RE: [www-qa] Re: Conformance and Implementations

On Tue, 23 Oct 2001, Karl Dubost wrote:

> It's why the test cases should be built by the WG or by an
> external resource WITH the WG at the earliest stage of a REC (I
> mean WD). Test cases are useful tools for developpers, but there
> are also an easier way to write a clear and unambiguous
> recommendation.

I do not believe that stating requirement X twice (once in the
recommendation text and once in the test case embedded in that
recommendation) will somehow make the recommendation less ambiguous.
IMO, this will only increase the number of ambiguities as some test
cases will sure contradict the text!

I believe it is better to have _one_ authoritative requirement. Since
most of us are better at reading human languages rather than RDF or
XML, that requirement should be formulated in a human language. Test
cases should be non-normative illustration whether they are developed
by the WG or not.

Having test cases helps improving documentation quality. Having
embedded test cases has a negative side-effects that are easy to avoid
by separating test cases from the authoritative documentation.

Alex.

Received on Tuesday, 23 October 2001 12:07:42 UTC