On 2/17/2015 2:04, Corey A Harper wrote:
> In Karen's initial message, it was suggested that some members see
> "class membership as the primary trigger for shapes". This seems
> reasonable. But the more absolute notion that class membership is the
> only trigger or that classes and shapes are the same construct seems
> to unnecessarily narrow the scope in which ldom will be useful.
I think Corey has summarized it nicely.
Thanks to everyone on this thread for their input. I don't want to
respond to each issue individually, but will use your input for the WG.
It is difficult to explain all the nuances of the ongoing discussion of
the WG here, so several misunderstandings are floating around. Let me
just make clear that having classes does by no means mean that rdf:type
triples are the only/best/primary way of checking instances. It is of
course perfectly possible to ask the system
"does X fulfill all constraints attached to class Y"
and any application can chose which combination of X and Y to use,
regardless of rdf:type triples. Then we have a notion of ldom:context
that allows people to fine tune whether they want to follow the default
definitions from an ontology publisher, or rather ignore them and
replace them with constraints from another context. All this will
hopefully be part of the standard and become clear in due time.
Yet, there used to be a notion of a Semantic Web, in which people were
able to publish ontologies together with shared semantics. On this list
and also the WG it seems that this has come out of fashion, and everyone
seems "obsessed" with the ability to violate the published semantics.
Yes this ability is important, but I doubt that it should limit the way
that people can publish the default semantics of a model on the Semantic
Web. Plus there are many applications in which ontologies are
custom-built anyway, and having a class-based model is simply the
best-established practice for those use cases. Why reinvent the wheel?
Thanks,
Holger
