Peter J. Acklam schreef: > I could have sworn the difference was 0 seconds between 1970-01-01 > and until the leap second in June 1972. I should have checked > > ftp://maia.usno.navy.mil/ser7/tai-utc.dat
By the way, according to the formulae on this page, TAI-UTC was 9.89 seconds on 1971-12-31, so DateTime::LeapSecond is probably wrong to introduce a leap second on 1972-01-01. It seems that before the leap second on 1972-07-01, UTC was corrected in 0.1 second steps, and the 1972-01-01 correction is very close to that. At http://cr.yp.to/proto/utctai.html, D.J. Bernstein, author of libtai, claims that the Unix epoch (1970-01-01T00:00:00) is 1970-01-01 00:00:10 TAI. This is only possible if 1972-01-01 has no leap second. http://www.phys.uu.nl/~vgent/astro/deltatime.htm shows TT-UT1 (approximately equal to TAI-UTC+32s) for 501BC to 2014AD. A real UTC <-> TAI conversion module should perhaps incorporate these tables... Between 501BC and 451BC, there was on average more than one leap second a month. I wonder how they managed to keep their clocks accurate. Eugene
