Re: Public feedback on HTML5 video

Jonas Sicking, Mon, 28 Dec 2009 23:31:32 +0100:
> On Mon, Dec 28, 2009 at 5:36 PM, Leif Halvard Silli:
>> I assume that the reason e.g Webkit has been eager to
>> buffer things, has been in order to provide a better user experience -
>> my theory has always been that improved user experience has been the
>> motivation for this. Therefore I can very well understand that one is
>> sceptical about putting the ultimate judgement into the hands of the
>> user agents.
> 
> Your assumption does not match the impression I got from Maciejs
> email. It sounded like they only wanted to download enough information
> to calculate the duration of the video, but limitations in the
> QuickTime framework used to implement <video> made it impossible, or
> too hard, at this time to do that without downloading the entire
> video.
> 
> So it sounded like the choice to download was one based on technical
> limitations, not one based on desired user experience.

The advantage of buffering is that the experience is more speedy. The 
disadvantage is that the content could be outdated. In the olden days 
Safari also had some problems with updating buffered stylesheets. I 
don't say that user agent vendors doesn't do as good as they can. But 
it is better if we can agree on what good behaviour should be. This 
would also help e.g. extension developers, as you mention below.

>> Can we be certain that the user agent _does not_ autobuffer when the
>> autobuffer attribute isn't used? Or is this also up to the user agent?
> 
> We can't ever be certain UAs will, or won't, do anything. As a
> developer of firefox *I* can't even be certain that firefox will, or
> won't, do anything. This is because users install addons into firefox
> that do all sorts of things that are outside of my control. I'd
> certainly expect someone to write an addon that affects how buffering
> will work in firefox, and I expect users to install that addon.
> 
> So yes, it is definitely up to the user agent.
> 
> What we can do, is to say what behavior is conformant to the HTML5 spec.

And the current spec text makes it a bit hard to say whether the user 
agent - or the extension - breaks the spec, since implementations are 
free to buffer even if there is no autobuffer attribute, and free to 
not buffer even when there is a autobuffer attribute.

I guess I liked Edward's proposal the best [1].

[1] 
http://www.w3.org/mid/3b31caf90912281544s7d45535ep11d1384e3fe6e64c@mail.gmail.com
-- 
leif halvard silli

Received on Tuesday, 29 December 2009 01:46:57 UTC