With ArcGIS Diagrammer being an unsupported Esri research project, GDB XRay being an unsupported tool to round-trip geodatabase design using Excel, and an uncertain future with Visio support for geodatabase design (not to mention being locked into a specific tool), Esri should reconsider developing a UML Profile of the Geodatabase. I've posted an entry to ideas.esri.com that you can vote on if you agree.
For those who don't know, http://ideas.arcgis.com is a newish site where you can propose and vote on new features for Esri products. Seeing lots of good stuff there that I've personally pined for in the Geodatabase and ArcSDE area (one I've championed in the past is getting GDBT to be core--and that seems to be getting traction). Glad to see Esri has provided a quantitative process for listening to and implementing customers' feature wish lists.
The "ideas" site is a great improvement over trying to petition your local sales rep, commiserating with colleagues in person or on the forums, and other suboptimal approaches to getting Esri to consider new features. So, regardless of whether you like the UML Profile idea (it may well be a stinker), please participate in the ideas.arcgis.com process. Surely it's not a true democracy (and the customer may not always be right), but I applaud the introduction of a more quantitative, centralized and systematic approach to dispositioning new ideas vs. a qualitative, decentralized, anecdotal one.
Here is a link to the UML Profile of the Geodatabase idea:
Anyway, since the XML Schema of the Geodatabase can exhaustively represent all geodatabase features--and one proof is ArcCatalog being able to import and export XML Workspace Documents--I believe it's possible in principle to develop a UML Profile of the Geodatabase. With a UML Profile of the Geodatabase, you could load that into the UML design tool of your choice and perform modern geodatabase design. Enterprise Architect (EA) from Sparx Systems, incidentally, is an inexpensive (around USD two hundred dollars per license) and seems to be used by Esri's Professional Services division (at least for application design on some consulting engagements):
You'd have the ability, at least with EA, to trace back GDB features to requirements and Use Cases. It would be immensely helpful to have a way to communicate geodatabase designs to non-GIS folks throughout the Enterprise--people like DBAs, Enterprise Architects, and others wanting to integrate with GIS.
Although GIS is maturing and becoming more consonant with Enterprise IT, geodatabase design has a bit of a way to go. I believe a UML Profile could be a good potential direction. Using ArcCatalog is too ad hoc to create and communicate Very Large Geodatabase designs and the other tools out there are incomplete. If it's not a UML Profile of the Geodatabase that improves this condition, that's fine. Just hope it's something that elevates geodatabase design closer to where relational database design is today. Hoping that by having posted an idea to ideas.arcgis.com it will be appropriately dispositioned--e.g. put to rest due to lack of interest from the community or elevated to "under consideration" if enough folks vote for it. Again, UML for Geodatabase Design may not be "the" answer or even "a" viable answer. But I feel its something that ought to reviewed seriously in relation to potential and current options available for geodatabase design; design in a fashion consonant with other modern design and documentation methods for data models and data modeling.
Something, whether it be a UML Profile or not, should better satisfy the need to mature the geodatabase design process. I understand Esri has attempted to provide a UML modeling capability with Visio, but that the approach is based on ArcObjects. It may be better to start fresh and create a UML Profile modeled with a view of the geodatabase and its features as things in themselves (non-GIS staff, and even some GIS staff, aren't necessarily going to understand or care about ArcObjects).
For reference, here's an older thread on using UML for geodatabase design:
Does Esri have any specific plans for maturing geodatabase design? Especially for VLGDBs (Very Large Geodatabases)--large with respect to the number of objects--that need to integrate with the Enterprise.
Seems there's no sense of urgency to move GDB design toward parity with modern relational database design. If there's insufficient customer demand, that's one explanation--it's tough to deploy resources toward something few people explicitly say they want. If there is demand, it's probably challenging to provide a solution that would satisfy the largest possible base of geodatabase modelers.
I'm open and interested in any good solution that steers the geodatabase modeling and design community away from ad hoc work, disconnected from requirements and the larger sphere of Enterprise IT. Doesn't have to be UML-based, but what are the alternatives?
Mostly, I'd like to see open and engaged conversation on this issue wherever it ultimately leads.