Re: yet another non-entailment

Jim:
>  2)  While in principle I think we definitely need some entailment
>tests, unless I misunderstand my math, the possible set of all
>entailment tests we could need is infinite and enumerating even the
>set of types to try could take a very long time.  Thus, I would ask
>Jeremy, Peter and anyone else interested to consider how best to
>bound the test set -- as far as I can tell, in the past the WG has
>discussed the Test cases being tied to open issues and the specific
>language features in Mike's document.

>From this chair's point of view, a
>limited and focused test set is preferable to an open and
>comprehensive one

My understanding is that the current intent is indeed a limited test set
focused around issues and features.

To move the feature part ahead we need to:
- analyse the OWL Features document (or perhaps the reference doc)
- identify the features to create tests for
- assign actions to test focus area members (and others) to create test
cases for the identified features

I believe the issue part is currently doing OK.
This will be met if:
- when we discuss issues we habitually use test cases (this appears to be
happening)
- when we resolve an issue we action someone to create test cases (someone
should search for closed issues with no test cases)
- hopefully in the critical cases the "create test cases" step is more a
selection process from the e-mail archive than anything else.


I would be keen to see WG members such as Peter who create examples in the
e-mail archive to actually create these as proposed tests (and given our
current lack of policy) in the test case directory.

My take is that the examples which we decide to discuss or by virtue of
that, highly likely to be focused on the issues we are discussing, and hence
a priori good candidates for inclusion in our test repository.

Jeremy

Received on Thursday, 29 August 2002 09:37:11 UTC