- From: Stefan Håkansson LK <stefan.lk.hakansson@ericsson.com>
- Date: Wed, 16 Jan 2013 14:45:17 +0100
- To: public-media-capture@w3.org
On 2013-01-16 08:45, Adam Bergkvist wrote: > Hi > > On 2013-01-16 01:08, Travis Leithead wrote: >> I'm still thinking about the user consent question. > > I think rendering frames on a canvas and grabbing them is as big of a > privacy issue as recording a stream with the recorder. So if we require > consent for one type of frame grabbing, we should do it for all. I don't think we need user consent. Even without the recording API the UA could ship (via a PeerConnection) the media off to a server which records. > > If I remember correctly, we talked about "tainted" streams. I guess we > could say that such streams are not recordable. The same goes for a > stream that has been created and associated with a specific receiver > identity. > >> For the settable parameters, I think it makes sense to follow the >> constraints pattern in the other spec—readable properties on the >> Recorder with a mechanism for overlaying constraints on top of that. > > I agree. We should use the pattern we've worked out consistently > whenever we have settings of this kind (i.e. non-trivial settings that > may not be supported, may fail or may not be combinable with other > settings). It's not the easiest pattern for developers to learn so they > should only have to learn one variant of it. > > In v6 of the settings proposal the methods: getConstraints(), > setConstraints(), applyConstraints(), and the rest of the constraints > methods belong to a "partial interface MediaStreamTrack". Perhaps we > should extract those methods into a separate interface > "Constrainable"/"ConstraintTarget"/"or something". That interface could > be implemented by MediaStreamTrack objects, the recorder, the dtmf > sender, the transport handler and any other objects that needs this > functionality. > > So an object using this would still list all its constraints as > attributes, but the functionality to modify them would be imported by a > single "InterfaceName implements Constrainable" statement. I think it > would save us a lot of work not having to duplicate spec text (and code) > and make sure the constraint mechanism stays consistent for all objects > that implement it. I think this is a good idea. I guess it would not only help spec writers and app developers, but also implementers. > > /Adam > >
Received on Wednesday, 16 January 2013 13:45:43 UTC