Re: shapes and classes: different

On 2/4/15 12:57 AM, Karen Coyle wrote:
> Holger, the problem that I see in your examples is that there is way 
> too much semantics.

There is a wide spectrum of use cases - some with very little and others 
with far more semantics than the current example. We should capture more 
such examples, and I am looking forward to your input on that.

> Also, you have assigned an rdf:type to each graph, thus the graphs are 
> 1-2-1 with classes. We have a vocabulary that has no classes defined, 
> and the IRIs are opaque. It looks like:
>
> ex:ResourceA
>    rdau:P60367 "The adventures of Tom Sawyer" ;
>    rdau:P60073 "1996" ;
>    rdau:P60093 dctype:text ;
>    rdau:P60434 <http://id.loc.gov/authorities/names/n79021164> ;
>    rdau:P20006 ex:ResourceC .

In the absence of either rdf:type or ldom:shape/oslc:instanceShape 
triples, how would a constraint checking engine know what it needs to 
do? If these are all "global" constraints, then we are in the realm of 
ldom:GlobalConstraint, which seems orthogonal to the question of classes 
vs shapes.

>
> That's not quite a valid example, but it will take a while to create 
> something meaningful, and I would probably need help to create 
> something that actually tests the use case of graphs vs. classes. But 
> I do think that the example is already weighted toward the class 
> decision.

"My" example is indeed a slight variation of a scenario introduced by 
the ShEx authors, so I assume it is not biased towards classes ;)

Thanks
Holger

Received on Tuesday, 3 February 2015 20:50:45 UTC