Tom,
> Yeah. If the schema-per-module answer were really a good answer,
> we'd have done it before now. But you need more infrastructure
> than just a schema to get good things to happen. Aside from the
> search-path-hell issue, a schema alone doesn't solve the problem
> of persuading pg_dump to dump a "load module" command rather than
> the individual module components.
I have faith in nothing historical regarding /contrib, which has been
largely characterized by maintenance neglect.
However, it sounds like waiting for something like Dimitri's package
manager is the way to go rather than messing with schemas without other
instruments in place. I'll just go through and replace the BEGIN ...
COMMITS in the SQL scripts, which Bruce admits he deleted without really
thinking about it.
--Josh