I'm building a large system on my local box (MySQL) which will be deployed on an enterprise server (Oracle). I'll be using binds and [C|B]LOBs. Ideally I'd like to write a query once, and let the abstraction layer figure out date formats, LOB handling, &etc.
I've found three promising classes for handling database abstraction:
Metabase -- Looks bulky and reports are that it's not optimized for speed. However, handles SQL translation and is written by Manuel Lemos.
Pear -- Tiny featureset. However, authored by PHP core team and will probably be heavily supported in future (if not compiled in). This doesn't emulate Oracle funtionality in MySQL, as Metabase does.
ADOdb -- Seems fast, but the benchmarks were in question. Does some SQL translation.
Does anybody have any recommendations, or have experience developing on a system which can alternate between two RDBMS? The most important factor is Speed and then abstraction capabilities, in that order.
-- flash
