The values of sh:argument (now called sh:parameter in the spec and the
upcoming 5.2) must be IRIs or bnodes that have a sh:predicate pointing
at a IRI. The properties don't need to be typed at all. I am not aware
of any other requirement here, but the usual suspects such as
sh:datatype and
If I use the 'create argument' option for sh:argument and sh:sparql, are
there any requirements that this definition must satisfy (other than have a
legal namespace)? Does the argument have to be of any particular type
(other than rdf:property)? Thank you.
Jack Hodges
Siemens WoS
--
You
Thanks Holger.
I think a solution could be to have 2 sets of arguments: template-local (by
default), and inheritable (defined so explicitly), or "tunneled".
Subclasses would not see the local arguments of superclasses, and it would
be an error to redefine tunneled arguments with a different
