Eve's already been mentioned here, and Chris' lecture at Cal was really interesting from a UX perspective. I'm sure you've seen it but if not, go watch it. The

Eve's already been mentioned here, and Chris' lecture at Cal was really interesting from a UX perspective. I'm sure you've seen it but if not, go watch it. The spreadsheet model is fairly intuitive for the end-user. He and Rich Hickey both are pretty convinced that a Datalog-esque query engine is the way to go. 20 years ago, Lotus' Improv's model (where you separated the data manipulation from the data type constraints) was way ahead of its time and guaranteed so much more internal/referential integrity than 1-2-3. Modern-day Quantrix is probably the closest you'll get to it, used pretty heavily in niche finance. MSR puts out some interesting work on the PLT behind alternate spreadsheet grammars[1] which are quite interesting too. Other than the obvious issues (ETL[2], ACLs/RBAC, presentation/extensibility) MSR has put out quite a few interesting papers out there. This is a very interesting problem domain to say the least. I've developed up-to-the-Pareto-point for this in the past for clients and if you get it right, it's a license to print money. [1]http://research.microsoft.com/en-us/um/people/sumitg/pubs/pldi11-table-synthesis.pdf http://research.microsoft.com/en-us/um/people/sumitg/pubs/pl... [2][both conventional a la SSIS and Informatica, as well as a Pipes-esque query; a premium feature you could easily charge into-the-Oracle-rates would be a Pipes-esque service with an SLA. I've been in enterprise consulting long enough to know that whether you're Ingersoll Rand, Haas, an REIT or Cravath, sanitized, guaranteed data is worth a lot, integration is worth even more.