As far as ACS 4.x goes, since aD is ditching it (over time, but in terms of new development ASAP), OpenACS has already in essence forked the code base. OpenACS 4.x will support both Oracle and PG, in fact getting the infrastructure in place to do so is what's been taking up Ben and my time the past two weeks.
We're almost there though it's been frustrating - I spent most of today rebuilding Oracle after it corrupted itself on my laptop, grrr...
As far as ACS 5 goes, aD has backed off the approach of putting much of the API in PL/SQL, specifically in order to make it possible to support multiple dbs. In other words, Uday's right. At least, that was the plan as communicated to Ben and I in January and AFAIK nothing has changed.
Ironically, we can deal with the API-in-PL/SQL approach fairly easily in PG, since it has a nice little PL/pgSQL that has much of the same functionality as PL/SQL (no packages, no default params and only positional params but we can handle most PL/SQL semantics no sweat).
This isn't true in certain other RDBMS engines, though, so you can presume that the goal is to make it easier to support any SQL92-ish RDBMS that has decent market share.
