ACTION-1311: Clarify that presentation role should preserve separation of nodes within tables/lists; work out in UAIG 1.1 and with implementors/vendors how to resolve this.

Email from Alex (to: Matt King, cc: Brett Lewis, David Bolter, Richard Schwerdtfeger, Joseph Scheuhammer):

Hi, Matt.

Layout table is presentational table, it provides table semantics to AT but it's not exposed to the user.

UAIG says to not create an accessible object for table element, so exposing role="presentation" table as layout table is not possible without breaking ARIA spec.

Also UAIG doesn't say we must create an accessible for TBODY/TR/TD/etc in this case. So Firefox implementation conform the spec. However having a generic accessible for presentational TD/TH doesn't seem breaking ARIA spec and it may resolve the problem.

Anyway, before making any attempts to fix the problem on Firefox side I'd like to get some discussion in ARIA group.

Thanks.
Alex.



On Fri, Nov 22, 2013 at 3:33 AM, Matthew King <mattking@us.ibm.com> wrote:

Alex,

Isn't a layout table a "invisible table?" It is for "presentation" only. It does not have "table semantics". So, removing table semantics is the same as calling a table a layout table. This is WG interpretation of the language and the way WG wishes UA developers to interpret.

When the WG wrote the spec, removing table semantics was not intended to remove all structure.

So, when the UA implements role presentation on tables and lists, there is still a requirement to maintain some form of structure that separates elements of the layout table.

Perhaps the UAIG could specify what the mapping of presentational elements should be fore each element and for each API. Do you feel that is necessary?

Will you please work to correct the current Firefox behavior and implement some mapping of presentational TD and TH that maintains a structure that AT can use? Right now, we lose all structure, making pages unreadable.

Matt King
IBM Senior Technical Staff Member
I/T Chief Accessibility Strategist
IBM BT/CIO - Global Workforce and Web Process Enablement
Phone: (503) 578-2329, Tie line: 731-7398
mattking@us.ibm.com



From: Alexander Surkov <surkov.alexander@gmail.com>
To: Matthew King/Fishkill/IBM@IBMUS,
Cc: Brett Lewis <blewis@freedomscientific.com>, David Bolter <dbolter@mozilla.com>, Richard Schwerdtfeger/Austin/IBM@IBMUS
Date: 11/21/2013 05:23 AM
Subject: Re: role="presentatinon"



I'd say ARIA spec shouldn't be so strict requiring the browser to remove native semantic of table and its descendants, it should say instead that presentational table should be a layout table. That'd be much easier for implementers. The same time if we exposed TD as a section role then it should be ok with ARIA spec but that would complicate already complicated browser implementation. So I'd suggested to rise this topic at WG.

Thanks.
Alex.


On Wed, Nov 20, 2013 at 7:59 PM, Matthew King <mattking@us.ibm.com> wrote:
Brett, Alex,

Today, JAWS, and other screen readers, have algorythms for identifying some tables as layout tables. In those cases, JAWS leaves the table semantics, or table structure, out of the JAWS virtual buffer. However, in those cases, JAWS still treats every cell in the table as a separate element and includes a line break in the virtual buffer between cells.

The result of applying role="presentation" to a table, or list, should be exactly the same. In other words, role="presentation" is meant to take the guess work out of identifying layout tables. ARIA now gives us a standards-based way of coding a layout table, or layout list, etc. That is the intent, and I thought the language of the spec made that pretty clear. If it does not, then perhaps we should entertain additional or alternate language.

Does that help you determine the best IA2 mapping for TD and TH in a layout table or LI in a layout list?

Matt King
IBM Senior Technical Staff Member
I/T Chief Accessibility Strategist
IBM BT/CIO - Global Workforce and Web Process Enablement
Phone: (503) 578-2329, Tie line: 731-7398
mattking@us.ibm.com



From: Alexander Surkov <surkov.alexander@gmail.com>
To: Brett Lewis <blewis@freedomscientific.com>,
Cc: Richard Schwerdtfeger/Austin/IBM@IBMUS, David Bolter <dbolter@mozilla.com>, Matthew King/Fishkill/IBM@IBMUS
Date: 11/20/2013 04:09 PM
Subject: Re: role="presentatinon"



Hi, Brett. I see. It makes me feel wrong about role presentation. If it is supposed to remove whole table semantic then that's what we do. If not then layout table should work for that.
Thanks.
Alex.


On Wed, Nov 20, 2013 at 5:48 PM, Brett Lewis <blewis@freedomscientific.com> wrote:
Hi Alex,
Attached is a little html example of a table used for layout purposes with role="presentation".
I definitely understand why Matt is having trouble with these types of tables, but I don't know a way around it
Rich are you proposing that instead of the IAccessibles for the table, rows, and cells just being removed that they be replaced with elements with IA2_ROLE_SECTION?

I personally feel that role="presentation" should just be removed from the table.
Brett


From: Richard Schwerdtfeger [mailto:schwer@us.ibm.com]
Sent: Wednesday, November 20, 2013 2:07 PM
To: Alexander Surkov
Cc: Brett Lewis; David Bolter; Matthew King
Subject: Re: role="presentatinon"
Matt, can you give Alex access to an example. I think it would help if he could see one. I am guessing we send him to greenhouse.



Rich Schwerdtfeger

Inactive hide details for Alexander Surkov ---11/20/2013 03:59:58 PM---Rich, I understand the problem but I'd need a help to fiAlexander Surkov ---11/20/2013 03:59:58 PM---Rich, I understand the problem but I'd need a help to figure out where is it. <span> and <div> are D

From: Alexander Surkov <surkov.alexander@gmail.com>
To: Richard Schwerdtfeger/Austin/IBM@IBMUS
Cc: Brett Lewis <blewis@freedomscientific.com>, David Bolter <dbolter@mozilla.com>, Matthew King/Fishkill/IBM@IBMUS
Date: 11/20/2013 03:59 PM
Subject: Re: role="presentatinon"



Rich, I understand the problem but I'd need a help to figure out where is it. <span> and <div> are DOM elements, role presentation is pure a11y thing, a11y has nothing to do with span or div so span/div terms are descriptive but it doesn't help in fixing. I guess no semantic accessible has accessible role that treated as a span by JAWS. But it'd great to have some details on this.
Alex.


On Wed, Nov 20, 2013 at 4:50 PM, Richard Schwerdtfeger <schwer@us.ibm.com> wrote:
When you remove say a TD you replaced it with a <span> vs. a <div>. Spanned text on the same row gets merged into a single line of text having different attributes in the AT vs. being processed as separate entities.

This is incredibly bad.

Rich


Rich Schwerdtfeger

Inactive hide details for Alexander Surkov ---11/20/2013 03:08:21 PM---A technical description of the regression would be helpfAlexander Surkov ---11/20/2013 03:08:21 PM---A technical description of the regression would be helpful (span/div terms don't make it clear with

From: Alexander Surkov <surkov.alexander@gmail.com>
To: Richard Schwerdtfeger/Austin/IBM@IBMUS
Cc: Brett Lewis <blewis@freedomscientific.com>, Matthew King/Fishkill/IBM@IBMUS, David Bolter <dbolter@mozilla.com>
Date: 11/20/2013 03:08 PM
Subject: Re: role="presentatinon"




A technical description of the regression would be helpful (span/div
terms don't make it clear with me). Brett?
Alex.

On Wed, Nov 20, 2013 at 4:01 PM, Richard Schwerdtfeger
<schwer@us.ibm.com> wrote:
> Hi Alex,
>
> A change went in for, I believe FF 20, that created a real problem when
> role="presentation" was applied to tables. The semantic tags were replaced
> with <span>s vs. <div>s which caused text in different cells to be merged
> together when used by the AT. That was not the intent of role="presentation"
> and it is breaking a lot of IBM web apps for blind users.
>
> Please change it back to <div>s. Matt is having a really difficult time now
> trying to use our products.
>
> Rich
>
> Rich Schwerdtfeger

FF now exposes IA2_ROLE_TEXTFRAME (MSAA+IA2) for the TD's in the presentation table. See:
https://bugzilla.mozilla.org/show_bug.cgi?id=945435

Also related:
https://bugzilla.mozilla.org/show_bug.cgi?id=1013584

Waiting on results of rationalized mappings for role="group" and <div> elements, and their relationship to the generic platform role. See ISSUE-678/ACTION-1515.

See also ACTION-1444.

Also waiting on ACTION-1581 -- panel role.

Edits made for MSAA+IA2.

https://github.com/w3c/aria/commit/ccc811073ae9969033d0dc27650dfd75273dfe13

Closing.