- From: Richard Schwerdtfeger <schwer@us.ibm.com>
- Date: Mon, 22 Feb 2010 12:32:57 -0600
- To: David Singer <singer@apple.com>
- Cc: Charles McCathieNevile <chaals@opera.com>, cooper@w3.org, cyns@exchange.microsoft.com, David Bolter <david.bolter@gmail.com>, Steven Faulkner <faulkner.steve@gmail.com>, Frank Olivier <franko@microsoft.com>, janina@rednote.net, James Craig <jcraig@apple.com>, "public-canvas-api@w3.org" <public-canvas-api@w3.org>, public-canvas-api-request@w3.org, surkov.alexander@gmail.com
- Message-ID: <OF9C9C2D29.900EA030-ON862576D2.006377F2-862576D2.0065E519@us.ibm.com>
Hi David,
It can't always be true. The reason is that if <canvas> is the actual
rendering then we cannot guarantee that the <canvas> sub-tree (fallback or
otherwise) is ever an accessibility tree representation of what is drawn on
the <canvas>. If it is not, and it is simply fallback content for browsers
that do not support canvas, the content is not an accessibility
representation for what is drawn on <canvas> and asking:
asking a test tool to test it for accessibility would have no relevance
to the UI being rendered. This would be the case today.
asking the browser to map the sub-tree to the accessibility API would be
inaccurate as it would not correctly represent the UI rendering
To illustrate this, imagine using Bespin where the fallback sub-tree is an
Iframe with an entirely different editor, say one that looks exactly like
Smultron, emacs or Visual Slick Edit but rendered in HTML. The controls
would be different, etc. Keyboard navigation alone would not follow what is
being drawn on the canvas.
It could also be the case that the sub-tree is meant for a user selectable
alternative rendering in the web application. In which case it does not
have a correlation to the visual rendering and should not be mapped or
tested. If however the user selected the alternative rendering the canvas
could be hidden and the alternative content would be rendered and in
neither case should adom be set to true.
If the flag were indeed true by default then that would restrict the author
as to the intent of the sub-tree content which is inconsistent with how it
is being used today. This would result in an accessibility API mapping and
an accessibility test scenario that is inconsistent with what is being
rendered on the visual canvas.
Consequently, having adom set to false by default would ensure that the
author establish the intent of the canvas sub-tree for accessibility test
tools and ATs at the time it is rendered.
Does that help?
Rich
Rich Schwerdtfeger
Distinguished Engineer, SWG Accessibility Architect/Strategist
David Singer
<singer@apple.com
> To
Richard
02/22/2010 10:53 Schwerdtfeger/Austin/IBM@IBMUS
AM cc
James Craig <jcraig@apple.com>,
Charles McCathieNevile
<chaals@opera.com>, cooper@w3.org,
cyns@exchange.microsoft.com, David
Bolter <david.bolter@gmail.com>,
Steven Faulkner
<faulkner.steve@gmail.com>, Frank
Olivier <franko@microsoft.com>,
janina@rednote.net,
"public-canvas-api@w3.org"
<public-canvas-api@w3.org>,
public-canvas-api-request@w3.org,
surkov.alexander@gmail.com
Subject
Re: Agenda: HTML 5 Canvas
Accessibility Meeting February 22,
2010 (resending)
Richard
I think you may be misunderstanding the position. I agree, today, the
fallback content is discarded by browsers that understand canvas.
We discussed the flag to indicate whether the fallback content should be
used as the (initial) accessibility content, and we also discussed what the
default for the flag should be. Someone suggested perhaps it should be
'true' by default (unlike today, where it is effectively false).
Now one asks: is it ever worth setting it to 'false'? In some cases, it
can save some work on the part of a script; consider, for example, the
case where the fallback content merely says "get a better browser", but the
accessibility content is a more complex structure that gives (for example)
alternative input method support. In this case, building the accessibility
DOM tree means you have to tear down that "get a better browser" sub-tree,
but that is easy.
So, if we make the flag true by default, is it needed at all? Can it be
always true?
On Feb 21, 2010, at 11:14 , Richard Schwerdtfeger wrote:
HTML <canvas> content should be accessible too. However, I can think
of none on the web that are today. Your statement of "partially
inaccessible" makes no sense either technically or logically based on
what has been implemented for <canvas> to date.
The purpose of the adom attribute is not an accessibility compliance
statement "flag." It is an indication to the browser to map what is
in the subtree to the accessibility infrastructure and to include it
in the keyboard navigation tree. The subtree, today, is not designed
to do that. It is basically ignored a browser unless it does not
support canvas. Setting the flag is not a guarantee that the author
has done a good job any more than anything else that is out there
today. That is what accessibility test tools are for.
The reality is that you should not map the subtree DOM to the
accessibility API and include it in the navigation order, by default,
because neither the AT, an accessibility test tool, nor the user have
any way of knowing the purpose of that content. As for Ian's content
about the content being non-conforming there is no restrictions on
what the fallback content is today. It can be anything. Cynthia made
a very important point, at one of the previous meetings, in that an
accessibility test tool need to know the content in the sub-tree is
directly related to the visual rendering to test it. That cannot be
known unless the author indicates it is.
There are authors that do not care about meeting accessibility
criteria and our making a blind assumption that the author should
care to do so would be irresponsible on our part. Without the
attribute you would have to REQUIRE that the author make the
structure and accessibility properties exactly match what you are
rendering on the canvas in all instances. That is an unrealistic
expectation of authors and creates an unmanageable situation for
accessibility test tools. We would also create a situation where two
people sitting down with a web page application (one sighted and one
not) will try to operate a web page application and the solution the
blind user has access to will have absolutely no correlation to the
one being rendered on canvas keyboard-wise, semantically, etc. It may
be an entirely different collection of components, which the neither
user cansee and behaves nothing like the visual rendering.
I am sorry James, but the argument you present holds no water to me
and I see no way of us ever reaching consensus on it.
Rich
Rich Schwerdtfeger
Distinguished Engineer, SWG Accessibility Architect/Strategist
<27158101.gif>James Craig ---02/19/2010 08:17:59 PM---I discussed
agenda item #2 (@adom) today with Maciej and David, and I've come to
agree with Ian's original argument that canvas
James Craig <
jcraig@apple.com>
Sent by:
public-canvas-api-reques <27027431.gif>
t@w3.org To
<27027431.gif>
Richard
02/19/2010 08:17 PM Schwerdtfeger/Austin/I
BM@IBMUS
<27027431.gif>
cc
<27027431.gif>
David Bolter <
david.bolter@gmail.com
>, cooper@w3.org,
janina@rednote.net,
Charles McCathieNevile
<?chaals@opera.com>,
cyns@exchange.microsof
t.com, Steven Faulkner
<
faulkner.steve@gmail.c
om>, Frank Olivier <?
franko@microsoft.com>,
"?
public-canvas-api@w3.o
rg" <
public-canvas-api@w3.o
rg>,
surkov.alexander@gmail
.com
<27027431.gif>
Subject
<27027431.gif>
Re: Agenda: HTML 5
Canvas Accessibility
Meeting February 22,
2010
<27027431.gif>
<27027431.gif>
I discussed agenda item #2 (@adom) today with Maciej and David, and
I've come to agree with Ian's original argument that canvas contents
should just be accessible by default. Though the idea of using an
attribute was slightly more palatable than an extra element, adding a
flag of any kind doesn't provide much benefit in the best case
scenario. In the worst case scenario, it would render partially
accessible content completely inaccessible.
James
On Feb 19, 2010, at 5:23 PM, Richard Schwerdtfeger wrote:
Monday, 2010-2-15
Time: 3:00pm-4:00pm Boston local
Name: WAI_PFWG(CANVAS)
Code: 92473 ("WAIPF")
One time
irc channel= #html-a11y
Agenda:
1. Identify Scribe
2. Final Review for spec. ready adom text for
Issue 74
http://lists.w3.org/Archives/Public/public-canvas-api/2010JanMar/0178.html
3. Progress on caret tracking:
http://www.w3.org/WAI/PF/HTML/track/actions/19
- Steve Faukner
Rich Schwerdtfeger
Distinguished Engineer, SWG Accessibility
Architect/Strategist
David Singer
Multimedia and Software Standards, Apple Inc.
Attachments
- image/gif attachment: graycol.gif
- image/gif attachment: pic30023.gif
- image/gif attachment: ecblank.gif
Received on Monday, 22 February 2010 18:33:45 UTC