+
+Note that the dbmdb lookup use, bypassing the dbfn.c layer,
+means that no file-locking is done.
+XXX This feels like a layering violation; I don't see it commented on
+anywhere.
+
+Future: consider re-architecting to support caching of the open-handle
+for hintsdb uses (the dbmdb use gets that already). This would need APIs
+for transaction locks. Perhaps merge the implementation with the lookups
+layer, in some way, for the open-handle caching (since that manages closes
+required by Exim's process transisitions)?