"Christopher Petrilli" wrote After some talking, I'd say our position vis a vis ODBC on UNIX is that it's simply not a good solution for application servers. ODBC real value lies in the client side, and on Windows NT, where often it is the easiest (or only) solution. On UNIX, all database vendors ship native libraries (except Microsoft, but don't go there ;-), which are substantially better choices. Some "swag" looks, ODBC for Oracle, for example, is 1/10th the performance level of the native OCI level interface, so why would you use this in your application server?
The other issue is simply that unless the vendor supplies an ODBC driver layer for your platform, you're still screwed. The only thing ODBC helps with is that you won't need to write a python DB-SIG compliant interface to a DB for which you have an ODBC driver. If you have some wierdo database that only provides an ODBC interface, then it will be useful - otherwise? meh. With the exception of DB2, almost every database I can think of has a mostly-portable DB-SIG compliant interface now. Well, ok, I think the msql one might still be a tad weird, but hey, unless someone's done something drastic, there's parts of that code that are 4 or 5 years old. Until the mxODBC license changes, that's also possibly an issue for people, too... Anthony.