Data sets should be extensible or flexible, similar to the characteristics of XBRL or RDF/OWL, rather than fixed schemas. This flexibility allows the data sets to evolve, it allows others to connect additional information to existing information. The notion of "linked data" as used by the Semantic Web people.
Both XBRL and RDF/OWL are modeled as graphs which are extremely flexible. Combine these graphs with the multidimensional model, all the data linked together with some sort of pivot table type interface is very flexible and powerful.