On 29 Dec 2006 at 12:08, Devon McCormick said: > Someone I know once calculated that there should be a 4000-year correction > as well, to whit: there is an exception (flip parity) to the "divisible by > 400" exception if the year is divisible by 4000.
Less than optimal. The current setup is equivalent to a mean year of 365.2425 days. The actual mean tropical year is 365.242199 days. The discrepancy is just over 26 seconds a year, or 1 day every %365.2425-365.242199 3322.259136 years. Omitting a leap year every 4000 years under-corrects, giving a mean year of 365+-/%4 100 400 4000 365.24225 days, which is still too long by 1 day every %365.24225-365.242199 19607.84314 years: you need to omit another leap year at about this interval (say, every 20000 years). A better match IMHO is to omit once every 3200 years. This gives a mean year length of 365+-/%4 100 400 3200 365.2421875 days, which is too short by one day in %365.2421875-~365.242199 86956.52153 years, which doesn't seem to admit a neat adjustment (though putting a leap year back every 86400 years has a certain attraction). However, since any change to the current calendar requires agreement among at least most of the world's governments, it probably won't get dealt with any time soon. ---------------------------------------------------------------------- For information about J forums see http://www.jsoftware.com/forums.htm
