comments inline
John Wieczorek wrote:
I am fine with the idea that the CollectionObject can be evidence for anything, I may have been projecting or misinterpreting what others said. Can you recommend a better definition than the one I provided?
I don't think that the definition of CollectionObject which you gave needs to be changed. Being derived from an Organism is the minimal requirement for a collection object. If it can do other things (be evidence for an Occurrence, support an Identification, or anything else) that's great but not required. If it has a collection date, location of collection, or both, which CollectionObjects usually do, then it would be evidence for an Occurrence as well as evidence that the organism existed.
Richard Pyle wrote:
One concern I do have, however, is in the many, many, many cases where there will be a 1:1 relationship between an Organism and a CollectionObject (i.e., the vast majority of all Museum specimens). Does that mean that data providers will need to generate two separate Ids (one organismID and one collectionObjectID) to represent all of these specimens?
Shockingly, I think I agree with everything Rich wrote in response to Gregor's email. As far as this question is concerned, I would say that whether one generates a single ID or two IDs is up to the user. Assuming that one has GUIDs of some form (LSID, HTTP URI, or whatever), then if one wishes to consider the dead fish in the jar both the Organism and the CollectionObject, then use a single GUID for both organismID and collectionObjectID. If one prefers to think of the fish as a Platonic ideal of Organism in all of its incarnations (living and dead) but the fish in a jar as the CollectionObject, then give them separate GUIDs. That really is a data management decision by the GUID creator. I was hung up on this issue for a very long time, but after mulling it over I realized that I was only hung up about it because I assumed that all classes in DwC had to be disjoint (sensu OWL). They almost always are, but in this case I don't see any reason why they would have to be. DwC simply uses classes to categorize things and to suggest the types of terms that one might use to describe instances of those classes, but otherwise stays out of our personal data management lives.
Richard Pyle wrote:
I think I might agree with this, but I want to ask a simple question:
To what objects would an Identification instance apply? In other words, an Identification instance represents a link between an instance of Taxon to an instance of [XXXXXXX].
In my mind, this should always be "Organism". To me, neither an Occurrence
Totally. Is this the first time I've agreed with everything Rich has said? :-) Steve