E-mail List Archives

Automation? (was Label vs ALT tag on form elements)


From: Michael D. Roush
Date: Mar 23, 2007 7:40AM

Patrick H. Lauke wrote:
> Again, it's the AT support that's the clincher here, and the impact this
> choice of markup would have on usability in general.

Not that I'm desperately actively searching for a project for the coming
year, but, here goes....

There are quite a few "automatic" code checkers out there. Each one is
slightly (or a lot) different. Mostly, they check our code against one
or more standards. There are also some tools out there that give you an
idea of whether your code will have difficulties being rendered by
various browsers.

I've been around for several discussions on this list about the relative
usefulness/uselessness of such tools, and I'm not trying to retrace old
steps. Let me just say that I'm coming from a perspective that
acknowledges the existence of such tools and some individuals' tendency
to use them to varying degrees, either as their entire accessibility
strategy or just a starting point.

Would it be helpful to have an automated checking tool that would
analyze a bit of code (i.e., a web page), and give a report about real
or potential difficulties with various assistive technologies?

For instance, the checker is scanning through the submitted code and
finds a "foo" tag with no attributes. The checker informs the user that
the "foo" tag is in use and that it is not supported by AT tools B, C,
and D, and that it is only supported by A if it has the bar="" attribute
(which is missing in my example).

I realize this would take a LOT of research and testing, preceded by
coming up with as complete a list as possible/desired of important
tags/elements to look for (labels, alt attributes, table summaries,
noscript tags, whatever), and assessing how various tools use or ignore
those tags/elements. But, if a tool like that could get us quickly past
the step of saying "I forget how such-and-such version of so-and-so does
with this kind of setup."

To anyone who just doesn't like automated checkers in the first place, I
realize you would never use the tool. I would never suggest that it be
a complete solution, but maybe a starting point in letting people who
are interested in not just producing "standard" code, but also usable
code, a tool to use to save them some scouring and help them go into a
live testing phase with fewer errors than they would have without the scan?

Just thinkin'.