Re: Thoughts on nested rowgroups?

> On Apr 2, 2015, at 8:30 AM, Joseph Scheuhammer <clown@alum.mit.edu> wrote:
> 
> Hi James,
> 
> On 2015-04-02 2:13 AM, James Craig wrote:
>> The rowgroup role was not intended to be mapped to AX APIs (they hadn't previously mapped thead/tfoot/tbody) and still isn't mapped by most User Agents. Firefox is the exception. IIRC, Alex added it a couple years ago, though the API mapping guide does require it.
> 
> The AAPI task force has been working on better mappings for role rowgroup for the past month, and the mapping specs are likely going to change.
> 
> We are converging on the following:  If an element has role="rowgroup", it is generally not mapped.  If, however, it is focusable, is interactive, or has some global aria-* property on it, then it will be mapped.
> 
> If you agree, how would AX API map it?

If there are other attributes on it, then presumably we'd map it as something other than a rowgroup. 

Tables and grids are tricky. WebKit glosses over as many author errors as it can, but when an author vends a malformed grid, WebKit avoids exposing it to the APIs as a malformed table. WebKit instead falls back to exposing the content in generic containers (e.g. just back to the div soup the author provided). A focusable rowgroup should probably be one of the heuristics that would cause WebKit to refrain from exposing the table to the API as a table.

I mentioned this during the CR phase of ARIA 1.0. Many of the test case examples were malformed (e.g. <div role="row"></div> with no child cells or containing grid). Mozilla mapped these directly and left it to the assistive technology to figure out, but WebKit avoids exposing orphaned child roles to the APIs.

James

Received on Friday, 3 April 2015 09:11:43 UTC