Re: "shape" as a relationship, not a class

I'm trying to understand how the subject of a triple can be either of 
type shape or can have a shape.

Let me make this more concrete. My subject is a book, which is a RWO, 
and it has an IRI. I'm going to make various statements about this RWO 
(it has a title , it has an author, etc.). It makes little sense to me 
to say that this RWO "has a shape/graph." The graph has a shape, but 
using the same IRI to represent the RWO and the graph violates a basic 
rule that each IRI references one and only one "thing."

It seems to me that the key difference between shapes and classes is 
exactly this: a shape is information about a graph; a class is 
information about the RWO. If the class of the RWO  is coincident with 
the graph that you wish to validate, then presumably the class can be 
used as a target for validation. However, that is making a use of the 
class which is not within the definition of class in RDF. I would find 
it inconsistent with RDF for us to encourage people to assign classes to 
RWO's that represent the graph itself.

kc

p.s. I continue to ponder the "validation vs. semantics" but I'm at all 
sure that I can gain sufficient clarity to be articulate about it.

On 2/5/15 2:48 PM, Holger Knublauch wrote:
> Another angle on the important classes/shapes discussion. To me, the
> term shape does not necessarily describe an entity, but is better used
> as a relationship:
>
>      :hasShape(?resource, ?class) : boolean
>
> or as a "magic property":
>
>      ?resource :hasShape ?class .
>
> *A resource has the shape of a class, if it fulfills all constraints
> attached to that class (regardless of its rdf:type triples)*. Classes
> are not only used to assert set membership (via rdf:type), but also to
> declare hypothetical sets of resources with similar characteristics.
> That's similar to how OWL uses owl:Classes.
>
> I think that terminology would still allow us to talk about shapes with
> proper meaning, without having to duplicate and reinvent established
> concepts.
>
> Holger
>
>
>

-- 
Karen Coyle
kcoyle@kcoyle.net http://kcoyle.net
m: 1-510-435-8234
skype: kcoylenet/+1-510-984-3600

Received on Sunday, 8 February 2015 18:37:07 UTC