W3C home > Mailing lists > Public > public-html@w3.org > April 2008

Re: One more thought about requiring the alt to add to the pile

From: Alfonso Martínez de Lizarrondo <amla70@gmail.com>
Date: Sun, 20 Apr 2008 11:57:35 +0200
Message-ID: <af2a8eab0804200257v3e8bac72wd5abea5c0e34783a@mail.gmail.com>
To: "Steven Faulkner" <faulkner.steve@gmail.com>
Cc: "Henri Sivonen" <hsivonen@iki.fi>, "public-html@w3.org WG" <public-html@w3.org>, "W3C WAI-XTECH" <wai-xtech@w3.org>, wai-liaison@w3.org

Wouldn't it be possible to prepare a set of pages, "randomly" taken
from the web, and then adjust them so there is at least on version of
each one of them in order to test the different hypothesis about what
is better in terms of accessibility?

Of course we all know that the best results are when all the important
images have a properly written alt, and even better if they have a
longdesc, decorative images have alt="" and there are no images
without alt.

But the reality is that in many pages that won't be possible: the
author doesn't care about validation, or uses a software that
automatically adds alt="" or the image has automatically uploaded and
there's no change that a proper alt describing the image can be
written.

So the test would be to create a modified version of each page where a
problem has been found (missing alt or alt wrongly set to ""), and
then narrate them to some tests subjects, simulating a perfect
behavior of a future screen reader. I mean a future because I don't
think that it is logical to be bounded by the current shortcomings of
the technology, but instead try to find out what is better, both from
the HTML side and the AT side, so instead of keep on guessing about
what could be the best solution, find out with real persons what they
find it's better for them to understand the content of those pages.

The steps would be:
1. Select a number (small) of pages to be tested.
2. Choose the hypothesis that we want to test (omited alt, setting
noalt, alt with magic values...)
3. Create a modified version of each one of the test pages.
4. Record a narration of each of the pages (so they can be tested
without a human narrator, this is of course optional)
5. Pick some people that are currently using some AT and ask them what
version of each page they do prefer (I repeat, the perfect version
with properly alt for each image won't be available in the test, we
want to test what happens when that info is missing)
6 Study the results instead of keep on theorizing about the perfect behavior.

Regards


2008/4/20, Steven Faulkner <faulkner.steve@gmail.com>:
>
>  hi henri,
>
>
>  >  There has now been a decade-long experiment with making alt a syntax
>  > requirement. I think this experiment shows that doing so has the downside of
>  > inducing bogus alt. When validation has downsides, as a validator developer,
>  > I want to work to remove the downsides.
>
>
> Where is the empirical data to support your assumptions? All we
>  currently have on both sides is anecdote and conviction.
>
>
>  >  A modal non-visual UI may be easier to invent, but in a non-visual UI it is
>  > also harder to tell what mode you are trapped in, so I don't think it is
>  > necessarily OK to introduce more modality even if there already is some. It
>  > appears that VoiceOver tries to avoid modality (apart from the VO key lock)
>  > just like visual Apple user interfaces.
>
>
> As far as AT is concerned we are not comparing like with like.
>  the windows and mac OS systems present different challenges for AT to overcome.
>  I do  not understand the reason for presence of modal UIs in windows
>  AT, but not in voiceover.
>
>  my suggestion was not to add another mode as both window eys and jaws
>  (for example) have the option available already to announce all
>  graphics, but currently this setting still ignores <img alt="">,
>  so what I am suggesting is that the vendors merely change the
>  functionality for this option so that alt="" is reported.
>
>  regards
>  stevef
>
> On 20/04/2008, Henri Sivonen <hsivonen@iki.fi> wrote:
>  > On Apr 18, 2008, at 11:07, Steven Faulkner wrote:
>  >
>  >
>  > >
>  > > > No. Now you are being so dogmatic about the alt attribute being there
>  > > > that you are willing to suggest modal UI to work around it. That's
>  > > > bad.
>  > > >
>  > > >
>  > > There is dogma on both sides of the debate, you  appear more dogmatic
>  > > about the idea of alt as optional, than I am to it being required, I
>  > > have publically stated that I am as yet unconvinced of the
>  > > desirability of a required alt. There is obviuosly no doubt in your
>  > > mind.
>  > >
>  >
>  >  I can be persuaded with empirical data.
>  >
>  >  There has now been a decade-long experiment with making alt a syntax
>  > requirement. I think this experiment shows that doing so has the downside of
>  > inducing bogus alt. When validation has downsides, as a validator developer,
>  > I want to work to remove the downsides.
>  >
>  >  It may be that there's a greater upside and that a situation that polarizes
>  > results but has a greater upside is better even if it also moves the
>  > downside further from neutrality. However, absent data about this, I think
>  > it is reasonable to default to removing the downside.
>  >
>  >  Also, I think the Image Review feature I have implemented in Validator.nu
>  > works better than merely flagging missing alt as a validation error would
>  > for validator users who want to maximize an accessibility measure. It
>  > remains to be seen how it affects validator users who don't care about an
>  > accessibility measure and are seeking to maximize a syntactic correctness
>  > measure.
>  >
>  >  In general, if you want people to maximize function f(), it is safer to
>  > tell them to do so than to tell them to maximize a more appealing function
>  > g() and then try to build an artificial correlation between the two. Because
>  > then people are really maximizing g() and if your artificial correlation
>  > setup isn't working, well, oops. So if your agenda is accessibility, the
>  > advocacy should be "accessibility, accessibility"--not "validity, validity"
>  > with an added attempt to tie them together.
>  >
>  >
>  > > There already are  "modal UI's" for most aspects of screen readers
>  > > content presentation, i think it is the nature of  presenting visual
>  > > UI's non visually or non linear content linearly.
>  > >
>  >
>  >  A modal non-visual UI may be easier to invent, but in a non-visual UI it is
>  > also harder to tell what mode you are trapped in, so I don't think it is
>  > necessarily OK to introduce more modality even if there already is some. It
>  > appears that VoiceOver tries to avoid modality (apart from the VO key lock)
>  > just like visual Apple user interfaces.
>  >
>  >  --
>  >  Henri Sivonen
>  >  hsivonen@iki.fi
>  >  http://hsivonen.iki.fi/
>  >
>  >
>  >
>
>
>
> --
>  with regards
>
>  Steve Faulkner
>  Technical Director - TPG Europe
>  Director - Web Accessibility Tools Consortium
>
>  www.paciellogroup.com | www.wat-c.org
>  Web Accessibility Toolbar -
>  http://www.paciellogroup.com/resources/wat-ie-about.html
>
>
Received on Sunday, 20 April 2008 09:58:14 GMT

This archive was generated by hypermail 2.2.0+W3C-0.50 : Wednesday, 9 May 2012 00:16:16 GMT