Magnus Hagander <magnus@hagander.net> writes:
> On Wed, Nov 5, 2014 at 7:02 PM, Bruce Momjian <bruce@momjian.us> wrote:
>> Because 1500 % 100 == 0, I think 1500 was not a leap year.
> I believe it was a leap year in the Julian calendar, maybe that's
> where the difference comes from?
Indeed. We won't be changing our code though, because we document that
we follow Gregorian calendar rules even before that calendar was instituted
(ie, proleptic Gregorian calendar). You could argue for doing that
differently, but then what are you going to do for dates before the Julian
calendar was instituted? In any case, this behavior appears to be
required by the SQL standard, which repeatedly says that datetime values
are "constrained according to the Gregorian calendar".
regards, tom lane