In also conservative on "normalising" dwc too much and replace all id
terms
used as foreign keys with the relations class. Of course we can model
things
that way (though im sure we'll start seeing more qualifiers on that class then), but dwc is largely successful because of its rather flat and simple nature, so we should be careful to preserve at least the important and
often
used id terms.
But DwC is already normalized this way. There is no LocalityID term in the Event Class. There is no EventID in the Occurrence class. In fact, the only "foreign key" term I know of in DwC (internet too slow here for me to consult it online) is "IndividualID", and this doesn't even serve the function even a "Foreign Key" because there is not yet any "indivudal" class to provide "PK" of the corresponding linked instance.
So...I've been asking various people how one is supposed to represent the fact that "This Event instance took place at this Locality instance", or "This Identification instance refers to this Taxon instance", or "This Occurrence instance involves this Event instance and this Identified Taxon instance". So far, I don't know the answer to this. For the BiSciCol project, these relationships between instances of DwC classes are the core item being indexed, so we are assuming that the purpose of ResourceRelationship is, in fact, to establish these relationships between instances of DwC classes.
I hope someone will correct me if I'm wrong.
Rich