- From: Niklas Lindström <lindstream@gmail.com>
- Date: Thu, 30 Nov 2023 15:53:21 +0100
- To: Andy Seaborne <andy@apache.org>
- Cc: public-rdf-star-wg@w3.org
Hi Andy, On Thu, Nov 30, 2023 at 1:57 PM Andy Seaborne <andy@apache.org> wrote: > > > > On 27/11/2023 21:37, Thomas Lörtsch wrote: > > I think it has been established that most use cases can be characterized > > as annotating tokens, not types (if that still isn’t common wisdom then > > that would be a very important discussion to have first). > > For proposals that do not also include types in the data model - > > Can there be two different tokens of the same type in a single graph? How? Yes. How depends on what the token is. If it's just old-school reification; two rdf:Statements with different identifiers (bnodes or IRIs) who both have the same values for rdf:subject, rdf:predicate and rdf:object (and exactly one of each). The "type" here is thus quite informal (but understandable, I believe). If it's based on named graphs with semantics (like option 3.4 in RDF datasets [1]), yes-and-no, since only the "name tokens" referring to those are in a single graph, and the type is based on the equality of the two other graphs that these tokens "name". And this equality depends on more than their "inherent" triples I think, specifically which entailments they're under and whether other graphs are "visible" (determined via relationships from/to the graph name; akin to owl:imports). This is certainly not formally good enough without semantics for datasets. (And I just posted an attempt which I hope is less far-reaching and closer to what you are thinking of.) > Conversely, can multiple triples refer to the same token > with requiring specific syntax (c.f Turtle ";") Yes, by using the same identifier (blank node or IRI). > With type and a triple to relate token and type, both are possible and > it is preserved on RDF merge. Yes. I am interested in the graph terms idea you have in this regard. Would two sets of triples with a graph term identifier in the fourth position merge in a redundant way (e.g. resulting in the same triple but with different graph term identifiers), or would there be a structural graph identity check and some relabelling? (I guess no, and that this would be left to a more expensive *union* operation, possibly relying on RDF C14N?) Best regards, Niklas [1]: <https://www.w3.org/TR/rdf11-datasets/#each-named-graph-defines-its-own-context> > Andy >
Received on Thursday, 30 November 2023 14:53:53 UTC