I just got a bunch of tests working for my Python ORM for various databases. Some stats just for fun, running the same 5 tests in x seconds. The fifth test spawns 99 threads and runs them in parallel. The rest are all simple serialized reads and writes. The times don't include creating or destroying the DB. In ascending order:
Python shelve (serialization to files): 1.102s
MySQL 4.1: 3.205s
SQLite 3: 4.116s
PostgreSQL 8: 9.633s
MS SQL Server (MSDE): 10.785s
ODBC (to MSAccess): 13.910s
MSAccess 2k: 19.338s
Just thought I'd share. :) The next step will be to make tools to do these same comparisons against *my* data, not example data. Part of why I'm writing my own ORM is to make that kind of comparison easier.