It would be terrific to have a UML Profile of the Geodatabase for Geodatabase Design. Then one could use the UML tool of one's choice for Geodatabase Design (Enterprise Architect, etc) and not be locked into using Visio. Diagrammer is great, but it's an unsupported research project; and it can't be used to trace GDB items back to requirements and use cases.
Since the XML Schema of the Geodatabase can exhaustively describe everything in a geodatabase, a UML Profile of the Geodatabase should be possible in principle. Translators / translation code may be an issue. But since ArcCatalog can import and export XML Workspace Documents, I don't see why a UML Profile couldn't be developed that, if used, could generate valid XML Workspace Documents and thus be loaded into ArcCatalog for geodatabase instantiation. We really need something better for geodatabase design than what we presently have.
It may ultimately be up to the user community to develop a UML Profile of the Geodatabase. UML may be a bit too obscure / underadopted by the geodatabase design community for Esri to commit the resources to this. However, I agree with commentators in other places who bemoan the inability to use good CASE tools such as those that exist for relational database design--and the difficulty in communicating geodatabases with the wider community of non-GIS IT folks like DBAs, Enterprise Architects, etc. A UML Profile of the Geodatabase, along with a way to provide full round-trip design, would go a long way toward integrating GIS into the broader bucket of Enterprise IT and beyond. GIS is growing up. Would love to see geodatabase design mature in step with this trend.