Re: Global property constraints (was: Re: Stand-alone Shapes and oslc:valueRange implemented in SPIN)

On 12/4/14, 7:40 PM, Dimitris Kontokostas wrote:
> Taking this a little further, we should be able to define global (on a 
> graph level) property constraints without the need to assign them to a 
> Class / Shape.
> Here's an example on how Wikidata does something similar.
> https://www.wikidata.org/wiki/Property_talk:P1047
> https://www.wikidata.org/wiki/Property_talk:P640
> https://www.wikidata.org/wiki/Property_talk:P1212

I cannot follow this line of reasoning. All examples above are 
properties that have a domain, i.e. they are assumed to be applied to 
instances of a specific class only. Therefore, why wouldn't it be 
possible to attach those constraints to the domain class, similar to how 
owl:Restriction and oslc:Property does it? In my original comment to 
Eric's example, I was also stating that the datatype could be attached 
to the Item class, not globally on the property.

My general opinion is that global property axioms such as rdfs:range, 
domain, and owl:FunctionalProperty are overrated and better avoided. No 
other mainstream technology seems to have such a concept of global 
properties. Even solutions like schema:domainIncludes and rangeIncludes 
are problematic, e.g. what happens if each have two values - which 
combinations are supported? If constraints are attached to classes, this 
is clearer.

Thanks
Holger

Received on Thursday, 4 December 2014 11:13:48 UTC