Re: CfC: Approve overview of testing in view of permissive CR exit criteria

The additional time granted to the HTML Accessibility Task Force for reviewing the HTML5 WG's "Overview of Testing in View of CR exit" (Interoperability Status Report) allowed us to conduct more research and testing on the 13 sections that originally caused us concern [1].  Through this testing and research we were able to identify a few areas of the spec that we would like to improve editorially (by filing bugs).  More importantly, we were able to agree that the following sections were not likely to impact accessibility in a way that would require their status be changed from "Interoperable" to "Requires testing."

- 2.5.6 Colors
- 2.5.10 Media Queries
- 3.2.3.2 The title attribute
- 3.2.5.2 Transparent content models
- 4.8.14.2 Processing model
- 4.9.1.1 Techniques for describing tables
- 4.9.2 The caption element
- 4.10.8 The button element
- 4.12 Links
- 4.12.5 Link types
- 4.13.3 Tag clouds

That only leaves us with 2 items for the Working Group to consider:

3.2.3.3 The lang and xml:lang attributes

- There is currently an open bug [2] related to the lang attribute that is considered to be critical on the path to CR [3].  While this bug does not seem to be related to accessibility and progress on closing it is being made, we may wish to reconsider its interoperability status once it is closed.

## 3.2.7.3 Strong Native Semantics [4]

- The spec states that an img element [5] whose alt [6] attribute value is empty must have a role of 'presentation'.  The Task Force conducted testing on nightly versions of Firefox, Chrome and Safari as well as IE10 by inspecting the accessibility API to confirm the role values of images that met this condition.  It was confirmed that the default role of an img whose alt attribute value was empty was not changed to 'presentation' or altered in any way.

We believe this is a critical aspect of accessibility in HTML5 and would like to see steps taken to insure it gets support from major browser vendors.

The spec also states that an element that is inert [7] must have the aria-disabled state set to "true."  The "inert" state was designed in HTML5 to be used by the dialog element (it is proposed as a global attribute in HTML5.1).  There are currently no known implementations of the dialog element or support for the inert attribute.  Because of this, we were not able to confirm or deny with tests the interoperability status of inert subtrees and their aria-disabled state.  Should the implementation status of the dialog element change, we would like to reconsider the interoperability status of this section.

## 3.2.7.4 Implicit ARIA Semantics [8]

As mentioned above, the dialog element [9] is at risk and currently not implemented.  The spec states that the dialog element must have an implied role of dialog.  Should the status of this element change, we would like to reconsider the interoperability status of this section.

The spec states that Headings level 1-6 [10] should have a role of 'heading', with the aria-level property set to the element's outline depth [11]. The Task Force conducted testing on nightly versions of Firefox, Chrome and Safari as well as IE10 by creating Headings that appeared at various levels of nesting in sectioning elements.  We confirmed, using accessibility api inspectors, that the aria-level property remains consistent with the Heading level specified by the element and not adjusted to its outline depth, which should be determined by the elements document nesting level.

The spec also states that the `<output>` element must have a role of 'status'.  Again, using accessibility API inspectors, we confirmed that the `<output>` element has a role of 'text', or 'editable text'.

Implicit ARIA semantics play an important role in accessibility and we believe that while parts of this section are currently implemented by major browser vendors, the examples of sections not currently implemented that are outlined above warrant a reconsideration of the status of this section.

Kind regards,

Mark Sadecki
On behalf of the HTML Accessibility Task Force

[1] http://lists.w3.org/Archives/Public/public-html-admin/2013Jul/0027.html
[2] https://www.w3.org/Bugs/Public/show_bug.cgi?id=18490
[3] http://lists.w3.org/Archives/Public/public-html-admin/2013Jun/0028.html
[4] http://www.w3.org/TR/html5/dom.html#sec-strong-native-semantics
[5] http://www.w3.org/TR/html5/embedded-content-0.html#the-img-element
[6] http://www.w3.org/TR/html5/embedded-content-0.html#attr-img-alt
[7] http://www.w3.org/TR/html5/editing.html#inert-subtrees
[8] http://www.w3.org/TR/html5/dom.html#sec-implicit-aria-semantics
[9] http://www.w3.org/TR/html5/interactive-elements.html#the-dialog-element
[10] http://www.w3.org/TR/html5/sections.html#the-h1,-h2,-h3,-h4,-h5,-and-h6-elements
[11] http://www.w3.org/TR/html5/sections.html#outline-depth
[12] http://www.w3.org/TR/html5/forms.html#the-output-element

On 9/3/13 4:01 PM, Mark Sadecki wrote:
> The following update, from the HTML Accessibility Task Force, was originally sent to the wrong mailing list (team-html-admin) on 8/16.  I was unaware of this mistake until my return from summer holiday and am now resending it to the proper list.
>
> My sincerest apologies,
>
> Mark Sadecki
>
> Original text:
>
> Please accept the summary below as a demonstration of our progress in finding evidence to support the report of interoperability issues found in our original [1] reply to the CfC: Overview of testing in view of CR exit[2].  A number of issues were either determined to be out of the scope of this CfC, or lacked evidence supporting the claim.  These are marked as having been "Withdrawn."  Those items for which we are currently developing tests are in an undetermined state.   We will notify this list as soon as a determination is made on those items.
>
> On 7/15/13 11:47 PM, Mark Sadecki wrote:
>> On 6/13/13 9:56 PM, Paul Cotton wrote:
>>>
>>> This is a Call for Consensus (CfC) to approve the parts of the following document that indicate which parts of the HTML5 specification can be considered interoperable as per the permissive CR exit criteria [1].  The items subject to this CfC are marked "Considered interoperable" and color coded in green in the document:
>>>
>>> http://www.w3.org/html/wg/tests-cr-exit.html
>>>
>>> Silence will be taken to mean there is no objection, but positive responses are encouraged. Objections should be made to specific entries in the document.  Objections of the form "features in 2.8.2.1 HTMLAllCollection are not currently interoperable" MUST be accompanied with specific evidence of non-interoperability, otherwise such objections will not be accepted by the Chairs.
>>>
>> The HTML Accessibility Task Force (TF) has prepared the following response based on analysis [1] of items marked as "Interoperable" in the HTML WG document "Overview of testing in view of CR exit" [2] to the extent possible in the time allotted for review .  It is the Task Force's belief that the items below should not be considered "Interoperable", exempting them from the formal testing effort required for the HTML5 specification to exit Candidate Recommendation [3].  We ask that the status of sections outlined below be changed from "Interoperable" to "Requires Testing".
>>
>> 2.5.6 Colors
>> Color is a critical component of accessibility.  While the parsing of color strings may not be a specific concern for accessibility, there was much discussion about color and the long standing dissatisfaction the Task Force and the Protocols and Formats Working group have with the deprecation of System Colors (see bug 13639 [4]).
> - Withdrawn
>>
>> 2.5.10 Media Queries
>> The TF anticipates Media Queries to be used for accessibility purposes.  Since media queries are a feature new to HTML5, we believe they should be tested to verify that they work as expected across different platforms.
> - Withdrawn
>>
>> 3.2.3.2 The title attribute
>> The title attribute, while widely implemented, is not implemented in a consistent or accessible manner.  The HTML spec acknowledges this in a Note [5] contained at the beginning of the section.  The TF believes this attribute requires interoperability testing.
> - Withdrawn
>>
>> 3.2.3.3 The lang and xml:lang attributes
>> Screen readers can use the lang attribute to perform language switching, allowing speech in the natural language specified. Since this attribute can be used on any element, the TF believes testing is required to insure its value is available to user agents in all contexts.
> - Attempts were made to find evidence of interoperability issues.
> - It does not appear that Accessibility API's provide access to the value of this attribute.
> - Current support by Assistive Technology is presumably through the DOM (JAWS and NVDA support natural language switching)
> - There are no valid, open bugs related to lang or xml:lang
> - Status update Email to TF <http://lists.w3.org/Archives/Public/public-html-a11y/2013Aug/0050.html>
>>
>> 3.2.5.2 Transparent content models
>> The "transparency" of an element depends on its context or usage. Most fallback content can be considered "transparent" as it inherits the context of its parent element.  The "re-mapping" of an element's content model must be properly conveyed in the DOM to insure that Assistive Technology (AT) can properly modify the element's behavior and provide the appropriate access to it.  As far as we are aware, there have not been any tests designed to verify this.
> - Withdrawn
>>
>> 3.2.7.3 Strong Native Semantics
>> and
>> 3.2.7.4 Implicit ARIA Semantics
>> The TF believes at least some of the implementation requirements in these sections are
>> either not interoperably implemented or require testing to demonstrate that they are.  For example:
>> * The h1-h6 elements are defined as requiring to have the aria- level property set to the element's outline depth [6].  The TF is unaware of appropriate implementation of this in any browser.
>> * The dialog element requires mapping to a dialog role [7]. The TF is unaware of appropriate implementation of this in any browser.
>> * The img element whose alt attribute's value is empty should have a default role=presentation [8] (removing it from the accessibility tree). The TF is unaware of appropriate implementation of this in any browser.
>> *NOTE: the current state of these sections has already been changed to "Needs testing" by the document's editor (Robin Berjon) based on discussions with Mark Sadecki on behalf of the TF*
> - Currently developing tests
>>
>> 4.8.14.2 Processing model
>> The TF has an issue with the spec text that makes an exception for how `<object>` elements are to be processed by user agents which do not support images [9].  The TF would like to see this issue resolved and the behavior tested in user agents as it affects the accessibility of imagemaps contained in `<object>` elements.
> - Withdrawn
>>
>> 4.9.1.1 Techniques for describing tables
>> The TF identified poor support for the techniques described in this section [10] by Assistive Technology.
> - Withdrawn
>>
>> 4.9.2 The caption element
>> The TF expressed concern about the removal of the summary attribute and its replacement, the caption element. Implementations of the caption element are buggy (i.e. some mobile browsers will cut off captions that are longer than the width of the containing table.) We believe this new element requires testing.
> - Currently developing tests
>>
>> 4.10.8 The button element
>> The TF believes that while the button element has existed for quite some time and is widely considered to be interoperable, there are many new attributes available on this element that may have an impact on accessibility (autofocus, formnovalidate, etc.) and have not yet been individually tested.
>> *NOTE: the current state of this section has already been changed to "Needs testing" by the document's editor (Robin Berjon) based on discussions with Mark Sadecki*
> - Currently developing tests
>>
>> 4.12 Links
>> Given that links are a critical component of the web and that the ability to navigate to them with the keyboard in all appropriate contexts is critical to users of AT (i.e. when they are used in fallback content) the TF believes that the keyboard accessibility of links be tested, especially when used in the context of fallback content.
> - Currently developing tests
>>
>> 4.12.5 Link types
>> While link types have been considered widely interoperable, HTML5 now supports the the use of of link type on the `<a>` and `<area>` elements in addition to the `<link>` element. The TF believes testing is required to determine interoperability of this new feature of HTML5.
> - Currently developing tests
>>
>> 4.13.3 Tag clouds
>> The TF believes that testing is required of the example given to markup a tag cloud. The TF does not believe the `font-size` property is interoperable [11].  This is supported by the variety of "CSS Reset" methods available to developers [12], particularly if used in conjunction with a user override of min-font-size.
> - Withdrawn
>
> [1] http://lists.w3.org/Archives/Public/public-html-admin/2013Jul/0027.html
> [2] http://www.w3.org/html/wg/tests-cr-exit.html
>
> Mark Sadecki
> On behalf of the HTML Accessibility Task Force
>>
>> Kind Regards,
>>
>> Mark Sadecki
>> On behalf of the HTML Accessibility Task Force
>>
>> [1] http://www.w3.org/WAI/PF/HTML/wiki/Results_of_Spec_Review
>> [2] http://www.w3.org/html/wg/tests-cr-exit.html
>> [3] http://dev.w3.org/html5/decision-policy/public-permissive-exit-criteria.html
>> [4] https://www.w3.org/Bugs/Public/show_bug.cgi?id=13639
>> [5] http://www.w3.org/TR/html5/dom.html#the-title-attribute
>> [6] http://www.w3.org/html/wg/drafts/html/master/sections.html#outline-depth
>> [7] http://www.w3.org/html/wg/drafts/html/master/interactive-elements.html#the-dialog-element
>> [8] http://www.w3.org/html/wg/drafts/html/master/embedded-content-0.html#the-img-element
>> [9] https://www.w3.org/Bugs/Public/show_bug.cgi?id=13451
>> [10] http://www.davidmacd.com/test/details.html
>> [11] http://css-tricks.com/font-rendering-differences-firefox-vs-ie-vs-safari/
>> [12] http://www.cssreset.com/
>>>
>>> If there are no objections by Monday July 15, this resolution will carry.
>>>
>>> /paulc
>>>
>>> HTML WG co-chair
>>>
>>> [1] http://dev.w3.org/html5/decision-policy/public-permissive-exit-criteria.html
>>>
>>> Paul Cotton, Microsoft Canada
>>>
>>> 17 Eleanor Drive, Ottawa, Ontario K2E 6A3
>>>
>>> Tel: (425) 705-9596 Fax: (425) 936-7329
>>>
>>
>>
>
>


-- 
Mark Sadecki
Web Accessibility Engineer
World Wide Web Consortium, Web Accessibility Initiative
Telephone: +1.617.715.4017
Email: mark@w3.org
Web: http://w3.org/People/mark

Received on Thursday, 5 September 2013 00:25:42 UTC