> That one's used, too, but the other is used as an upper bound.
> n_distinct tends to come out too small on large tables, so that
> formula is prone to overestimation. Actually, both formulas are prone
> to overestimation.
>
Right - thanks.
> When this happens depends on the values of a whole boat-load of GUCs...
Well, then doesn't the 'proper' number of buckets depend on a whole
boat-load of GUCs...
-Nathan