>> I meant we use them in many cases to link entries, and in
>> pg_description for descriptions and lots of other things
>> that may use them in the future for system table use.
pg_description is a point I hadn't thought about --- it uses OIDs
to refer to pg_attribute entries. However, pg_description is pretty
broken in its assumptions about OIDs anyway. I'm inclined to change
it to be indexed by
(object type ID, object OID, attributenumber)
the same way that Philip proposed indexing pg_depend. Among other
things, that'd make it much cheaper to drop comments during a DROP
TABLE. You could just scan on (object type ID, object OID), and get
both the table and all its columns in a single indexscan search,
not one per column as happens now.
regards, tom lane