- From: Jan-Ivar Bruaroey <jib@mozilla.com>
- Date: Mon, 28 Sep 2015 15:10:23 -0400
- To: Martin Thomson <martin.thomson@gmail.com>, "public-webrtc@w3.org" <public-webrtc@w3.org>
On 9/24/15 7:26 PM, Martin Thomson wrote: > Apparently these serve a purpose: shibboleth. > > We can't have an attribute that is a dictionary. This breaks > localDescription and remoteDescription on RTCPeerConnection. > > Thus, I need to restore an interface for RTCSessionDescription. > > I've tried to think of ways to make this less awful, but I can't see > any way to do so. Ideally, we'd make RTCSessionDescription attributes > immutable, but applications have come to expect the attributes on > RTCSessionDescription to be mutable. Thus, we have to have a copy of > the actual session description from the local/remoteDescription > getter. Having to do that is why WebIDL forbids dictionary > attributes. So we get the worst of all possibilities. Yeah that does stink. But aren't we in the process of deprecating localDescription in favor of currentLocalDesc? Should it be getCurrentLocalDesc() maybe, and return a dictionary? FWIW this needn't hold back the other changes, as I believe one can pass an RTCSessionDescription interface to a method expecting an RTCSessionDescriptionInit dictionary (because the interface has a serializer). Could we keep the interface just for local/remoteDescription purposes, and still do the other changes (not requiring users to use the constructor)? That still seems like a win to me. .: Jan-Ivar :.
Received on Monday, 28 September 2015 19:10:52 UTC