Re: Domain/Range: conjuntion or disjuntion??

>   Thanks Mr. Kampman, I was reading the document you sent, in concrete the
> chapter for, and I understand what you explain there, but you can fix it in
> other way, don't you?
> 
>   A -1-> a, B -2-> b
> 
>   (note: Subject -Predicate-> Object)
> 
> and property 3 is subproperty of 1 and 2 (first of all, I never understood
> subproperties, I would delete it from the schema)
>   3 will be a property that expect a instance of A and a instance of B in
> the domain and a instance of a and b on the range, you can do something
> like this.
> 
>   C -subClassOf-> A, C -subClassOf-> B
>   c -subClassOf-> a, c -subClassOf-> b
> 
>   3 -domain-> C, 3 -range-> c
> 
>   You have more sentences but you have exactly the same, and you can add
> that 
>   3 -subProperty-> 1, 3 -subProperty->2 
>   without any interpretation, without inherit ranges and domains.

You can't just ignore inheritance of domain- and range restrictions. Domains
and ranges of subproperties should always be at least as restricted as their
superproperties because of section 2.3.3 of the RDF Schema spec:

"If some property P2 is a subPropertyOf another more general property P1, and
if a resource A has a P2 property with a value B, this implies that the
resource A also has a P1 property with value B."

Assume P1 has its domain restricted to D1, and P2 does not have any domain
restrictions. This would allow you to use P2 with resources that are not
instances of D1. But as P2 is a subPropertyOf P1, D1 will also have property
P1. This again implies that the resource must be an instance of D1.

> 
>   Finally, I want to repeat, this is a solution for Sesame a RDF based
> product but not for all RDF based products, RDF should be independent of
> concrete product necessities.

I cannot agree with this more, but all of these concrete products should
adhere to one single interpretation to assure interoperability. Sesame
deviates from the original union-semantics because that can lead to
inconsistencies in the model, which is a no-go for all RDF based products.

Regards,

Arjohn

Received on Wednesday, 21 November 2001 06:18:55 UTC