I should add my 2 cents here.
I think it is my role (when I can get the documentation thing out the way) to pull together an overview of the whole TDWG 'thing'. This will be part of the architecture along with a vision of where we should be going. A glossary of terms is a start but their needs to be more than that. We need a Bush (senior) "Vision Thing" here.
I imagine the vision will crystallize out after the GUID meeting at which point lots of people will say "it will never work" and we will have to change things or build demo systems or whatever. At the moment it looks likely (as Donald says) that we will need to move toward an RDF/OWL way of doing things but I don't see this as being based on big triple stores as I don't think all services will give the ability to search on all possible assertions (I have not had a chance to look at the logs I received properly but it doesn't look to me that people want to do complex searches and we should perhaps have clear test cases before we engineer a solution to support arbitrary requests rather than basing it on what we thing people might want to do). What is clear is that people want to extend and mix and match schemas. ABCD has been added to repeatedly and DarwinCore has been forked. Also we don't make use of metadata standards like DublinCore of vCard which means that engineering generic solutions is likely to be difficult. RDF/OWL type approach definitely solves these kinds of problem. Even if we still exchange templated chunks of XML those chunks can also be RDF compliant, served through Tapir and reference a central OWL ontology. Everything changes but things remain the same.
Hope this helps,
Roger