Proposed change to rdfs:domain of sh:declare and rdfs:range of sh:prefixes

Dear SHACL CG,

there has been some discussion on the data shapes git repo, and as the 
official successor of the data shapes WG I would like to make this CG 
aware of a proposed change to the shacl.ttl file. The changes would 
eventually go "live" in the SHACL namespace document hosted at 
http://www.w3.org/ns/shacl#

To cut a long story short, there was divergence between the prose of the 
spec and the TTL document. The intention has always been that sh:declare 
SHOULD (but does not have to) be used at the named graph resource (which 
is often an owl:Ontology), and that sh:prefixes SHOULD (but does not 
have to) point at such a named graph resource. However, I probably was 
too eager using rdfs:domain and rdfs:range statements to back up those 
recommendations, to aid form builders and other tools place those 
properties. However, these domain/range statements cause undesired RDFS 
inferences, so it is better to take them out.

It was suggested to switch to schema:rangeIncludes and 
schema:domainIncludes, which are semantically close but safer. I don't 
have a super strong opinion but I would prefer to avoid bringing in 
unnecessary references to external namespaces, esp if they are outside 
of the base standards such as RDF and RDFS. If we would go down that 
route then I guess we should also add similar triples to accompany our 
other usages of rdfs:range and rdfs:domain. This would open a can of 
worms for rather small benefits.

If anyone has concerns about this pull request, please comment on the PR 
itself:

     https://github.com/w3c/data-shapes/pull/133

Unless I hear concerns I would merge this into 'master' by the end of 
the week.

Thanks

Holger

Received on Monday, 18 January 2021 23:39:32 UTC