MediaStream Image Capture

[dontcallmedom]

The WebRTC and Device & Sensors Working Groups would like the TAG review of MediaStream Image Capture, an API to capture still pictures from a MediaStream generating camera.

[travisleithead]

The ergonomics of the API seem aligned with good principles. E.g.:

• Uses Promises for potentially long-running operations.

Exposes ImageCapture as an "add-on" object to a MediaStreamTrack (a new constructable interface:

• Why make this a separate constructable object, with the reflective 'videoStreamTrack' member? Could this be streamlined into the MediaStreamTrack as a partial interface to it instead?
• Seems like 'getPhotoCapabilities' and the MediaStreamTrack's 'getCapabilities' are nearly the same. Is there a reason to have the separate method and property list?

Is 'grabFrame' really necessary? Isn't this already possible using a video/canvas element?

Consider allowing a DOMPoint in places where a Point2D dictionary could be used (DOMPoint from http://www.w3.org/TR/geometry-1/#DOMPoint). The DOMPoint handles points in 3d space, but the 'z' and 'w' coordinates are not required, and this may be useful for platform consistency. 

[travisleithead]

Should do a re-review at next F2F to see how the spec has been developing.

[torgo]

Discussed at Tokyo f2f…

[triblondon]

Discussed at London F2F

[torgo]

Agreed to close at f2f. @travisleithead will follow up on remaining issue w3c/mediacapture-image#142 .