Re: On reserving and releasing resources

Jim Barnett wrote:
> I think that tying permission and possession of the device (your
> definition 2) is a simpler model, and easier to understand.  On the
> other hand, it's less flexible.

I actually have the opposite view of its flexibility. Mandating that 
resources are automatically released when a stream has no consumers 
means that if I want to, say, disconnect the camera from the current 
PeerConnection (because the call drops, for example), I need to add it 
to a fake consumer right away or I risk losing it (if I'm trying to 
re-dial). I'm sure others could think of other examples where this might 
make sense.

But I do agree that 2) seems easier to understand. Particularly for 
users... if you let a page ask for permission as soon as it loads, but 
it doesn't actually use that permission (and there's no green light or 
whatever to remind them the camera is on), they might be quite surprised 
(and have a quite different opinion) by the time the page does actually 
want to turn the camera on.

Received on Tuesday, 8 May 2012 16:23:47 UTC