Re: Issue 204 CP objections [Was: TF Teleconference Minutes for 26 April]

On Fri, Apr 27, 2012 at 2:01 AM, Leif Halvard Silli
<xn--mlform-iua@xn--mlform-iua.no> wrote:
> Benjamin Hawkes-Lewis, Thu, 26 Apr 2012 22:33:40 +0100:

[snip]

>> But it's also plausible that such authors might use @hidden with the
>> intent of hiding elements even from accessible names and descriptions.
>> For example, @aria-describedby might reference text that has the
>> @hidden attribute until it is relevant to the state of the
>> application, for example error text. Consider the following example
>>
>>   <label for=username>Username:</label>
>>   <input id=username name=username aria-describedby=error>
>>   <div hidden id=error>Invalid username. Must be six or more
>>                        alphanumeric characters.</div>
>>
>> If "hidden" means never ever render the content, then the error
>> message doesn't get included in accessible description calculation
>> until the error occurs and the "hidden" attribute removed.
>>
>> If "hidden" means do not render the content except for ARIA name and
>> description calculation, then the error message is inappropriately
>> included regardless of the actual state of the application.
>
> To make the above not work, one must change ARIA so that
> aria-describedBY="section-where-aria-hidden-is-set-to-true"
> does not work.

Or for HTML to define the "text alternative" of elements with @hidden
to be the empty string. Host languages define how text alternatives is
computed from their native semantics.

> One could of course try to help by flagging it as an error to make
> aria-describedby point to @hidden and @aria-hidden="true" sections.

Jonas is trying to help users of content written by authors who don't
use conformance checkers.

> However, ARIA *also* says that one MUST do
> <p style="display:none" aria-hidden="true">Hidden text.</p>

I've queried the appropriateness of that requirement:

http://lists.w3.org/Archives/Public/public-pfwg-comments/2012AprJun/0007.html

> the use of @hidden does lead to
> some undpredictability too: The element could be made visible, via CSS,
> despite the @hidden attribute. What happens then?

As the spec stands, user agents should render it. See especially:

   http://dev.w3.org/html5/spec/rendering.html#hidden-elements

(Personally, I think it would be simpler if, like <noscript> with JS
enabled, it was never rendered.)

> When I read the rest of that letter, the I came to think that it
> perhaps should be forbidden to do
> aria-describedby=hidden-anchor-element.

Based on how aria-describedby is currently specified and implemented, maybe.

--
Benjamin Hawkes-Lewis

Received on Saturday, 28 April 2012 20:40:44 UTC