some codecs (e.g. VP8) there may be no packet sent to > indicate dropping a layer, just a reconfiguration of the encoder, whereas in others (e.g. H.264) > a codec-specific packet might be sent (SEI layer drop). > > In MST SVC it is possible that the situation could be different if we can assume thatl WebRTC implementations > understand TMMBR and therefore a TMMBR message with maximum bandwidth of zero might be > used to pause individual layers (or simulcast streams, for that matter). > > Peter Thatcher said: > > "I still think signalling is up to the app. If the sender wants to > signal to the receiver that start/stop/pause/resume/whatever is called > on the sender side, in whatever signalling mechanism the app defines. > The question I would ask is "what API does the receiver need to pass > down information into the browser when such information has been > signalled from the sender?" I can't think of any of the the top of my > head, but there might be something." > > [BA] I agree that signaling is up to the app, but in addition, it looks to me like the actions taken > as the result of a stop(), pause() or resume() may also need to be under application control > as well. Yes, I agree that the sender should be able to control what is sent when it calls stop/pause/resume/whatever.
Received on Tuesday, 21 January 2014 18:59:49 UTC