Re: Dumb Thought on alt Text (or Smart Thought?)

Well, I think the PS is pretty smart. And I think that if HTML was different
(say, was SVG, or what me might like to have in XHTML 2, or some other
language) your idea might be applicable. But the semantics of HTML and in
particular of the relationship between img element and alt attribute (and
coincidentally to some extent between the object element and its content)
lead me to conclude that your fundamental idea of not using alt for anything
except text content is wrong.

This basically feeds from the ideas I ahve tried to express about the
difference between "all content of a document" and "the information a
document is trying to convey", and why I think those are important concepts
for us to be able to use and identify, in the xtech thread on definition of
content (nearly all of the posts in
http://lists.w3.org/Archives/Public/wai-xtech/2002Jan/thread are part of this
thread).

There is nothing in the HTML specification that suggests that an img combined
with an alt and
perhaps a longdesc and/or title is insufficient for conveying information,
and that the information in that collection of content must be reproduced in
any other form. I amm not sure that the specification is correct about this,
but I don't think you have established that it is wrong, either. I think in a
number of cases the use of a graphic identifier (which is the only kind of
text I think is reasonable to leave in images, and precisely because I think
it is a graphic device that may be related to the shapes of letter in the way
that my singing may be related to music) and a text equivalent probably is
sufficient, and in other cases adding a full equivalent via a longdesc
reference means there is enough there.

cheers

Chaals

On Sat, 2 Feb 2002, Kynn Bartlett wrote:

  A dumb thought which I'll share here first instead of sharing on IG:

        Maybe alt text should _only_ be used for images which are
        text-as-graphics.  Everything else gets null alt text.

  Okay, expanding on that a little more. The purpose of alt is to be
  a textual equivalent/substitution/alternative for an image. The only
  textual equivalent for an image is, literally, the text on the image.

  So if an image is my name in a fancy font, the alt text is "Kynn".
  If it is a picture of me, the alt text is "" (quote quote).  Why?
  Because there's no text in there for the alt to be the equivalent
  of.

  Let's look at a case raised on the IG list earlier this week:

  At 9:59 AM -0800 1/31/02, Charles F. Munat wrote:
  >Alt attributes are NOT FOR DESCRIPTIONS.
  >"Ducks on the lake on a warm summer day" is a DESCRIPTION of the image.
  >This makes NO SENSE as alt text.
  >The alt attribute text should flow with the rest of the text. After all,
  >that is how it will be read. [...]
  >Now, suppose that this lake with ducks is right next to Big Hall, and
  >that the function of the image is to give visual users a sense of
  >what life is like in Big Hall. Then an appropriate alt text might
  >read: "Big Hall overlooks Campus Lake and its duck denizens."
  >Here is how this flows:
  >"... There are four residence halls on campus. The biggest is Big Hall,
  >located at the northernmost point on campus. Big Hall overlooks Campus
  >Lake and its duck denizens. The next biggest is Middling Hall..."

  I agree with Mr. Munat's statement that alt attributes are not for
  descriptions. I disagree with his next suggestion, though, of hiding
  image interpretation within the alt text.

  New term: image interpretation. We have "text alternative" which I am
  redefining here to be "the text contained with the graphic"; we have
  "description" which is "what the picture looks like"; and now we have
  "image interpretation" which is "what the image is meant to convey."

  It is my suggestion that "image interpretation" is as appropriate for
  alt attributes as descriptions, which is to say NOT AT ALL.

  Image interpretation should be part of the visible, visually accessible
  text of any context in which the image appears. In other words, the
  page's visible text content should contain the intended interpretation
  of the hall and the ducks, which is the point of the image.

  How is this encoded? It should either clearly appear within the page's
  textual content in context, or should be associated via markup. There
  are few markup methods for doing this reliably; one method which is
  barely acceptable is a one-cell data table with a <caption>.  The lack
  of clear way to associate interpretive and/or descriptive text with
  an <img> is an obvious shortcoming of HTML; this should be passed along
  via PF to the HTML Working Group (XLink may help here).

  Anyway, so the three ways of giving textual information about an
  image would be:

  textual equivalent - via alt attribute
  description - via longdesc
  interpretation - via actual content and/or markup

  To answer one objection: I am aware that we discourage the use of
  textual graphics. This discussion has no real bearing on that except
  to note that in an ideal world where CSS were implemented and there
  were strong associations between images and description/interpretation,
  there would be no further use for the alt attribute as I have proposed
  its use here.

  Thoughts? Is this merely proof I've been up too late at night and my
  mind isn't working properly, or is it a stroke of genius, or somewhere
  in between?

  --Kynn

  PS:  The term 'longdesc' is dangerous to accessibility. This term
        should simply be 'description' or 'desc' and future versions
        of XHTML should drop the 'long' prefix. The name of the
        attribute supports the mistaken notion that alt text is for
        description -- what is longdesc a longer description than?
        Another suggestion to pass along to HTML WG.

  CC: Courtesy carbon copy to Charles Munat, since I'm not sure if he's
       on the WCAG working group.


-- 
Charles McCathieNevile    http://www.w3.org/People/Charles  phone: +61 409 134 136
W3C Web Accessibility Initiative     http://www.w3.org/WAI    fax: +1 617 258 5999
Location: 21 Mitchell street FOOTSCRAY Vic 3011, Australia
(or W3C INRIA, Route des Lucioles, BP 93, 06902 Sophia Antipolis Cedex, France)

Received on Sunday, 3 February 2002 18:01:31 UTC