Re: An HTML language specification

On Fri, 21 Nov 2008, John Kemp wrote:
>> All I was suggesting was that the exact same HTML5 language which is
>> being defined in the current HTML5 draft is put into a separate
>> document, describing only the actual language. I believe that is what
>> Michael has already attempted.

>> I fail to see how that particular operation affects the possibility to
>> achieve well-defined specifications or interoperable implementations.

Ian replied:

> Mike's document doesn't contain implementation requirements. I'm not sure
> how to extend Mike's pattern to a spec that also includes implementation
> requirements. For a simple example, consider <video src="">, whose
> implementation definition is deeply integrated with the definition of the
> HTMLMediaElement API. How would you separate the two cleanly?

Could you be more explicit?

As best I can tell, Mike's document says:

> src = uri
> The src content attribute on media elements gives the address
> of the media resource (video, audio) to show. The attribute, if present,
> must contain a valid URL.

You might want to add that a video must have either the src attribute
or at least one contained <source> element, but not both.

But I don't see any reliance on the HTMLMediaElement scripting API.

I see that a browser needs to know about both.

I see that some web applications will want to use scripts to provide
a richer or more tailored UI.

But if I as an author (or the writer of a simple authoring tool) just want
to embed a video with default video look and feel, default controls, etc
... similar to what I get today with object ... I don't see why I would
need to know anything about the HTMLMediaElement scripting API.

So leaving that in a separate processing spec seems perfectly
reasonable; the processing spec would depend on the markup
semantics spec, but the dependency would be in only one direction.


Received on Sunday, 23 November 2008 03:24:33 UTC