The important thing to me is that, within
the range of supported dates for any implementation, functions like WEEKDAY()
should always give the correct answers, and any function that takes a range
of dates works correctly for ranges that are entirely before March 1st,
1900, entirely after March 1st, 1900, as well as date ranges that include
March 1st, 1900. This includes date arithmatic as well. You
should not be able to get a non-existant date by adding 1 to '2/28/1900',
or subtracting 1 from '3/1/1900'.
If we want an normative ISO reference
for the Gregorian Calendar, we can refer to ISO 8601:2004 "Representation
of dates and times", section 3.2.1 "The Gregorian calendar"
-Rob
wrote on 10/18/2006 08:00:34 AM:
> Hi David,
>
> On Tuesday, 2006-10-17 18:35:55 -0400, David A. Wheeler wrote:
>
> > Excel thinks that 1900 was a leap year
> >
> > OpenFormula avoids this.
> > We do NOT require that implementations get 1900 WRONG.
>
> Btw, this is the reason why the "usual" null-date in OOoCalc
is
> 1899-12-30 instead of 1899-12-31 as in Excel, being able to both
> correctly handle 1900 as a leap year and to get correct dates from
> 1900-03-01 on, with the consequence that all dates between 1900-01-01
> and 1900-02-28 when calculated from serial date values differ by one
day
> between Excel and OOoCalc.
>
> Should we add this quirk to the notes of section "4.2.2 Date
and DateTime"?
