Re: WD-WAI-PAGEAUTH-0203 and the use of ALT text

For the benefit of those reading the archives, the original message to
which this is a reply was sent to the IG list rather than the GL list.
On Thu, 5 Feb 1998, Alan J. Flavell wrote:

> The principle from which I have proceeded is that the ALT text is
> literally "alternative text", i.e it provides a textual substitute for the
> _information_ or _purpose_ of the image.  There are other attributes
> (TITLE and, now, LONGDESC) available for other purposes such as
> _describing_ the image. 

This is indeed precisely the principle which I have been advocating in
this working group for some time, and which I understand to have been
accepted as a broad position of concensus.
> 
> At:
> 
> "Provide alternative text for all images and image maps."
> 
> it says:
> 
> "Each image should have alternative text that describes the function of
> the graphic"
> 
> I disagree.  What is wanted is not a _description_ of the function, but
> the function itself.  Of the examples given, "Section Title: Banana
> Products,"  and "Search Button." seem fine, but "Graph of population
> versus age," is _no_ alternative text presentation of the _information_. 
> Rather, this would be more appropriate IMO as the TITLE attribute.  What
> the text mode reader needs to know is not _that_ there is a graph present,
> but what the heck does the graph show that is relevant to the topic under
> discussion?  If the text already makes that clear, then maybe ALT="" is
> appropriate.  Surely, calling the text-mode reader's attention to a graph
> that they can't read, while telling them nothing about what information it
> contains, is worse than useless?  
I partially agree. The ALT text, however, is appropriate in that it both
identifies the presence of the graphic and indicates the nature of its
contents. However, it would be inadequate unless accompanied by a long
description, which would convey the detailed information presented in the
graph, for instance by means of a table or by a description of the graph.
In an educational setting in particular, or if a person with a disability
is discussing the information with her/his colleagues who are viewing the
graph, it is helpful to know at least which axis is used to express which
variable.

The example should be modified to include the LONGDESC attribute.

 > > And
in the "possible strategies", again we have a
description
of > information, instead of the information itself.  <IMG src="logo.gif" 
> alt="XYZ Logo"> cannot possibly be right, in my view.  If the logo is
> there in order to identify the company, then the name of the company would
> be appropriate, e.g ALT="XYZ Corp", whereas if the company name is already
> present in clear text, then ALT="" would seem more appropriate, unless
> their logo is somehow pertinent to the topic under discussion, rather
> than mere identification.
The question here is whether it is important for the reader to know that
there is a logo present at this point in the document, as well as the name
of the company. The best solution to this problem would actually identify
both the name of the organisation and indicate the signifiance of the
logo. For an excellent example, see http://www.srl.rmit.edu.au
 > 
> With a little thought it's possible to choose ALT texts such that the text
> reads naturally, and the text-mode reader is scarcely aware of the
> "seams". Pages that have been composed "by rote", putting ALT="XYZ logo" 
> whenever the logo occurs, tend to read idiotically on a text mode
> browser.  Almost as bad as those that have ALT="large red bullet" etc.
I agree. The logo is best described perhaps only once, on the home page.
Thereafter, it could be treated as decorative and receive a null ALT
value.

> I would rather say "The ALT text should substitute for the
> _purpose_ of the image; the mere description of the graphic will rarely be
> appropriate for doing that". 
> 
I agree.

Received on Thursday, 5 February 1998 19:25:20 UTC