Re: Use of hidden labels/descriptions in the wild? [Re: HTML-A11y TF minutes, 17th May 2012]

On Tue, May 22, 2012 at 12:56 PM, Charles McCathieNevile
<chaals@opera.com> wrote:
> On Sat, 19 May 2012 01:57:58 +0200, Benjamin Hawkes-Lewis
> <bhawkeslewis@googlemail.com> wrote:
>
>> On Fri, May 18, 2012 at 8:07 PM, John Foliot <john@foliot.ca> wrote:
>>>
>>> Benjamin Hawkes-Lewis wrote:
>>>>
>>>>
>>>> In the minutes John Foliot is recorded as saying: "Benjamin is saying
>>>> that it shouldn't be allowed to happen. It's already happening, so all
>>>> we're trying to do is document the truth."
>>>>
>>>> Did he just mean that some browsers today expose @hidden content to
>>>> accessible name and description calculation or is the assertion that
>>>> there's some actual web content in the wild that depends on this
>>>> behavior?
>>>
>>>
>>> "Still works when the link is hidden using the HTML5 hidden attribute or
>>> CSS display:none"
>>>
>>>
>>> http://www.paciellogroup.com/blog/2012/05/firefox-14-image-long-description-via-link-using-aria-describedby/
>>
>>
>> Cheers for the clarification. Browser behavior is comparatively easy
>> to change to whatever the WG decides is best.
>
>
> The lesson from history (most recently writ large by the -webkit- prefix
> debate, but repeated pretty often) is that this is only true if you convince
> the browser makers that what the working group decides will be in the best
> interests of their customers. Which is no surprise - authors take the same
> approach to working groups, as shown by the same example.
>
> Expecting the world to work some other way isn't a very effective approach
> to developing standards in a free market.

I agree.

My point is simply that nobody has shown that authors are *today*
relying on aria-describedby pointing to @hidden and being surfaced as
an accessible description. Consequently, there are no demonstrated
compatibility constraints on user agents or conformance checkers;
instead, people are making arguments based on plausible conjectures
about what authors might or might not do. Contrast this with (say)
<canvas> where one of the major reasons for HTML WG to adopt <canvas>
into HTML5 was that it was already used in the wild so UAs had to
support it.

(I've been wondering if the Common Crawl corpus for 2012 might provide
some evidence on current usage of @hidden one way or the other, but
haven't yet freed up enough time/energy from my day job to properly
investigate.)

--
Benjamin Hawkes-Lewis

Received on Tuesday, 22 May 2012 15:16:15 UTC