- From: Holger Knublauch <holger@topquadrant.com>
- Date: Wed, 24 Feb 2016 11:56:11 +1000
- To: public-data-shapes-wg@w3.org
On 19/02/2016 12:32, Peter F. Patel-Schneider wrote: > And neither of the special cases for sh:Class appear to hold for > sh:directType, although that is contrary to the first sentence in 3.1.5. > > The wording for sh:classIn is also confused. Yes, we will clean this up once we have a resolution on what the semantics of sh:class should be. Thanks for pointing this out. The viable alternative seems to be to a) Limit sh:class to "real" rdf:type triples b) Add a warning that sh:class rdf:List may lead to surprises c) Add more sh:NodeKind instances d) Probably add something like sh:memberShape as an idiom for list-typed properties Related to this: Am I the only person wondering whether we made a wrong turn with all these sh:class, sh:datatype, sh:classIn, sh:datatypeIn, sh:directType properties? How come that other languages such as OWL and RDFS just have a single property (owl:allValuesFrom, rdfs:range). We could just have a single property such as sh:type and distinguish the various use cases based on the types of values that it gets. Our current design also makes it impossible to define mixed properties that can take either strings or objects, such as those found in Dublin Core and http://schema.org/address. I am afraid these things will come back to haunt us when exposed to more users. Holger
Received on Wednesday, 24 February 2016 01:56:45 UTC