Philipp Hancke

Trying to get an estimate for the BWE probe overhead (service wide) which as you say is accomplished in many ways. Bonus: being able to visualize it live in the...

@henbos thoughts?

well, the unified plan that ships in Chrome is actually a bastard thing as it has a=ssrc:12345 msid:something lines in addition to the standard a=msid. We should/could update the example...

I do not think the spec definition is very useful. It would have prevented detection of the infamous ["macos audio bug"](https://bugs.chromium.org/p/webrtc/issues/detail?id=4799) which showed up as "no bytes sent". Easy fix:...

You are proposing to intentionally curtail an implementation behavior that has demonstrably proven to be more useful than harmful for the sake of not updating a text that is not...

which is a great catch and I agree the current behavior of some implementations is potentially harmful in some rare cases. But the spec behavior is more harmful so the...

> vendors exposing stats ahead of negotiation which I hope we can agree is unnecessary. Correct, that should go away. I think such a change won't cause too much breakage....

> So then, in your mind, when would stats appear? After negotiation? After connection (and negotiation)? After negotiation so a test that drops all ice candidates can check they exist....

#302 looks like the intent was "after negotiation". But then JSEP tells us there shall be no ssrc lines. I think we need the following test assertions: 1. createOffer->SLD does...

but if we have "standard signaling" shouldn't the stats have a `mid` attribute (the transceiver has)?