Re: Survey ready on canvas proposal

On Feb 22, 2010, at 4:13 PM, Richard Schwerdtfeger wrote:

> On Feb 22, 2010, at 2:23 PM, Michael Cooper wrote:
>
> >> There is a one-question survey ready on the proposal from the  
> canvas
> >> sub-team:
> >>
> >> http://www.w3.org/2002/09/wbs/44061/20100225_canvas/
> >>
> >> Please take a moment to provide your view. While the survey will be
> >> open until Thursday, if you can, try to provide your input by the
> >> end of Wednesday in order to allow input to be incorporated prior  
> to
> >> Thursday discussions.
> >
> >I'd like to ask for a clarification on something that is not fully
> >clear to me from the proposed spec changes. If adom is not set, are
> >user agents forbidden to expose the children of <canvas> to assistive
> >technologies? Or is it meant as a hint that they don't need to, and
> >the contents may not meet the stated requirements?
> >
> >Specifically, this is the sentence that is not totally clear to me:
>
> >  "The default value for adom is false to indicate that the canvas
> >subtree is only used as fallback content and may not be used as an
> >accessible DOM subtree representation of what is drawn on canvas."
>
> Maciej,
>
> What we are saying is that when adom is set it indicates the subtree  
> is mapped to the accessibility API as the representation of the  
> <canvas> rendering. If it is set to all it is NOT the accessible  
> representation of the <canvas> rendering and should not be exposed  
> to the AT. However, if the fallback content is rendered instead of  
> <canvas> (canvas is hidden) then by all means it should be mapped.  
> It is for this reason we did not make a MUST not. Are you suggesting  
> that to address this point we should make it a MUST NOT?
>
No, to the contrary. I think it should be always allowed to always  
expose the content of the <canvas> element to AT. I did not see a  
justification for ever not exposing <canvas> content in the proposal  
linked from the survey.

Therefore I will propose adopting this proposal with one of the  
following changes:
A) Allow <canvas> children to always be exposed to AT, even if adom is  
not set; OR
B) Provide a rationale for not exposing this content to AT in some  
cases (this would likely include not exposing it for any currently  
existing <canvas> elements).

You can see what I entered on the Results page: <http://www.w3.org/2002/09/wbs/44061/20100225_canvas/results 
 >.

I note also that my comments are purely personal opinions about the  
technical issues, and are stated with my HTML WG co-chair hat off.
> >Is the "may not" in that sentence meant to be a UA requirement, and  
> is
> >it meant to be mandatory or optional? Does "used as an accessible DOM
> >subtree representation of what is drawn on canvas" apply to any form
> >of exposing the content to AT?
>
> I think I answered this above. If not please let me know. ... Should  
> we be more prescriptive in the text?
>

I think in official conformance language, "may not" simply means the  
UA is allowed not to. It doesn't mean the UA is required not to. Thus  
I think a MUST NOT would better match your intent, as you described  
it. Though as I said above, I am not sure I support that approach.

Regards,
Maciej