Re: TEST: proposal for test case structure

On Thu, 2002-07-18 at 06:27, Ian Horrocks wrote:
> 
> On July 16, Jeremy Carroll writes:
> 
> [...]
> 
> > 
> > 4.3 Consistency test
> > 
> > A positive consistency test consists of one document.
> > The document is consistent.
> > 
> > A positive consistency test consists of one document.
> > The document is inconsistent.
> 
> What do you mean by an inconsistent document? Is it an OWL KB for which
> there is no model?
> 
> What are the allowable answers to consistency/entailment questions?
> Are they two valued (T/F or whatever) or three valued (including "don't
> know")? If you allow "don't know" answers, then a system can be fully
> compliant simply by answering "don't know" to all
> consistency/entailment questions. If you don't allow "don't know"
> answers, then incomplete systems will not be compliant.

Exactly; the point of the tests is to say "for this one,
you have to answer yes" and "for that one, you have to
answer no". If the WG approves a test where the
answer is yes, and an implementation says "I don't know",
it doesn't conform.

The non-entailment tests are relevant too: the WG
might decide that our spec doesn't justify going from
	John in intersection(Student, Employee)
to
	John in intersection(Employee, Student)

that would leave implementations free to answer "I don't know"
in that case.

Hmm... hard to see how to actually test software
against the non-entialment tests... maybe their
main function is documentation.


Hmm... maybe I'm just confused. It's a good question, anyway.

I'm pretty sure I understand how (positive) entailment tests work:
if the WG approves them, a conforming OWL reasoner (let's say)
must answer "yes, that conclusion follows from those premises."
I'm not sure about the rest.

-- 
Dan Connolly, W3C http://www.w3.org/People/Connolly/

Received on Thursday, 18 July 2002 14:49:12 UTC