Re: Selectors+accessibility (the XML threat)

On Wed, 20 Aug 2003, Silas S. Brown wrote:

> There is a disturbing trend for an increasing number of Web
> documents to be written in XML (not HTML) and to have the
> CSS applied directly to the XML.  This means that
> accessibility stylesheets don't work, because the HTML
> elements that they apply to simply aren't there.

This is a much wider problem. Consider a specialized browser that uses no
style sheets (e.g., because it operates in ways that make current CSS
completely or almost irrelevant, such as rendering in Braille).
It has really nothing to base the rendering on, if the document is in XML
using "vocabulary" (in the XML sense) that is not known to the browser.

CSS is relevant in the big picture though. It's one of the technologies
that have made tag soup XML popular. By tag soup, I mean XML data that is
well-formed and perhaps even valid but lacks any public, intelligible and
clearly formulated semantic definition. Even if such definitions exist,
they are virtually unsupported by mainstream browsers, not to mention
specialized browsers, which are important to disabled people. Lots of
warning signals have been sent about this, but I can just unhappily state
that my predictions in 1998 ("Lurching Toward Babel: HTML, CSS, and XML",
http://computer.org/computer/co1998/pdf/r7103.pdf )
pretty much describe what has happened.

So what can we do? Just state that luckily the progress of CSS is slow?
Especially after Microsoft's announcement of its policy on IE,
CSS 3 is mostly just an expert game, or a practical joke.

> If people increasingly use their own XML DTDs instead of
> HTML then something more is needed for accessibility.

When people do that, they are fighting against accessibility (among other
things), and there's little we can do about it.

> *::[style:font-weight=bold]

If I understand your idea as a whole correctly, it is based on an idea of
reverse-engineering a style sheet in order to deduce the logical structure
that it is supposed to reflect. There's probably something in the idea
that we can try to guess the intended structure from the presentational
suggestions. Usually bolding is meant to be some kind of emphasis, for
example.

But it would probably be easier, and more productive, to base the reverse
engineering on the actual appearance, as captured e.g. though the same
routines by which a screen reader gets its input, peeking at data going
onto the screen. It would _also_ help to detect e.g. the very common use
of <b> in HTML to indicate emphasis and to render the assumed emphasis in
some suitable way. Or even the use of bolding in text processing programs.

Thus, I don't think there's any reason to make CSS more complex for such
reasons.

-- 
Jukka "Yucca" Korpela, http://www.cs.tut.fi/~jkorpela/

Received on Saturday, 23 August 2003 06:47:38 UTC