Re: entailment-from-inconsistent-graph: new test case request

[message cut and re-ordered, hopefully still makes sense]

On Fri, 31 Oct 2003, Dan Connolly wrote:

> On Fri, 2003-10-31 at 09:48, Jan Grant wrote:

> [ proposal to suggest splitting PFPS' test case into two pieces ]
>
> I don't mind these two tests *in addition* to the one Peter
> requested, but to do anything less than include his test
> as suggested says to me that we do *not* accept that
> this entailment holds, or that it's not illustrative
> of a very special issue.

If PFPS insists on the original test case standing, then my suggestion
would be to add these two test cases as well, and mark (in the
description) PFPS' test case as a special case of (1) the inconsistency
in rdfs case, and (2) the "basic rules of entailment" test case.

[ on "basic rules of rdfs entailment" test case ]

> That suggests that false premises are typical in RDFS. But they're
> not. It's only this one very, special case of XML literals that
> allows falsehoods to be stated in RDFS.

I don't agree that the test case makes any such suggestion; rather, it
underlines that a false premise *-entails anything. We could certainly
indicate in the test descriptions that rdf can't construct inconsistent
graphs; that rdfs can only do it via the XMLLiteral, or whatever.

Basically I'm concerned that PFPS' test case, as it stands, is (possibly
deliberately) overly obscure: therefore I suggest negotiating with him
to see if he's satisfied that the addition of a test case that more
transparently demonstrates the construction of an RDFS inconsistency,
and an "entailment 101", which demonstrates that any conclusion at all
can be drawn from a false premise (such as an inconsistent graph),
adequately address his concern.

If he _is_ so satisfied, then I think the test cases in question will be
better off for it, since their illustrative purpose will hopefully be
more successfully met.

If he's _not_ satisfied by that, then I'd certainly agree that we owe
him such a test case* if he honestly thinks it's useful; but I'd be
inclined to add the illustrative ones as well and indicate that PFPS'
test case is a particular instance of those more general ones.

jan

* I've used DanC's point of view on this before to bolster the addition
of test cases that I thought were useful, and I'd hate to be
hypocritical here.

-- 
jan grant, ILRT, University of Bristol. http://www.ilrt.bris.ac.uk/
Tel +44(0)117 9287088 Fax +44 (0)117 9287112 http://ioctl.org/jan/
Usenet: The separation of content AND presentation - simultaneously.

Received on Friday, 31 October 2003 11:46:01 UTC