Interesting paper comparing approaches to upper ontologies from the MITRE Corporation, via the Ontolog forum. The authors' focus is supposedly on US Government applications, but I think that the conclusions or analysis would apply to any large ontology-based information system. Particularly useful high-level summary of different modelling choices in building an upper ontology, though some of it assumes familiarity with the terminology (e.g. 3D vs 4D ontologies are mentioned but not explained). Paper abstract and pdf version.