User Agent: Mozilla/5.0 (Macintosh; Intel Mac OS X 10_10_2) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/41.0.2272.89 Safari/537.36 Steps to reproduce: I have two cameras: - camera A (integrated FaceTime camera) - camera B (external Logitech USB camera) In the dialog following a call to getUserMedia: 1. Select camera A 2. Set "always allow" access. - Firefox (36.0.1) - OS X (10.10.2 (14C1510)) Actual results: Subsequent calls to getUserMedia will receive input from camera B. Expected results: Subsequent calls to getUserMedia should receive input from camera A.

"Always allow" remembers that you granted access to a camera; it doesn't remember which camera. The code (at http://mxr.mozilla.org/mozilla-central/source/browser/modules/webrtcUI.jsm?rev=56da5d401fe2#304) just uses the first camera of the list when there are persistent permissions. In most cases this seems to do the right thing, but I'm not really surprised that the result can be unfortunate in some cases. I vaguely remember a discussion in the past where I argued that we should separate device access permission (what the prompt is for) which is site-specific and device selection, which should probably apply to all websites, and could be set in the preferences window. I think someone replied that in most cases the default device can be changed from the operating system's settings, so duplicating that in Firefox's preferences wouldn't be useful. While it would not fully fix the problem, I think letting the user change the device after the permission has been granted (bug 880312) would mitigate the issue.

Also, we expect to land device enumeration/ID support soon; which will enable a site to request a specific camera (once it has permissions), such as "the same camera I got last time".

Once Bug 1212501 is fixed, it should be possible to implement memory here that doesn't conflict with constraints (more in that bug).