Well I mentioned above that I didn't really like tying it into the acs-subsite package-proc code ... but, that's where all the underlying pieces it uses live, so for now at least it feels like the right place.
db_* should remain a low-level interface to the database driver, I think. This is not a low-level interface and in the PG case depends on explicit metadata (including the defining of default param values) that aren't part of PG.
