Ron Mayer <rm_pg(at)cheapcomplexdevices(dot)com> writes:
> Tom Lane wrote:
>> BTW, I'm thinking that a "cost constant" probably ought to be measured
>> in units of cpu_operator_cost.
> Any chance that costs could eventually change to real-world units?
Define "real world units".
If you like you can try to adjust things so that cpu_operator_cost bears
some relation to elapsed-msec-on-your-own-machine, and scale everything
else accordingly; indeed that's why we added seq_page_cost in 8.2.
But it's awfully hard to see the numbers being very portable.
regards, tom lane