On Sat, Nov 28, 2009 at 4:57 PM, Tom Lane <tgl@sss.pgh.pa.us> wrote:
> Eduardo Piombino <drakorg@gmail.com> writes:
>> I see current criteria and all the SQL-standard compliance policy, but
>> wouldn't it still make sense to be able to store a date reference, along
>> with a time zone reference?
>> Wouldn't it be useful, wouldn't it be elegant?
>
> It seems pretty ill-defined to me, considering that many jurisdictions
> don't switch daylight savings time at local midnight. How would you
> know which zone applied on a DST transition date?
Yeah, I think the only reasonable way to define a date with a timezone
would be as some kind of interval, starting at 00:00:00 and going
until 23:59:59.99999 (or < 00:00:00 next day, whichever is more
accurate. On spring forward / fall back days it would be 23 or 25
hours respectively. I'm not sure what you'd DO with it though.
> TIME WITH TIME ZONE. We only put it in for minimal spec compliance.
Yeah, it's kinda twilight zonish to me.