RE: content-072.htm is invalid

On Thursday, October 14, 2010 7:27 PM L. David Baron wrote
>
> On Thursday 2010-10-14 19:24 -0700, "Gérard Talbot" wrote:
> >
> > > On Thursday 2010-10-14 18:26 -0700, "Gérard Talbot" wrote:
> > >>
> > >> > http://test.csswg.org/suites/css2.1/20101001/html4/content-072.ht

> > >> > m relies on an attribute expansion mechanism that would expand
> > >> > <script defer> to <script defer="defer"> as far as what the CSS
> > >> > attr() function sees.  No such mechanism exists in
> > >> > http://www.w3.org/TR/html5/ , so this test is invalid.  I tend to
> > >> > think it should be removed.
> > >>
> > >> The default, implied value for defer attribute is "defer" in HTML
> > >> 4.01.
> > >> http://www.w3.org/TR/html4/interact/scripts.html#h-18.2.1

> > >
> > > That's not true in HTML5, and in HTML4 it was only a piece of
> > > fiction.

HTML5 has a bug here then because this is a big compatibility issue with CSS and the attr() function. Sites could be relying on the expansion and if user agents begin to support HTML5 parsing then sites will break. HTML5 needs to correct this and make attributes be expandable in all cases.

Second issue it that the CSS 2.1 spec clearly calls out the HTML 4.01 spec as a normative reference this means that any tests for CSS can use the entire range of what is valid in HTML 4.01. Until that reference changes to HTML5 the tests is completely valid and needs to stay and be supported by user agents.

> > Then, let's test with checked or selected or readonly.

The attr() function should also be able to get the value from these attributes.

> >
> > > The CSS test suite is not the place to test that HTML handling is
> > > based on SGML DTD processing.
> >
> > I agree that the CSS 2.1 test suite should not be an HTML4 test suite
> > at the same time. But then a browser should also have a decent
> > compliance with HTML4 to begin with.
> >
> > So what's your opinion on testing on minimized attribute checked or
> > selected or readonly?
> 
> We should not be testing for the values of minimized attributes in the CSS
> test suite.

We absolutely should be testing everything that is in HTML 4.01 it’s a normative reference in the CSS spec. This is how we determine if a user agent is supporting things correctly for CSS. We are not testing HTML 4.01 at the same time. We assume that user agents tested and fully support HTML 4.01. That is what we have to assume going into the testing of CSS. They are independent pieces of a complete system. If the CSS spec needs to be updated to also have a reference the HTML5 spec then this changes things but until that happens then this case is valid per the CSS 2.1 spec.

The test is testing exactly what CSS 2.1 requires and in this case it requires that it get a value from the attribute that HTML 4.01 and the user agent should support. The case as I see it is in no way invalid for CSS 2.1 however it may be 'na' for your implementation.

--
Thanks,
Arron Eicholz

Received on Monday, 25 October 2010 17:07:28 UTC