Re: RDF / RDFS semantics for the liberal baseline

On Mon, Jan 13, 2025 at 5:31 PM Franconi Enrico <franconi@inf.unibz.it>
wrote:

> Let’s continue the discussion about the RDF / RDFS semantics for the
> liberal baseline.
> After the finalization of the liberal baseline simple semantics, it is
> necessarily the case that the RDF / RDFS semantics have to be at least as
> follows.
> RDF SEMANTICS
>
> RDF interpretations add the following new IRI with the namespace prefix
> rdf: rdf:reifies.
>
> The rdfD2 RDF entailment pattern is replaced by the following entailment
> pattern, so to have a sound and complete set of metamodelling RDF
> entailment patterns: ⏪
> if the triple structure appears in S then S RDF entails
> *rdfD2-ts* sss aaa ooo aaa rdf:type rdf:Property .
>
I tend to believe in this. It appears integral to the abstract grammar: if
it appears as a predicate, it is a property. That is, all propositions in
RDF have properties in their predicate position.

But if completeness is not necessary, I have no strong arguments for adding
the rule. I do think applications should be able to rely on this being the
case; presumably the domain of RE is enough for that.


RDFS SEMANTICS
>
> The rdfs4a and rdfs4b RDFS entailment patterns are replaced by the
> following entailment patterns:
> if the triple structure appears in S then S RDFS entails
> *rdfs4a-ts* sss aaa ooo sss rdf:type rdfs:Resource .
> *rdfs4b-ts* sss aaa ooo ooo rdf:type rdfs:Resource .
>
 Is this trivially true? (It appears to be if I interpret what Peter said
correctly.)


The question is what people want to do with rdfs:Proposition.
> Possibilities:
>
>
>    1. I want that triple terms are of type rdfs:Proposition.
>       1. YES
>       2. NO
>
> Yes. (A triple term denotes a proposition.)


>    1. I want that objects of rdf:reifies triples at top level are of type
>    rdfs:Proposition.
>       1. YES
>       2. NO
>
> Yes. I'm fine with that in RDFS only (rdf:reifies rdfs:range
rdf:Proposition).


>    1. I want that objects of rdf:reifies triples at any nested level are
>    of type rdfs:Proposition.
>       1. YES
>       2. NO
>
> No. I don't see it as necessary as it doesn't appear to be integral to the
abstract grammar. (Similarly, a proposition of unknown truth value with
rdf:type as the predicate does not imply that its object resource is an
rdfs:Class. Some applications might have trouble with these not always
holding for some reason, but that appears to be extensions on their end,
wherein restrictions might need to be enforced.)

Follow-up question: did we decide on rdfs:Proposition, or is that or
rdf:Proposition an open question? (I recall a discussion in December, but
not the closing argument. I have no strong arguments, but e.g. [1] needs to
track this.)

Cheers,
Niklas

[1]: <https://github.com/w3c/rdf-schema/pull/31>



>
>
> —e.
>
>