Inconsistencies in WAI guidelines

I'm crossposting this to w3-wai-ig, because it seems to go beyond the
scope of ER alone.

As members of WAI/ER will know, I am developing a new Site Valet tool
that combines WCAG-based accessibility reporting with validation.
The WCAG reporting is based on Rick Joliffe's Schematron, and takes
as its starting point the published WAI schema example.

On test-driving the new tool (which is currently available online
but is very much under-construction and could generate inconsistent
results - caveat visitor), I am - perhaps unsurprisingly -
confronted with a few gotchas, that appear to highlight
inconsistencies in the WCAG itself.  I'll deal with the first
of these in detail now; perhaps we can establish some guidelines
for how a tool such as mine should handle borderline cases.

+++++++++++++++++++
(1) Point the tool to almost any page containing a Form:
e.g. to Site Valet itself at <URL:http://valet.webthing.com/>

When confronted with Form elements, it complains
	"WCAG Checkpoint 1.1
	An input element should have some descriptive text:an alt or
        longdesc attribute"

This is correct according to the document linked at that point:
	<URL:http://www.w3.org/TR/WAI-WEBCONTENT/#gl-provide-equivalents>

As regards whether this is sensible, it seems to me to be at best
marginal.  Form Inputs as rendered in graphical browsers do not
carry cues that are denied to other readers, so a page author who
fails to provide adequate information is failing their mainstream
users just as much as accessibility-challenged users, and should
presumably be aware of the fact.  Exceptions to this observation
would appear to me to be pathologically-constructed cases.

On following the "Techniques for checkpoint 1.1" link to
<URL:http://www.w3.org/TR/WAI-WEBCONTENT-TECHS/#tech-text-equivalent>
I find that Form Inputs are here excluded from the list of elements
that should have text equivalents.

Neither is it required by
<URL:http://www.w3.org/TR/WCAG10-TECHS/#tech-text-equivalent>,
which is rather confusingly linked from the previous document as
"Checkpoint 1.1", while in fact being identical to it.

++++++++++++++++++++++++

I have discussed my own view regarding a particular checkpoint.
But the purpose of this tool is to implement the WCAG, not my views.
That leaves open the question of how to deal with cases where the
WCAG itself appears inconsistent, other than to exercise my own
judgement.

You can test-drive the prototype tool at
<URL:http://valet.webthing.com/xslt/waitest.html>
(bearing in mind that it is under active development, and may
change at any time).

Any thoughts?  Comments?  Suggestions?

-- 
Nick Kew

Site Valet - the essential service for anyone with a website.
<URL:http://valet.webthing.com/>

Received on Sunday, 18 November 2001 00:15:18 UTC