Re: How to make complex data tables more accessible to screen-reader users

>
>  ( ) I support the design of the HTML4 working group.
>     (Including the summary="" attribute on tables.)
>
>  ( ) I support the design currently in Ian's HTML5 proposal.
>     (Suggesting that tables should be described in captions.)
>
>  ( ) I support the design currently in Rob's HTML5 proposal.
>     (Allowing summary="", but saying it doesn't work.)
>
>  ( ) I have another proposal. Describe it below.
>
>

First, just a note to set expectations. I'm not an accessibility
person, and when I give my opinion, I'm doing so as a general web
designer/developer who is interested in providing accessible pages,
not as part of any specific accessibility group.

First, thank you for pulling all of the relevant past research on
summary into one post. It's helpful to ensure we all start from the
same basic starting point.

I do have a point of clarification:

I haven't seen Rob's new document, but in the emails we exchanged
regarding it, he wrote, "The first edit I am planning is to make the
summary attribute conforming, with a note explaining that it hasn't
always worked very well, and may continue not to." That's not the same
as saying, "it doesn't work". It's basically a statement that it
hasn't always worked well in the past and may not continue to.

I would expand on this option by also providing more details in the
HTML5 document about how to use summary correctly, in line with the
expanded discussion on how to use HTML tables correctly. All of the
empirical data you've linked also demonstrates that few people have
used HTML tables incorrectly, and so it's difficult to separate the
two and say one is a keeper but with a better definition, the other
can be tossed. The usage is basically "corrupt" (sorry can't think of
a better word) for both.

In addition, I have to question about whether forcing the descriptive
data into visibility will make a difference. If people can't be
bothered to use summary correctly, I have a feeling that telling
people put this information that they've not provided in summary into
caption, or incorporating it into the prose surrounding the table just
won't improve accessibility. We're using technology to attempt to
change human behavior, which tends to have a high rate of failure.
We'd be better off focusing less on the markup, and more on changing
the human behavior.

Regardless, I also agree with Lachlan, in that we should not combine
caption and summary because their content and purpose are different.

Speaking more generally, and I apologize for this being somewhat
off-topic, but I'm reluctant to start a new thread, I think it's fine
to look at web scrappings to see how people are using HTML elements,
but I think you're over relying on such. We have no control over how
the raw data is collected, which means we have no way of assessing how
old the pages were, or how often the web bots were prohibited from
accessing actual pages because of robots.txt instructions, or how
often they just plain failed.

We just cannot tell that the raw data used for the queries is truly a
representative sampling. If nothing else, we have to question the fact
that the sampling size seems disproportionately small.

As for using screen scraping for data as a decision point: The web has
been around for several years now, but much of the web is still caught
up in old legacy static pages, created when we didn't even understand
how to use paragraphs correctly. The fact that so many of the tables
are navigation and layout tables demonstrates this.

The web has changed, though, in the last 5-7 years, with more reliance
on content management systems to generate web pages. Rather than have
to convince a million people not to use HTML tables for their sidebar
items, we now only need to convince widget or CMS developers not to
use HTML Tables for navigation, and in the next upgrade of the
software, the HTML table problem has been resolved for potentially
millions of pages.

The sampling could be showing massive number of errors and incorrect
usage now, but query again in a year, and you may find drastically
different results. Especially as the old static web page hosting
sites, such as Tripod, close down their sites, literally removing
millions of really bad web pages from any selection set. And again, we
don't know if those collecting the raw data update their sets to
reflect dropped or modified pages.

So, to make a long explanation short: I think it is a mistake to build
the future of the web primarily based on the mistakes of the web of
the past.

I think the data and queries are one good tool to use in making a
decision, but they shouldn't be the only tool used, or even the tool
given the most importance.

Shelley

Received on Sunday, 5 July 2009 16:33:59 UTC