Sorry, I maybe did not make my sarcasm clear enough at that point.
Support to longdesc is virtually nonexistent in mainstream browsers,
and why would a user switch to a special browser just because he
wants to check the long description of an image?
I fully agree.
Actually most users probably have no idea of the longdesc attribute,
and extremely few pages use it, so I don't think users _should_ know
about it.
This is a self-fulfilling statement. If no-one uses it, there's no
point in implementing it. That's sort-of true. If no browser
implements it, there's no point in using it? Well, no, not exactly.
Using it does no harm to anybody, and may occasionally stimulate some
action on the browser side, such as implementation in the browser, or
a plug-in, or a bookmarklet etc. etc.
The alt attribute is sufficient for most images. If an image needs a
long description, then it should be provided using normal methods,
as text on the page near the image or on a page linked to from
there, using a normal link. That's simple, that's understandable,
and it works across browsers.
There's probably a wide range of situations where that procedure
would be appropriate. In any case, it's arguable that for web
browsers it's better to use <object> instead of <img>, with a graceful
fallback for the browser-like OS component.
http://ppewww.ph.gla.ac.uk/~flavell/www/obj-for-img.html
That brief sketch doesn't go into detail about how one would
produce the counterpart of longdesc (it wasn't really composed for
that purpose), but I think it should be reasonably obvious.
The longdesc attribute is just too poorly supported and too implicit
for anything serious. I would go as far as saying the same about the
title attribute, but you might not follow me there.
If I ever get to write this page I was talking about, I would
certainly cover the browser difficulties with longdesc. Oh, right, I
do disagree with you about title= - browser implementations for title=
aren't ideal, but at least they are there. And also, title="" is a
handy workaround, if no title is wanted, for MesSIE's habit of
popping-up alt texts at inappropriate moments.
The only reason why it could make sense to _hide_ the reference to a
long description is that we might think that the long description is
only for those who cannot see the image and would therefore see or
hear the reference instead. But that's not how browsers work, and
that's not how they are required to work. Besides, even if the
longdesc attribute worked technically, how could we know that the
long description does not benefit people who _do_ see the image but
just don't understand it?
Those are all fair comments, indeed.
There's one more reason not to tell authors to use longdesc. In
addition to being practically useless, it complicates things. It's a
new concept to be learned, remembered, and applied. It's an
unnecessary complication in a world where most authors still don't
understand even the simple principle behind the alt attribute.
I can see what you're getting at, but in general I think you're being
unduly pessimistic. While most of the world seems to be beavering
away with some grand-scale experiment of finding just how many faults
they can get into a web page without it obviously breaking in their
supposed mainstream viewing situation, there's still a hardcore of
folks who are trying - and even succeeding - in doing a proper job,
producing pages which behave with reasonable grace and comfort in a
wide range of browsing situations.
So much for the general. As you say, longdesc is a specific problem.
But I wouln't be trying to base the argument on not using longdesc, on
the fact that most alt texts seem inept. Yes, I know that even the WAI
keep getting it wrong, and keep switching from "textual alternative
for the purpose of the image" (good) to "description of the purpose of
the image" (usually wrong).
