[whatwg] <video> ... <script> race condition

[Philip Jägenstedt]

On Fri, 13 May 2011 12:25:39 +0200, Henri Sivonen <hsivonen at iki.fi> wrote:

> On Fri, 2011-05-13 at 11:40 +0200, Philip Jägenstedt wrote:
>> Problem:
>>
>> <video src="video.webm"></video>
>> ...
>> <script>
>> document.querySelector('video').oncanplay = function() {
>>    /* will it run? */
>> };
>> </script>
>>
>> In the above the canplay event can be replaced with many others, like
>> loadedmetadata and loadeddata. Whether or not the event handler has been
>> registered by the time the event is fired depends on how fast decoding  
>> is,
>> how fast the network is and how much "..." there is.
>>
>> This is not new information, Simon Pieters wrote extensively about the
>> problem in
>> http://dev.opera.com/articles/view/consistent-event-firing-with-html5-video/
>>
>> The reason I write now is that in one of our projects this exact race
>> condition happens to be hit for http://html5demos.com/video which is  
>> where
>> the above code snippet comes from.
>>
>> Does anyone have ideas about how to deal with this problem, which *is*
>> going to appear on many real-world pages?
>
> <iframe src=foo.html></iframe>
> <script>
> document.querySelector('iframe').onload = function() {
>    /* will it run? */
> };
> </script>
> has the same problem. The solution is using the onload markup attribute
> that calls a function declared in an earlier <script>:
>
> <script>
> function iframeLoaded() {
>   /* It will run! */
> }
> </script>
> <iframe src=foo.html onload=iframeLoaded()></iframe>
>
> Separation of markup and event handlers is overrated.
>

I wasn't asking how to work around the problem once you know it exists, I  
was wondering if any browser vendors have done anything to make this  
problem less likely to happen on pages like http://html5demos.com/video  
that don't do the right thing?

-- 
Philip Jägenstedt
Core Developer
Opera Software