On Fri, 2011-05-27 at 16:43 -0400, Alvaro Herrera wrote:
> One of our customers is interested in being able to store original
> timezone along with a certain timestamp.
Another thing to consider is that this will eliminate any useful total
order.
You could define an arbitrary total order, of course, just to allow
BTrees for equality searches. However, I don't think you should define
">" (and other non-equality comparator operators) according to that
total order -- they should be more hidden like "~>~". ">" should not
exist as an operator over this type at all.
I also do not like the idea of having "=" mean "equivalent after
timezone adjustment". If we're making a distinction between "2000-01-01
10:00:00 +03" and "2000-01-01 9:00:00 +02", then "=" should not obscure
that distinction.
Regards,Jeff Davis