All - I know that at this point Alma and Stevan would expect me to point out that - as well as OA IRs - there are other systems maintained by funders and research institutions.
These are called CRIS (Current Research Information Systems)and there is an EU Recommendation to member states (i.e. a standard) named CERIF (Common European Research Information Format) which has formal syntax and defined semantics and is thus ideal for interoperation even in a multilingual environment. Since researchers move quite a lot and may have multiple (simultaneous or sequential) affiliations to research organisations, funders etc the data model has to have timestamped role-based relationships between major entities like persons, institutions, publications etc. This is much more expressive than DC-type metadata. CERIF also has the necessary attributes to generate appropriate publication metadata in Dublin Core, MARC and the various bibliographic reference standards (like APA, Vancouver, Chicago, BibTex etc). Finally, CERIF also provides context for the research i.e. research project information and what (if any) facilities and equipment used, events attended etc. Details at www.eurocris.org/cerif So instead of trying to add (non formal (or semi-sructured)syntax and variable semantics, DC-type) metadata into IRs I recommend strongly using the data structure of CERIF and link it to the full-text (multimedia) in the IR. Of course the same technique works for research datasets and software etc etc Best Keith