> I have been reading all this technical talk about costs and such that
> I don't (_yet_) understand.
>
> Now I'm scared... what's the fastest way to do an equivalent of
> count(*) on a table to know how many items it has?
Make sure to analyze the database frequently and check pg_class for
reltuples field. This gives 0 time approximations of # row in table at
the time of the last analyze.
Many other approaches...check archives. Also your requirements are
probably not as high as you think they are ;)
Merlin