Re: index vs. seq scan choice?

Alvaro Herrera <alvherre@commandprompt.com> writes:
> Tom Lane wrote:
>> (The default statistics target is 10, which is widely considered too
>> low --- you might find 100 more suitable.)

> Does this mean that we should look into raising the default a bit?

Probably ... the question is to what.

The default of 10 was chosen in our usual spirit of conservatism ---
and IIRC it was replacing code that tracked only *one* most common
value, so it was already a factor of 10 better (and more expensive)
than what was there before.  But subsequent history suggests it's
too small.  I'm not sure I want to vote for another 10x increase by
default, though.

            regards, tom lane