W3C home > Mailing lists > Public > public-css-testsuite@w3.org > March 2009

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

From: Sylvain Galineau <sylvaing@microsoft.com>
Date: Wed, 11 Mar 2009 09:13:17 -0700
To: Bjoern Hoehrmann <derhoermi@gmx.net>
CC: "public-css-testsuite@w3.org" <public-css-testsuite@w3.org>
Message-ID: <5D97C7EB4695104AB6345E56FE356B19408A7B45C1@NA-EXMSG-C125.redmond.corp.microsoft.com>
> The reasoning behind the test in question is sound, but it requires a
> markup language where following well-defined processing rules results
> in an attribute name beginning with the digit "1". That makes XML 1.0
> unsuitable for the test. Perhaps "HTML5" would be a suitable language.

Yes, it does seem as if we are arguing this backwards. Because XHTML must be well-formed then these tests must be well-formed, instead of : it is better for CSS testing to reduce or eliminate dependencies on the handling of invalid markup by HTML parsers - if you can't depend on proper handling of valid HTML, you have a larger problem than testing CSS - therefore, as a general principle, testcases should aim to use valid markup. And XHTML enforces that.

But it does not follow that no test can ever use invalid markup. No test suite that aims to be complete can limit itself to valid input. Or eliminate a test because the effect it checks for has its cause tested somewhere else. There will always be some redundancy (and there should be some in a test suite imo). It does not follow either that a test should be rejected on the sole basis of its lack of well-formedness. We're not testing XHTML well-formedness. But if parsing errors prevent a valid testcase from even rendering then its format needs to change.
Received on Wednesday, 11 March 2009 16:14:34 GMT

This archive was generated by hypermail 2.2.0+W3C-0.50 : Monday, 20 September 2010 17:51:58 GMT