Moving on with noaccess/peerIdentity

We've had a lot of confusion about the role of noaccess and
peerIdentity in WebRTC over the time that they have been proposed,
discussed, etc...  I want to concentrate on finalizing what is in gUM
first, because while there are still some small issues to resolve on
the WebRTC side, gUM is straightfoward.

I want to propose the following set of changes.

1. remove noaccess (that's been discussed for a while now, and the time is ripe)

2. move the location of the peerIdentity constraint.  As a
fully-fledged constraint, it turns out that it is difficult to
properly vet input for a mixed audio+video stream:
  var constraints = {};
  constraints.audio = { peerIdentity: 'x@nfcform.domain' };
  constraints.video = { peerIdentity: 'x@nonnfcform.but.equivalent.domain' };
  navigator.getUserMedia(constraints, ...);
This is important, since it makes no sense to have different
constraints on different tracks in the same stream.  Mixing isolation
restrictions causes great headaches already to the point where Firefox
at least just marks mixed content (!) as fully isolated.  So I
propose:
  navigator.getUserMedia({audio: true, video: true, peerIdentity:
'x@example.com' }, ...);

Now, since I've implemented these, I can be pretty confident that
these are sound, so I made a pull request on the spec that does this.

https://github.com/fluffy/webrtc-w3c/pull/10

This includes a refined text regarding mixed isolation levels for
streams that contain peerIdentity and non-peerIdentity tracks.

Received on Thursday, 6 February 2014 00:50:46 UTC