Re: [whatwg] [whartwg] Video processing ability for MediaStreamTrack

Hi, all,
Thanks for Anne's feedback. After irc with Anne, we got some consensus.

1. addVideoProcessor()  doesn't need to check duplicates due to it will
create a new MST. So the worker will append to the new created MST. It
means every MST will contain only one video processor. I will explicitly
point it out in next version.
2. It is a typo.
3. The event need a constructor and use Exposed. I will change it later.
4. outputImageBitmap=null => My bad. I will fix it later.
5. The ImageBitmap is immutable. It is not ImageData. So this might not an
issue.
6. Also Anne suggests me to write down the algorithms including a. dispatch
event, b. how to use outputImageBitmap... will add those into next version.

BR,
CTai

2015-08-11 13:59 GMT+08:00 Anne van Kesteren <annevk@annevk.nl>:

> On Fri, Aug 7, 2015 at 8:56 AM, Chia-Hung Tai <ctai@mozilla.com> wrote:
> > http://chiahungtai.github.io/mediacapture-worker/
>
> Given that removeVideoProcessor() does not take arguments, should
> addVideoProcessor() not check for duplicates?
>
> VideoProcessEventThe looks like a typo.
>
> The events don't define constructors. They probably should. You should
> also use IDL [Exposed=] syntax to indicate they are available in
> workers. (Or dedicated workers, anyway.)
>
> The outputImageBitmap=null syntax is wrong. You want to remove "=null"
> there and define the default in prose.
>
> You also want to define the processing model a bit more carefully I
> think. E.g., it seems to be that for processing the event instance is
> modified and then once the dispatch flag is unset that data is copied
> somehow. But how is it copied? Is it a structured clone transfer that
> detaches the buffer?
>
>
> --
> https://annevankesteren.nl/
>

Received on Sunday, 16 August 2015 09:13:49 UTC