Fwd: Proposal for output device selection

The proposal below was recently made to the WebRTC Media Capture Task Force.

A discussion with chairs and staff indicated that this is possibly a 
better fit with the scope of the
HTML working group; therefore I'm forwarding the proposal here.

Commentary on the Media Capture list has been largely supportive of the 
concept, but we have some worries about the need for fingerprinting 
protection and possibly asking the user for permission to get 
information about the devices attached to the user's computer.

Comments (including comments on where it should be worked on) welcome!

             Harald Alvestrand, chair, WebRTC Media Capture Task Force


-------- Original Message --------
Subject:  Proposal for output device selection
Resent-Date:  Mon, 12 Aug 2013 23:44:13 +0000
Resent-From:  public-media-capture@w3.org
Date:  Mon, 12 Aug 2013 16:43:25 -0700
From:  Justin Uberti <juberti@google.com>
To:  public-media-capture@w3.org <public-media-capture@w3.org>
CC:  Harald Alvestrand <hta@google.com>, Victoria Kirst 
<vrk@google.com>, Tommy Widenflycht (ᛏᚮᛘᛘᚤ) <tommyw@google.com>, Tommi 
Gunnarsson <tommi@google.com>



WG,

With the work done on MediaStreamTrack.getSources 
<http://MediaStreamTrack.getSources> (previously known as 
getSourceInfos/getSourceIds), we now have a clean way to enumerate, 
select, and remember audio and video input devices. However, we are not 
yet able to do this for audio output devices. This is a significant 
problem for scenarios where you want audio output to go to a headset 
that is not the default device, e.g. a USB or Bluetooth headset.

Note that this goes outside the bounds of WebRTC - once we have an 
output device ID, we need a way to tell other HTML5 APIs, such as an 
<audio/> tag or Web Audio context, to use the specified device. 
Therefore locating this output device enumeration API on 
MediaStreamTrack (similar to getSources) is probably not the right fit.

We therefore propose the navigator.getMediaSinks method as an API to use 
for output device enumeration, and a new HTMLMediaElement.sinkId 
property to allow the ids returned by getMediaSinks to be supplied to 
<audio/> and <video/> tags. (See full details below)

getMediaSinks works overall similarly to getSources - it asynchronously 
returns a list of objects that identify devices, and for privacy 
purposes the .label properties are not filled in unless the user has 
consented to device access through getUserMedia.

If we like this design, it may make sense to move 
MediaStreamTrack.getSources to also be on the navigator object, for 
consistency.

----------------------------------------------------------------------------------------------

*New enumeration API*

// async API, returns results through SinkInfoCallback
void navigator.getMediaSinks(SinkInfoCallback)

// similar to SourceInfoCallback
callback SinkInfoCallback = void (sequence<SinkInfo>)

// similar to SourceInfo
dictionary SinkInfo {
  DOMString sinkId;
  DOMString kind;
    DOMString label;
};

*New API on HTMLMediaElement*

// when set, specifies the desired audio output device to use
DOMString HTMLMediaElement.sinkId

*Usage*

// print out the available audio output devices on the console
function listAudioDevices() {
navigator.getMediaSinks(printAudioDevices);
}

function printAudioDevices(sinks) {
   for (var i = 0; i < sinks.length; ++i) {
      if (sinks[i].kind === "audio") {
  console.log(sinks[i].sinkId + " : " + sinks[i].label);
      }
   }
}
*
*
*// *set the audio output for the <audio/> tag with the id "audio-renderer"
function selectAudioOutput(sinkId) {*
*
**document.getElementByID("audio-renderer").sinkId = sinkId;
}