W3C home > Mailing lists > Public > public-html@w3.org > February 2012

Using @aria-describedby for long described image links [Was: Using an image map for long described image links [Was: Revert Request]]

From: Benjamin Hawkes-Lewis <bhawkeslewis@googlemail.com>
Date: Thu, 2 Feb 2012 10:33:05 +0000
Message-ID: <CAEhSh3frjgQzeOUPfZGMZQ=ftFCMLreCiT9xWhi02vL2D6iUVQ@mail.gmail.com>
To: John Foliot <john@foliot.ca>
Cc: Matthew Turvey <mcturvey@gmail.com>, Leif Halvard Silli <xn--mlform-iua@xn--mlform-iua.no>, Silvia Pfeiffer <silviapfeiffer1@gmail.com>, Laura Carlson <laura.lee.carlson@gmail.com>, Sam Ruby <rubys@intertwingly.net>, Paul Cotton <Paul.Cotton@microsoft.com>, Maciej Stachowiak <mjs@apple.com>, HTML WG <public-html@w3.org>
[This conversation has derailed from discussing image maps]

On Thu, Feb 2, 2012 at 4:44 AM, John Foliot <john@foliot.ca> wrote:
> 2) You couldn't just park it off screen somewhere, or 'hide' if with
> @hidden, and then link to it with aria-describedby because:
>
>        a) You would lose both the hyperlink to the actual speech, as well
> as the semantic markup of <abbr>, as both would be flattened to string text.
>
>        b) If somehow you could overcome the flattening-to-string-text
> problem, to activate the hyperlink you must put tab-focus on the link - how
> do you focus on something that is hidden? And what of sighted users,
> (perhaps using a tool such as ZoomText Magnifier/Reader, which is both a
> screen reader and screen magnifier -
> http://www.aisquared.com/zoomtext/more/zoomtext_magnifier_reader/) who would
> hit the tab key and not see a visible tab focus (failing WCAG 2.4.7 Focus
> Visible: Any keyboard operable user interface has a mode of operation where
> the keyboard focus indicator is visible. (Level AA)) - what exactly should
> the magnifier magnify?

It's a good question, but again we need to ask the editors of ARIA
this question as they are the ones who are saying UAs may and AT
should provide a way to navigate to navigate to even hidden structured
information.

    http://www.w3.org/WAI/PF/aria-implementation/#mapping_additional_relations_reverse_relations

    http://www.w3.org/WAI/PF/aria-implementation/#include_elements

One possible UI solution to this would be to generate some sort of
popup display rendering the hidden content. You could even mimic
precisely how JAWS uses @longdesc by rendering the hidden content into
a data URI and then asking the browser to open the data URI in a new
window. But it might be get a better UI by generating a lighter weight
popup (think the popup displays used by Apple VoiceOver) based on the
accessibility tree itself.

>        c) The aria-describedby attribute (as well as its companion
> aria-labelledby and aria-label attributes) are read as the Accessible Label
> in the Accessibility API's - those APIs do not recognize the fact that this
> is additional and supplemental information that should be served to the user
> on demand - they instead supply their labeling/describing text as part of
> the regular speech flow.

This doesn't sound right to me? In the suggested API mappings,
@aria-labelledby and @aria-label are mapped to accessible name and
(where available) labelled by relations but @aria-describedby is
mapped to accessible description and (where available) described by
relations:

    http://www.w3.org/WAI/PF/aria-implementation/#mapping_role_table

APIs do not generate "speech flow". AT is free to use the semantics in
the accessibility graph however they like, and additionally to query
the DOM for more information. It may be (I haven't see any tests
around this) that all text-to-speech ATs currently read the
description property automatically, but that is not a limitation of
the information available to AT. For example, where APIs support
relations (iAccessible2, UI Automation, AT-SPI), AT can locate the
description nodes in the accessibility tree. Where APIs that do not
support relations (legacy MSAA and Apple Accessibility API), AT could
query the DOM to find the referenced description elements and map them
back to the accessibility tree. Having located the description nodes,
rather than automatically reading the description property, AT could
give the user the option of moving focus to those nodes.

Querying the DOM is of course more cumbersome than using relations,
but I believe this how JAWS supports @longdesc in IE.

Gecko appears to provide special methods for opening long descriptions
in nsHTMLImageAccessible:

    http://mxr.mozilla.org/mozilla-central/source/accessible/src/html/nsHTMLImageAccessible.h

    http://mxr.mozilla.org/mozilla-central/source/accessible/src/html/nsHTMLImageAccessible.cpp

I'm not quite sure what these actions look like in the various
accessibility APIs, but there's no fundamental reason that opening a
data URI or some other sort of popup pulling together the elements
referenced by @aria-describedby could not be exposed by the same
mechanism. It would look the same to legacy AT.

--
Benjamin Hawkes-Lewis
Received on Thursday, 2 February 2012 10:33:33 UTC

This archive was generated by hypermail 2.3.1 : Monday, 29 September 2014 09:39:29 UTC