Re: attribute-value-selector-004.xht not well formed

Sylvain Galineau wrote:
> David,
> 
> Shouldn't the statements below be reversed i.e.
> 
> p { color: green; }
> p, [1badAttr] { color: red; }
> 
> However, the only way to verify that the red rule is not applied to [1badAttr]
> is to have markup with that attribute in the page. We can certainly assume
> that if color:red not applied to p then it's not applied to [1badAttr] i.e.
> that the user agent handles selector error recovery properly for all the
> selectors in the group. The original testcase does not make that assumption.
> It may be conservative but it would also reveal a bad bug that the alternative
> wouldn't.
> 
> Imho, testing CSS should take precedence over the well-formedness of the markup
> in some of the testcases, as long as the broken markup relates directly to the
> feature being tested. Given how much invalid or broken markup there is in the
> real world, limiting ourselves to testing valid XHTML seems restrictive and may
> allow a PASS that should FAIL.

It is the opinion of two test suite Peers, L. David Baron and Anne van Kesteren,
that David Baron's suggestion to rely on the normative error handling rules as
above is sufficient to test that the invalid selector [1badAttr] is correctly
ignored. Arron Eicholz disagrees. The vote tally is 2 vs. 1 and the thread's at
a standstill.

It is my opinion as the test suite Owner, that David Baron's suggested test is
sufficient to test the assertion "Attribute values cannot begin with numerals."
If Arron wishes he can escalate the issue to the Working Group. Otherwise this
discussion ends here and the test must be revised as indicated before being
accepted into the W3C test suite.

~fantasai

Received on Saturday, 28 March 2009 12:24:21 UTC