Re: <canvas> and the 2D context API

Anne van Kesteren wrote:
> On Thu, 13 Aug 2009 17:59:48 +0200, Toby Inkster <> wrote:
>> On Wed, 2009-08-12 at 18:29 -0700, Maciej Stachowiak wrote:
>>> But HTML5 would likely have a normative reference to a Canvas API
>>> spec, so there is a limit. Per W3C rules, you can't normatively
>>> reference a document that's more than one maturity level apart.
>> The solution then seems to be to only include an informative reference.
>> An example of the current draft with just an informative reference to
>> the canvas API might be:
> This does not help postMessage(), which takes an ImageData object.
Is that the only instance where the HTML 5 specification drills down 
into the Canvas element's 2D API directly?

Again, this is a handshake issue, whereby one requirement the HTML 5 
specification would have is the ability to obtain a serialization of the 
Canvas object. This is an unfortunate consequence of merging and 
breaking down borders between API and declarative syntax, but I don't 
see it as a showstopper.

There's nothing in this _requirement_ that would overly inhibit 
innovation on the Canvas object. Nor is there anything likely to happen 
with the Canvas object, the 2D immediate mode API, that would impact on 
this handshake between specifications.

But you are right, in this would be normative, not informative. I don't 
see this as a showstopper though, unless we're constrained by HTML 5 not 
being able to specify a normative reference to the new 2D API, because 
the 2D API doesn't have a published specification yet.

In which case, I would think it better to remove the reference to 
ImageData from PostMessage. It could be taken up again in a later HTML 
specification, as folks are fond of saying. I've looked at the existing 
implementations of PostMessage for the various browsers. None have 
implemened ImageData yet. Even if they had, this is the risk they take 
for implementing what's current in a working draft.

As Philipe stated in the HTML-WG IRC, implementations may not support 
passing ImageData objects directly[1], at least with Worker Threads.

This is not a showstopper.



Received on Thursday, 13 August 2009 16:50:45 UTC