I strongly disagree.  Domain and Range are not validation features in RDF, they allow additional inferencing only.

By enumerating all of the domains and ranges, the ontology is led into the trap of defining class specific predicates when a more general predicate would have sufficed.  Just look at all of the unnecessary (foo)Assigner predicates that have identical semantics, just to have the domain specified.

I would argue that once the predicates have been deproliferated, the domains and ranges will be less necessary than they are now.  Conversely the documentation describing the usage of the predicates will need to be improved, and clear examples provided.


On Wed, Apr 1, 2015 at 6:21 PM, Joseph Kiegel <[log in to unmask]> wrote:
In regard to the domains of BIBFRAME properties, the BIBFRAME Vocabulary: Terminology and Conventions states:

"Domains and Ranges may be restricted to one BIBFRAME Core Class, i.e. Work, Instance, Authority or Annotation.  If applicable to all Core Classes they have domain Resource [i.e. bf:Resource]; if applicable to more than one but not all they have no domain specified".

In practical work with BIBFRAME, it is problematic that domains are not specified when a property is restricted to a set of core classes, that is, more than one but not all.  Restrictions exist, but when they are not stated explicitly, implementers cannot tell what they are.

For example, bf:titleVariation does not have a specified domain. Presumably, it can be used with both bf:Work and bf:Instance.  But the fact that we have to make a presumption is not good practice.

bf:musicKey, bf:musicMedium, bf:musicMediumNote and bf:musicNumber contain information that, in the WEMI model, belongs in rdac:Work.  Their domains are not specified, so we must presume they are included in bf:Work.  Perhaps these properties also apply to bf:Instance.  It is unclear what is intended.

A puzzling case is bf:immediateAcquisition.  Conceptually, this property seems as though it can apply only to bf:Instance, yet the domain is not specified, so it must apply to another core class as well, but which one?

A complete specification of the domains of BIBFRAME properties would add clarity to the model.

Rob Sanderson
Information Standards Advocate
Digital Library Systems and Services
Stanford, CA 94305