- From: Pat Hayes <phayes@ihmc.us>
- Date: Mon, 12 Dec 2005 11:58:32 -0600
- To: "Christoffel Dhaen" <christoffel@landcglobal.com>
- Cc: <public-swbp-wg@w3.org>
>Hi all, > >I realize I should never have mentioned the >domain-range, and never should have used >"inferred". I understand why Gary, and I imagine >a lot of other people, got very confused when I >brought this up. This is beyond the scope of the >query language since "new" information is >"added" to the graph and would require a >reasoner or similar mechanism. This was never my >intention. > >All I wanted to point out was transitive >closure. And has been noted by E.Franconi: " you >need a full fledged "graph" query language, >while I guess that SPARQLl suffers of a mixed >origin: a bit of SQL/relational data model and a >bit of graph data model." > >Thatıs all. RDF is a graph, it has native transitive properties. That phrase does not make sense to me. Can you please explain what you mean by 'native transitive'? RDF uses a graph *syntax*, but that is merely a syntactic convention. Nothing follows from this about transitivity or otherwise of the relations described by an RDF graph. >Being able to express in query that only the >direct nodes should be taken into account, or >that the transitive nature of the properties has >to be taken into account is a logical step. What is a 'direct node'? RDF does not require *any* properties to be transitive. RDFS requires rdfs:subClassOf and rdfs:subPropertyOf to be transitive, but nothing else. Are you referring to queries that assume RDFS completion of the query graph, i.e. that it is closed under RDFS entailment? > No more, no less. >The complexity of introducing this: minimal. I >only mentioned it because it has some >implications, it won't implement itself. I just >got the impression that there was a concern that >implementing this would be far to complex. It >isn't. >Hierarchical queries do something similar, and >their implementations and syntax vary. >And thatıs exactly what I meant by >"non-standard": some implementations will add >their own keywords to SPARQL to add this >feature. Others will opt to define the indirect >variant of a transitive property, like >ns1:indirectSubclassOf for rdfs:subClassOf, or >as has already been suggested, >skos:inNarrowerClosure for skos:narrower. >Namespaces will vary, keywords will vary, >implementations will vary, and interoperability >between systems will be lost. It is far from >unwarranted. > >The /*+RULE*/ example was not a great one, but >the comparison to hierarchical queries still >holds: For a standard database there was already >a need to create hierarchical queries. Rdf has a >hierarchy in classes and properties because the >property is transitive. RDF makes no transitivity assumptions at all. RDFS makes it for subClass and subProperty. Are you talking about RDF or RDFS? Pat Hayes -- --------------------------------------------------------------------- IHMC (850)434 8903 or (650)494 3973 home 40 South Alcaniz St. (850)202 4416 office Pensacola (850)202 4440 fax FL 32502 (850)291 0667 cell phayesAT-SIGNihmc.us http://www.ihmc.us/users/phayes
Received on Monday, 12 December 2005 17:58:41 UTC