Re: Against the notion of reification well-formed graph (i.e., atomicity)

On Tue, Jan 23, 2024 at 12:55 PM Peter F. Patel-Schneider
<pfpschneider@gmail.com> wrote:
>
>
>
> On 1/23/24 05:26, Doerthe Arndt wrote:
> > Hi all,
> >
> > I am joining rather late to the discussion,  so apologies if all I write has
> > already been said or my problem here is already solved (in the latter case I
> > am very thankful for your explanations :) )
> >
> > So, to look to Enrico’s point from another direction, let’s assume that we
> > have RDF and we would like to simply add support for owl:sameAs (for whatever
> > reason).
> >
> > Now, we would normally have that from
> >
> > :s :p :o.
> > :o owl:sameAs :k.
> >
> > we could derive
> >
> > :s :p :k.
> >
> > But what happens if we have
> >
> > :t rdf:subject :s;
> > rdf:predicate :p;
> > rdf:object :o.
> >
> > and
> >
> > :o owl:sameAs :k.
> >
> > following the same logic as above, we would derive
> >
> > :t rdf:object  :k.
> >
> > But with that derivation, we violate our wellformedness constraint which is a
> > purely syntactical constraint. So, here, we seem to have a problem because we
> > mix up syntax and semantics.  Maybe your proposals already contained a
> > solution which I missed? How do we avoid such derivations?
> >
> > Kind regards,
> > Dörthe
>
>
> Not so.  There is no notion of adding the triple back into the RDF graph.  So
> there is no ill-formed RDF graph here.

This is key. Along with a working notion of well-formedness (for both
reification and lists), it could pretty much seal the deal from my
perspective.

> What implementations do internally is not affected.  All that implementations
> are permitted to do is to reject ill-formed RDF graphs on input.  What they do
> internally is not changed - an implementation that adds consequences to its
> internal data structures is on its own.

What goes for querying and output? For instance, having something like
`FILTER [NOT] ENTAILED [VAR*]` to SPARQL would be valuable (a
standardized alternative to the mechanisms discussed in [1]). To be
able to selectively query over the entailed triples could also solve
some/most of the problem of referential opacity or transparency.

Best regards,
Niklas

[1]: https://github.com/w3c/sparql-dev/issues/124


> peter
>
>

Received on Thursday, 25 January 2024 09:03:37 UTC