Re: Change proposal for ISSUE-85

On Tue, Jun 15, 2010 at 11:19 PM, Ian Hickson <ian@hixie.ch> wrote:
> The role="" attribute gives us the unique opportunity to catch this
> particular error as a syntax error. This is a good thing, not a problem.

This encourages authors to migrate from detectable errors to
undetectable errors.  For example, if alt were not a required
attribute on <img>, most authors would omit it entirely, and AT would
apply whatever algorithms they use when no alt text is provided.
Since it is required, most authors still omit it, but a substantial
number just make up some nonsense to shut up the validator.  They use
the empty string, or the filename, or some random other string that
doesn't actually give any more info.  This means that the AT can't
tell whether an image with (say) empty alt text a) is actually just
decorative, or b) has some contents that the user might profitably be
alerted to, but the author suppressed them to shut up the validator.

This doesn't mean that alt text should not be required on <img>.  But
adding more conformance requirements can have downsides, even if it
would be great if everyone followed them.  You seem to be deciding
what pages should look like (ARIA attributes not contradicting
semantic markup), and then requiring that, without regard to the
practical effect it will have on typical authors who bother to use
validators.

In this case, it seems very likely that authors who run into this sort
of error will be tempted to just remove the role attribute rather than
use another element.  This is a disadvantage that should at least be
considered, but you don't mention it anywhere -- like most of your
change proposals, this one has "None" for "Negative Effects".

By the way, I don't see why search engine spiders, text-only web
browsers, and so on couldn't support the role attribute somehow.

Received on Wednesday, 16 June 2010 17:12:35 UTC