action 224 on issue-62

I have reviewed issue 62. This seems to be the issue as originally raised by Sandro in 2003, referring to http://www.w3.org/2000/10/rdf-tests/rdfcore/amp-in-url/Manifest.rdf :

" I think you need to use an rdf:List for test:entailmentRules.  As you
> have it now, the test:entailmentRules arcs can be dropped by RDF
> simple entailment, but doing so renders the test statement false.
> For example, a PositiveEntailmentTest on RDFS entailment is likely to
> have its conclusions no longer follow from its premises if the
> entailmentRules arc is dropped."

First comment, this issue has been poorly titled. It might be better titled, "Bad RDF Practice in test cases manifest". It does not report an actual error in the manifest, in the sense that any of the tests would give a wrong or misleading result. What it does point out is that the manifest uses RDF with a closed-world assumption which is not exactly in accord with the RDF semantics. To be precise, the manifest, while itself correct, (simply) entails an RDF graph which is not a correct manifest. 

The response in 2004 (composed by Jan Grant) was basically to say, Oh Dear, but Never Mind because it would be too much trouble to fix this (and, implicitly, nobody cares, and nothing breaks.):

"The working group accept this comment that the test case manifest format
currently has some closed-world assumptions.

To be specific, test cases exist with multiple entailment rules,
supported datatypes and/or premise documents. A full fix to this would
require a change to the way those properties of a test case are
expressed.

It is felt that a change to the manifest format at this stage would be
potentially counter-productive, requiring effort from all maintainers of
test case harnesses in order to run the same set of tests.

While such a fix "would be nice", it is not felt to be critical to
delivering the test cases at this point.

Therefore the working group will create a postponed issue to track this
concern."

I propose that we now basically repeat this non-action, but put a more positive spin on it, and use the manifest as an example to illustrate the fact (as we now all understand it to be) that RDF semantic extensions can in fact impose what amount to non-monotonic conditions on RDF documents. So being-a-correct-test-manifest is an extension condition on RDF graphs that is not preserved under simple entailment. It is not the only one, we now know, eg being legal OWL2-DL is also not preserved under simple entailment, for a very similar reason (it requires declarations to be present), and I am sure there are many other closed-world and document-centric uses of RDF out there. So we should make this point explicitly, and the manifest provides a handy way to do that. 

I believe this discharges my action 224.

Pat



------------------------------------------------------------
IHMC                                     (850)434 8903 or (650)494 3973   
40 South Alcaniz St.           (850)202 4416   office
Pensacola                            (850)202 4440   fax
FL 32502                              (850)291 0667   mobile
phayesAT-SIGNihmc.us       http://www.ihmc.us/users/phayes

Received on Thursday, 24 January 2013 18:11:25 UTC