Re: [html4all] HTML5 Alternative Text, and Authoring Tools

For what it's worth, Adobe GoLive CS, when I drag an image in, and 
then save as HTML, inserts an empty alt attribute.  But clicking on 
the image shows (in the inspector where most entry and adjustment is 
done) a clear type-in field for alt text, so it's easy for the user 
to fix it.

More intriguing is Microsoft Word.  I'm using 11.3.5 (Word 2004 for the Mac).

You can insert images in documents, and you can save documents as 
HTML.  Now, generating HTML is not the primary purpose of this tool. 
In Word documents, there is no place to enter (as far as I can tell) 
alt text, and indeed, when you save as HTML, it generates an IMG 
without an alt attribute at all.

Imagine that Word did NOT have "save as HTML" and you are a junior 
engineer assigned to writing this functionality, and you are told 
that your HTML must be conformant, and you can't ask for changes in 
the rest of Word.  You are told that as a 'save as' plug-in you have 
no interactivity, no access to the user.  (And even if you did, I 
cannot imagine how a user would react if asked for alt text for all 
images at save time, and anyway, who says the person doing "save as 
HTML" was the original author anyway?)

What do you do?

a) leave the alt attribute out;  if it's required, that's 
non-conformant, you might get a bad review;
b) Heh, it's a WORD document, everything else, including images, is 
clearly decorational, put alt="";
c) make something up, like the filename or whatever, so as to keep 
the letter of the law:  alt="IMG_2745.jpg".
d) refuse to save any document with images in it, because you cannot 
fill in the alt attribute honestly and accurately.  (And lose your 
job, probably).

Word 2004 for the Mac does (a).  I'm sure others of you (probably 
most) would have done (b) or (c).  I suspect few would choose (d) 
(though at least one person has indicated that option).  Maybe there 
are other choices.  But four is already three too many.  I think that 
the editors are right in feeling we're going to get better interop. 
if we have more predictability and the UA can make reasonable 
deductions from the prescriptions (and proscriptions) in the spec., 
and if the spec. therefore says which one is the least bad 
('preferred') choice.

Given how poor all these choices are, some have suggested noalt in 
this case (which seems to be semantically equivalent to (a), though 
it reinforces that the tool *knew* it was in a bad position, that 
leaving alt out was not accidental).  I've (half-seriously) suggested 
alt-trustworthiness-level ranging from 0 (it's completely worthless, 
I only put in a value so as to pass syntax conformance) to 100 (the 
alt text is a brilliant and insightful piece of masterly writing that 
could not be improved on in so few words).
-- 
David Singer
Apple/QuickTime

Received on Thursday, 15 May 2008 19:06:25 UTC