I never got around to documenting this for 1.0M4, but there is some maintenance that you should do in a script on a semi-regular basis.

For PostgreSQL:

delete from pg_largeobject where loid not in (select bodyid from messagedata)
delete from storeitemmetadata where pid not in (select loid from pg_largeobject)

Should be done as either postgres or a user with "CREATEUSER"

For DBs similar to MySQL or HQL (using the paged store -- tablenames may not be exact in case/existence of _'s):

delete from paged_store where id not in (select m.bodyid from messagedata m)

delete from paged_store_pages where blobid not in (select b.id from paged_store_pages b

delete from storeitemmetadata s where s.pid not in (select b.id from paged_store b)

For DBs similar to Oracle using JDBC3 (table names may not be exact):

delete from blob_mail_store where id not in (select m.bodyid from messagedata m)

delete from storeitemmetadata s where s.pid not in (select b.id from blob_mail_store b

The issue is that we don't expunge large objects regularly. I intended this for performance but I've not yet decided whether the tradeoffs on the various dbs of doing this irregularly (or in a script or something) outweigh the benefits. For the next release we'll likely make this self executing as part of the expunge/close/dropcarrier or pop/quit/dropcarrier... Figuring their overhead should be moderately low and at an "unimportant" point from a performance perspective.

I'll document all this in the wiki when I get time or around the 1.0M5 release.

Thanks,

Andy