Re: [mediacapture-main] Clarify behavior of using wrong-type constraints

I think that the only satisfactory answer here is that - if supported 
- mandatory constraints that are nonsensical cause the track to become
 overconstrained (or not selected).  That's consistent with asking for
 fillMode on a camera that doesn't have it.

Non-mandatory constraints can say anything they like.  

Side question: are implementations required to count infinities when 
determining fitness distance?  For example, if there are two sources 
and source 1 doesn't support constraint 1 and source 2 doesn't support
 constraint 2, then you have to fall back to counting fitness distance
 on both of them based on the constraints they do support.  If you 
just count to infinity, you lose that ability.

That said, you might not always need to count infinities, just 
remember that there is one.  If source 1 doesn't support two 
constraints and source 2 doesn't support one constraint, then falling 
back on the constraints that each supports is better than saying that 
source 2 is better because 1∞+n fitness distance is always better than
 2∞+m.  Of course, if n or m is unset, then maybe you do need to fall 
back to counting infinities...

Did I mention that I think constraints are the best thing ever?

-- 
GitHub Notification of comment by martinthomson
Please view or discuss this issue at 
https://github.com/w3c/mediacapture-main/issues/409#issuecomment-252823357
 using your GitHub account

Received on Tuesday, 11 October 2016 06:16:40 UTC