RE: use of alt attributes in decorative images

I agree that this particular problem is a bug with JAWS.
As you know, I am a big fan of spec-compliant coding and UA.
As far as browsers go, we've got maybe IE on the Mac and Netscape/Mozzilla 6
that come close to rendering HTML 4 tolerably well.  What behavior would be
reasonable for a browser for code like:

	<a href="foo.html" title="Big World of Foo."><img src="foo.png"
alt="Come to Foo!"></a>

The current state of the art is, I would argue, fairly reasonable and
predictable.  With a screen reader, the link title and graphic alt are read
in sequence.  It is actually clear which belongs to which.  Screen reader or
not, I can preview the link title before activating the link.  Graphically,
a mouse-over pop-up has a choice of two content pieces and, quite
reasonably, the title of the link takes precedence (and this additional
information is rendered rather subtly).  Oh yeah, I also have the image or
the associated alt -- depending if graphics are loaded or not.  Okay then,
with screen reader or not, my browser has unobtrusively presented me with
(1) a graphic or ALT content, (2) URL and associated (title) text.  Life is
good.

Now, suppose I want to be on the bleeding edge.  Or maybe I am confused
about what makes good ALT content -- and I am trying to make everybody
happy.  I mistakenly figure that more is more, so I code something like:

	<a href="foo.html" title="Big World of Foo."><img src="foo.png"
title="Foo Globe Icon" alt="Come to Foo!"></a>

Sure, this is perfectly logical and valid.  But face it, this gives me THREE
textual associations for the graphic.  As the author, there is no way I
could guess what the user is going to get!  I cannot come up with a scenario
that would reveal all three (or even two out of three) in any natural way
(with out requiring extra keystrokes on the part of the user, or some other
artifice) for a graphical user agent.  Can you?

Please Kynn, if you think titles should be included everywhere, please
explain how they might be revealed in the graphical user interface!  It gets
even more complicated, since the above example could be inside a P or DIV
with its own title!  Yeech!

TITLE is largely superfluous, except on links and ABBR and ACRONYM.  Until
user agents figure out how to handle the attribute, it should be avoided.
This IS a practical, real world, accessibility issue.

Kynn, depending on the weather, you are frequently in favor of backwards
compliant coding and even minor code tweaks to make up for idiosyncratic
behavior of Assistive Technology.  I would argue that this belongs in that
category.  Unless you are certain of the resultant browser behavior, avoid
TITLE.

-- Bruce

P.S.  One _could_ make the argument that the behavior of JAWS is NOT broken:
TITLE is a newer attribute than ALT and therefore an author is more likely
to put more thought into TITLE (if present) than ALT (especially since we
know how bad ALT tends to be).  Besides, IE visually renders TITLE in
deference to ALT -- so that is the one people pay more attention to.
Therefore, if both are present, TITLE must be the more important one -- so
that's the one that should get read by the screen reader.


> ----------
> From: 	Kynn Bartlett
> Sent: 	Monday, February 5, 2001 1:00 PM
> To: 	Bailey, Bruce
> Cc: 	'w3c-wai-ig@w3.org'; 'poehlman1@home.com'; 'kynn@idyllmtn.com'
> Subject: 	Re: use of alt attributes in decorative images
> 
> At 09:26 AM 2/5/2001 , Bailey, Bruce wrote:
> >Okay, I am using the latest version of JFW on a Windows 2000 box.
> >When I come across an image (not inside a link) that has BOTH a title and
> >alt attributes, JAWS reads ONLY the title -- and I can not get it to read
> >the alt content!
> 
> This sounds to me like a user agent problem, not a markup problem.
> The markup is done properly, but the assistive technology fails
> on it; whose responsibility is it to correct this problem?
> 
> In my opinion, it's on the shoulders of the JFW people -- but if
> someone is specifically designing an interface for JFW (such as
> an edaptive system that recognizes JAWS), the optimal code for that
> user agent should be delivered.
> 
> >IMHO titles belong on links -- and that's about it.  (Yes, Kynn, I know
> the
> >specs allow them just about everywhere, but that's not the point.)
> 
> Sure it is -- the user agent is not respecting the spec, and neither
> is your humble opinion, so why should we listen to either?  Title is
> meant to be used, potentially, EVERYWHERE, and a user agent which
> breaks on such a circumstance is a faulty user agent.
> 
> The question then becomes whether or not you should compensate for
> that problem in the user agent (in my model with an edaptive
> interface, I would say yes), but you surely should not condone this
> type of faulty behavior by suggesting that _only_ links can have
> titles, because JFW got it wrong.

Received on Monday, 5 February 2001 15:22:55 UTC