substantive semantics change?

pat:
> > 8. (related). The definitions of rdf and rdfs entailment have been
> > simplified so that they do not make explicit reference to a
> > vocabulary. This is actually more conventional; and Herman pointed
> > out that the more complicated definitions meant that entailment might
> > not be transitive (aargh). The motive for introducing this
> > complication in the definition in the first place has been removed by
> > subsequent changes. This doesn't change any test cases.
> >
jeremy:
> My understanding is that test cases do change - in particular the
> test cases
>
>
> >   rdfms-seq-representation/Manifest.rdf#test002
> >   rdfms-seq-representation/Manifest.rdf#test004
>
> which were incorrect according to the LC2 semantics doc, now
> become correct.
> So this is a substantive rather than an editorial change. (I am awaiting
> feedback from HP implementors concerning this)
>

Further, as far as I understand, the Last Call 2 doc had the empty document
entailing only a finite number of ground triples, whereas the current
editors draft has the empty document entailing an infinite number of ground
triples.

HP is unlikely to implement such a change, and would oppose it, if it were
formally proposed.

Moreover, since this substantive change (assuming my understanding is
correct) had not been highlighted before the PR vote, we may need to ask the
chairs to rewind history a little.

I don't follow the transitivity argument ...

G entails H if every interpretation of G is an interpretation of H.

this is clearly transitive, and a "counter-example" like

*empty*
  entails
rdf:_1 rdf:type rdfs:Resource
  entails
rdf:_1 rdf:type rdfs:ContainerMembershipProperty

is flawed because the first entailment is false according to the LC2
semantics.

Is it possible to undo some of the changes to semantics without it falling
apart, leaving hermann and/or peter dissatisfied?

Jeremy

Received on Monday, 10 November 2003 10:53:09 UTC