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

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

From: Boris Zbarsky <bzbarsky@MIT.EDU>
Date: Fri, 20 Feb 2004 14:57:41 -0500
Message-Id: <200402201957.i1KJvgAA022471@no-knife.mit.edu>
To: Mikko Rantalainen <mira@cc.jyu.fi>
Cc: WWW Style <www-style@w3.org>

> However, it might make some sense to interpret invalid byte 
> sequences in UTF-8 file as invalid characters and proceed instead of 
> halting

There are some serious security implications involved in doing this, so I can
understand UAs being unwilling to do it....  If done, it needs to be done
_very_ carefully.

> I REALLY think we shouldn't specify new stupid rules simply because 
> there already exist some stupid tutorials. Fix the stupid tutorials 
> and specify good rules.

The tutorials are not the problem, really.  The massive installed base of
stylesheets that would break if we suddenly started treating them as UTF-8 is a
problem.  Unfortunately, quirks mode does not help much, since a number of
pages out there are in standards mode in modern.... and do not have sheets
properly labeled.  A good fraction of these are not in UTF-8.

Put another way, how many sheets out there _do_ have the charset labeled?
Other than testcases written to test algorithms such as the ones discussed
here, the only ones I've seen have been those written by people _very_
knowledgeable about CSS (people who regularly post on this list).  And even
then, they tend to be unlabeled [1].

In other words, the first UA to stop guessing would suddenly fail to render
every single site that triggered standards mode.

Boris

[1]  A brief survey of randomly selected sites that would have charsets set
properly if any would:

1)  http://ln.hixie.ch/ -- no HTTP header, no @charset rule, no attribute on
    linking element
2)  http://dbaron.org/log/ -- charset set in HTTP header.  Bravo!
3)  http://bjoern.hoehrmann.de/ -- links to
    http://www.w3.org/StyleSheets/Core/Swiss which has no HTTP header and no
    @charset rule.  Doesn't have a charset attribute on the linking element.
4)  http://tantek.com/log/ -- no HTTP header, no @charset rule, no attribute on
    linking element
5)  http://weblogs.mozillazine.org/bz/ -- no HTTP header, no @charset rule, no
    attribute on linking element
    http://web.mit.edu/bzbarsky/www/ -- same

    (OK, so using myself as an example is a little questionable).
6)  http://www.w3.org/ -- no HTTP header, no @charset rule, no
    attribute on linking element
7)  http://www.cc.jyu.fi/~mira/ -- no HTTP header, no @charset rule, no
    attribute on linking element

Needless to say, every single one of these pages triggers standards mode (in
Mozilla, at least, but I suspect they would in most browsers).  The only one
that would use the right charset is David's.

Now I did not look at the actual content of the sheets, so it may be possible
they are all ASCII and hence would actually work as UTF-8....

-- 
We are all agreed that your theory is crazy.  The
question which divides us is whether it is crazy enough
to have a chance of being correct.  My own feeling is
that it is not crazy enough.
                                     -- Niels Bohr
Received on Friday, 20 February 2004 14:57:47 GMT

This archive was generated by hypermail 2.2.0+W3C-0.50 : Monday, 27 April 2009 13:54:26 GMT