W3C home > Mailing lists > Public > www-style@w3.org > February 2004

Re: [CSS21] response to issue 115 (and 44)

From: Jukka K. Korpela <jkorpela@cs.tut.fi>
Date: Wed, 25 Feb 2004 19:38:21 +0200 (EET)
To: WWW Style <www-style@w3.org>
Message-ID: <Pine.GSO.4.58.0402251927100.7731@korppi.cs.tut.fi>

On Wed, 25 Feb 2004, Henri Sivonen wrote:

> Working around these issues
> would require lot of work for something that is classified as error
> recovery.

That's why I suggested that error recovery be left to user agents, except
in a few clear-cut cases.

> > How would that reduce the usefulness of a specification?
> You could not rely on the specification to tell you what to implement.

You could implement as expensive error processing as you can afford,
and claim conformance.

> Perhaps, but experience with HTML suggests that leaving the behavior in
> error situations undefined is a bad idea and leads to reverse
> engineering the implementation with the largest installed base.

CSS already has error processing rules for the actual style sheets.
There's a point in this, despite the market leader's refusal to take those
rules seriously at all, but I don't think the principle of setting
requirements on error processing needs to be extended to metadata.

> The expectation is that if the the spec defines error recovery, the
> market leader doesn't have to invent its own error recovery behavior.

They already have their "error recovery" for actual style sheets (far
more important, aren't they?), despite the very explicit rules
in CSS 1 and in CSS 2. If browser vendors wish to design different error
processing strategies for metadata, I don't think it's a problem.

> Is it better to always guess wrong and know it or to guess
> sometimes right and sometimes wrong and not know when the guess is
> right and when it's wrong?

I don't think we have such alternatives. The simple alternatives are
rigorous error processing rules (which are known to result in wrong
guesses, unless you impose the rule of not rendering a document if there
is an error) and leaving error processing rules to user agents.

In fact, if style sheets are used properly, the principle of ignoring the
entire style sheet when its encoding cannot be decided would not prevent
the page from being used. In this sense, even showing the page without the
style sheet (maybe showing a small flag saying that the author tried to
suggest some rendering but didn't do that properly) would be an acceptable
behavior. A cynicist, or a realist, might say that in such a case, there's
no reason to expect that the styled presentation would be any improvement.
Anyway, user agents should have such options; they should not be forced to
obey style sheets that are known to violate publised specifications - and
this is what rigorous error processing rules would imply.

Jukka "Yucca" Korpela, http://www.cs.tut.fi/~jkorpela/
Received on Wednesday, 25 February 2004 12:38:23 UTC

This archive was generated by hypermail 2.3.1 : Monday, 2 May 2016 14:27:11 UTC