[whatwg] ARIA

[James Graham]

Aaron Leventhal wrote:
> James Graham wrote:
>> Dave Hodder wrote:
>>> The current HTML 5 draft doesn't mention ARIA anywhere.  Perhaps it 
>>> should clarify the relationship (or non-relationship as it is at 
>>> present), even if it's only a brief mention in section 1.1.
>>
>> Unfortunately a brief mention is insufficient as aria functionality
>> overlaps substantially with HTML functionality and so processing
>> requirements for aria-in-html need to be carefully considered (so we can
>> answer questions like "how does <div aria-role='heading'> affect the
>> outline algorithm"). This has not yet happened.
>>
>>
> Okay, so I can speak to this. I developed first browser implementation 
> of ARIA -- the one in Firefox.
> 
> ARIA doesn't really overlap with HTML, because ARIA only reports what a 
> JS developer is using elements for.
> So ARIA semantics should not affect  behavior. Any code for dealing with 
> ARIA markup should be strictly in the accessibility API support code 
> (MSAA/IAccessible2/ATK/AT-SPI/UI Automation/Universal Access).

This seems like a poor idea from the point of view of authors actually 
using the markup correctly as they will have to use special 
accessibility-API aware tools to inspect the effects of their aria-* markup.

> A div need not affect the outline algorithm, etc. any more than a div does.
> Thus it should not be complicated to mention ARIA in the spec.

So, given markup like:

<section>
<h1>foo</h1>
<div aria-role='heading'>bar</div>
</section>

would AT questioned on what the headings and subheadings of the 
<section> are  report on the <h1> alone (as specified by the current 
outline algorithm)?

Similarly, if I have a table like:

<tr><th id="a">1</th><th id="b">2</th></tr>
<tr><td>A</td><td aria-describedby="a">B</td></tr>

would AT report the cell headings for B as 2, not 1, despite the 
aria-describedby attribute?

If the aria-* attributes overrule the HTML-native algorithms then aria 
creates an odd (and IMHO unpleasant) situation in which a document's 
semantics depend on the type of client being used to view it.

What's the easiest way to test existing aria implementations on 
Mac/Linux (I don't have access to a Windows box)?

-- 
"Mixed up signals
Bullet train
People snuffed out in the brutal rain"
--Conner Oberst