Josh Berkus <josh(at)agliodbs(dot)com> writes:
>> It's the same, because the limits are calendar based (particularly,
>> the Julian-date functions) and not dependent on the representation.
> Hmmm? Just storing dates for the range described (until the year
> 294,000) takes 8bytes by my calculations. And that's without the 3
> bytes for the time zone. Is my math off?
timestamptz stores GMT; it doesn't store timezone separately.
(If it did, we'd need more than 8 bytes...)
> And, of course, this doesn't answer at all why time with time zone is so
> huge.
Because we haven't lifted a finger to optimize it.
regards, tom lane