If I understand this correctly, it assumes that there would be a general "property" entity. An instance of this entity would then be the property "shape", which has a list of generally applicable values. Some of these values would apply to leaves as well as to butterfly wings, other may be more specific.
Actually, what I'd like to see as a requirement is more modest than this. I simply propose that there should be the possibility of explicitly recording a property name, which need not have any global significance.
Thats actually what I was suggesting (albeit probably not that clearly). i.e <property name="shape"> or similar. "shape" may have meaning within a particular domain. If possible that meaning ought to be standardised. If it can't, then the loss is in interoperability between groups in that domain.
Inter-domain exchange of information is a much more difficult nut to crack.
The property could be expressed in many ways, but for the sake of an example I suggest that we might possibly have a DELTA character that looked like this:
#2. body <@property degree of convexity>/ 1. strongly flattened/ 2. slightly flattened to moderately convex/ 3. strongly convex/
Ideally both the body character, its property, and the values would be standardised - but thats a different issue, and one that is orthogonal to the design of the information model.
As Jim Croft said, the information model is the important point, and the first 'deliverable'. After that, when people actually start using the model, standardisation can be explored if necessary. But that process shouldn't stop the model from being derived in the first place.
Cheers,
L.