There are some interesting parallels with this project, oriented to
compound objects, not dissimilar from many datasets, and UMBELs
objectives. Look at their recent white paper --
http://www.openarchives.org/ore/documents/CompoundObjects-200705.html-- and just think of UMBEL every time you encounter the phrase
"compound object".
This paper is very well written and helps explain broader issues of
the Web and interoperability as well.
Mike