docfaraday

The more I think about this, the more I think it is a mistake to omit the rid when there is a (single) rid in SDP negotiation. This is because...

> > ...what's "currently being used". > > Unless... is early media still a thing? Is a receiver in "have-local-offer" or "have-remote-offer" prepared to receive any payload types defined in...

Similarly, do we want a rollback to _restore_ encodings that were removed by a sRD(offer)? Including any scaleResolutionDownBy etc that was set by JS?

> So [should the following](https://jsfiddle.net/jib1/2yzw5ef1/) produce 1,1,1 or 4,2,1? _That's_ an interesting question. It gets more interesting when you ask what happens when that setParameters occurs right after a sRD(initial...

Or, what happens if only some of the scaleResolutionDownBy are unset?

I think my inclination is to just set to 1 regardless. Too many wonky corner cases pop out if you do anything but that. Or, I guess we could throw...

There are a handful of types of race here, on further thought. addTrack, then a race between sRD(simulcast recv offer) and setParameters(unicast) - There are multiple questions here. First we...

> One way to lift our ban on _"[remotely initiated RID renegotiation](https://w3c.github.io/webrtc-pc/#set-description)"_ (to comply with JSEP O/A) without giving up these invariants would be: once envelopes have been established, defer...

But you _have_ gathered candidates. The ICE restart has not been committed yet, and the old ICE transport is still being used.

As for why we should care, having the gathering state sometimes transition through new if the timing is just right (wrong?) makes it really difficult to write tests.