You'll want to use the APM because the bootstrap installer now uses that to determine which datamodel to load, which query files to load, etc.
This doesn't mean you have to support both Oracle and Postgres datamodels in any packages you write - it is perfectly reasonable to write a package that only contains an Oracle datamodel, and generic SQL92+(when necessary)Oracle-specific queryfiles.
Or PostgreSQL-only, of course.
If you write within the framework and if the code can be GPL'd, then others can port any new packages you write to other RDBMS systems much more easily if you follow the framework from the beginning. And working within the framework's easy.
Regardless we highly recommend you follow the new queryfile structure. If you have existing packages you can run the query extractor, stuff your sql/*.sql files into sql/oracle/*.sql, run the APM on it to build a new .info file, and be up and running under the OpenACS 4 scheme in a half day at most (more if you need to ask us questions, etc).
Non-profit support is one of my personal driving motivations for being a part of the OpenACS project. I'm not alone in that. You can be certain that the community will be willing to help you out with any questions you might have (though I'll personally be unavailable from June 6th-June 29th).
