Name: el35337 Date: 04/04/98


In both 1.2beta and 1.1.5, the method GregorianCalendar.isLeapYear()
returns false for all years before 45 BCE. A look at the class's code shows
that the behavior is intentional (it was the behavior suggested in bug report
4051768), but it's wrong.

While it's true that the Julian calendar wasn't introduced until that
year, one does not get the pre-Julian calendar merely by omitting the leap
year. The Roman republican calendar was entirely different. Many of the
months were of different lengths than they are now, and it employed a leap
month inserted at irregular intervals. Because this calendar is not
reconstructable by formula, historians and astronomers both adopt the
convention of projecting the Julian calendar back proleptically.

The formula used internally to convert between time and the year-month-day
fields uses this proleptic julian calendar, so for dates before
JULIAN_ONSET, you will have years that have a February 29, but all functions
and fields that depend on isLeapYear(), e.g. Calendar.DAYOFYEAR, will
be incorrect.
(Review ID: 27737)
======================================================================