The only problem is that each DB has a different subset of SQL that is available.

I think the problem is that each DB has a different superset of SQL. SQL is the Cobol of data-access languages: its universality has given it a momentum out of all proportion to how well it actually solves the problem.

If anyone has any candidates for the ML, or even the C, of data-access languages, let me know. (Python plus a good DB abstraction, or C++ plus Microsoft's DB abstraction du jour might be close to the mark.)

Peter