On Mon, Feb 24, 2003 at 11:06:38PM -0500, Tom Lane wrote:
> Robert Treat <xzilla@users.sourceforge.net> writes:
> > I guess adding 1 day to 1752-09-02 should give us 1752-09-14, but your
> > right, it gives us 1752-09-03.
>
> As was pointed out at length just recently, the transition from Julian
> to Gregorian calendars happened at different times in different places.
> So the above claim is only correct for some places.
>
> The conclusion from the previous discussion was that our existing
> behavior (extrapolate Gregorian rules backwards indefinitely) is as
> defensible as anything else that would be likely to get coded.
To quote SQL1992:
4.5.3 Operations involving datetimes and intervals
[...]
Arithmetic operations involving items of type datetime or inter-
val obey the natural rules associated with dates and times and
yield valid datetime or interval results according to the Gregorian
calendar.
Ross