Re: PFWG-ISSUE-701: Modify aria-owns such that the idref order pertains to the order of the associated children

Hi, Joseph. This one looks an edge case since I don't see any practical
outcome of doing this. So I would fall it back into common procedure, i.e.
if we make the hierarchy to respect aria-owns order then these elements
have to be swapped. I would suggest to change that wording to something
like:

"Author SHOULD NOT use aria-owns to change parent/child relationship
presented in the DOM if this breaks logical relationships between elements."


On Tue, Feb 17, 2015 at 11:15 AM, Joseph Scheuhammer <clown@alum.mit.edu>
wrote:

> Alex,
>
> On 2015-02-15 11:07 AM, Alexander Surkov wrote:
>
>> The order the referred elements listed in the value should be preserved
>> when their parent-child relationship is set. All explicit unreferred
>> children should be considered followed aria-owns elements
>>
>
> That sounds right to me.
>
> But, there is text in the spec for aria-owns [1]:
>
> "Authors SHOULD NOT use aria-owns as a replacement for the DOM hierarchy.
> If the relationship is represented in the DOM, do not use aria-owns."
>
> What if the author does the following -- a modification of your example:
>
> <div role="grid">
> <div role="row" aria-owns="c2 c1">
>   <div role="gridcell" id="c1">cell1</div>
>   <div role="gridcell" id="c2">cell2</div>
> </div>
> </div>
>
> 1. The document order of the cells in the row are c1, c2.
> 2. The aria-owns order is c2, c1.
> 3. Author's shouldn't use aria-owns in this case, since the parent/child
> relationship is expressed in the DOM.
>
> Nonetheless, what should the browser do when it encounters this
> situation?  Respect the aria-owns ordering or the DOM ordering?  I don't
> see a compelling reason for either approach.  Perhaps the spec should
> simply stipulate which ordering wins.
>
> However, what if the author uses table markup?
>
> <table role="grid">
> <tr aria-owns="c2 c1">
>   <td role="gridcell" id="c1">cell1</td>
>   <td role="gridcell" id="c2">cell2</td>
> </tr>
> </table>
>
> It's more of an issue here since there is an inconsistency between the
> conceptual tabular organization and the one exposed through accessibility
> APIs.  I wonder if there are use cases where authors cannot help expressing
> the child order one way in the DOM, but intend a different semantic
> ordering.
>
> [1] http://w3c.github.io/aria/aria/aria.html#aria-owns
>
> --
> ;;;;joseph.
>
> 'Array(16).join("wat" - 1) + " Batman!"'
>            - G. Bernhardt -
>
>
>