RE: [CSS21] Test Suite

> -----Original Message-----
> From: www-style-request@w3.org [mailto:www-style-request@w3.org] On Behalf
> Of Chris Lilley
> Sent: Thursday, July 22, 2004 2:24 PM
> To: Boris Zbarsky
> Cc: www-style@w3.org
> Subject: Re: [CSS21] Test Suite
> 
[snip]
> 
> Right. I agree that putting this into something served as text/html is a
> really bad idea. Since the discussion was xhtml I assumed it was served
> with the correct MIME type.
> 
> BZ> Of course sending XHTML 1.1 markup (which these tests are) as
> text/html is an
> BZ> issue in itself...
> 
> Yes, its a broken misfeature that should never have been suggested.
> Trying to send XML to a tag-souper and saying "don't use this" "leave a
> space here it might help" etc is a disaster.

So... I understand the points that Boris, Chris, and the others are making.
I am properly chastised, and I have changed the settings of IIS to send the
application/xhtml+xml Content-Type for the XHTML pages.  With that said, my
fear is we're losing the forest for the trees.

My suggestions were designed to expand the set of UA's that would have a
running chance at passing the CSS 2.1 interoperability test suite.  Using
constructs in the test suite that knowingly exclude HTML4 UA's is
arbitrarily limiting.

There's nothing in CSS that requires XHTML, and XHTML was specifically
designed so that "XHTML documents can be written to operate as well or
better than they did before in existing HTML 4-conforming user agents." [1]
Why not take advantage of this feature?

Enclosing the content of <style> elements in CDATA section markers is one
example of artificially restricting the set of UA's.  Although the section
markers are a requirement for arbitrary embedded style sheets, it is
optional when the embedded style sheet contains neither ampersands, nor
opening SGML tag delimiters (less than symbol).

The tests can be designed to meet this requirement (they do so far), and
more importantly SHOULD be designed to function without the use of CDATA
section markers, IMO.

So, my question to the others is:  Shouldn't the tests be written in as
inclusive a manner as is reasonably possible?

I'm not asking to leave aspects of CSS 2.1 untested, or to go to unnatural
lengths to support a wider set of UA's.  Even the cool, but convoluted CDATA
escape mechanism [2] is more than I would argue for.

On the other hand, minimizing the image support of the UA to a single
format, or simply dropping the CDATA section markers seem like very
supportable suggestions that demonstrably expand the set of available UA's.

-MM


[1] http://www.w3.org/TR/2002/REC-xhtml1-20020801/#xhtml
[2] http://lists.w3.org/Archives/Public/www-style/2004Jul/0138.html

Received on Thursday, 22 July 2004 18:17:41 UTC