On Sun, Apr 25, 2010 at 02:31, Doug McNutt <dougl...@macnauchtan.com> wrote:
> Agree on a format for storing fractional atomic seconds. There are > proposals for two word integers with one of them being micro or nano seconds > and the other seconds. I prefer IEEE floating point with atomic seconds as > the unit of measure. That's a bit like MS Excel where days are used for the > unit. The Excel scheme is not as clean because it assumes 86400 seconds in a > day and it's a day off between what it thinks was Feb 29, 1900 and its epoch > at the start of 1900. > The Excel scheme is nowhere as clean as you think it is, because there is no _single_ Excel scheme. There are two epochs, one is the buggy 1900 epoch, the other is the 1904 epoch. In addition, there's an "extended" date format, which doesn't handle the Julian -> Gregorian transition very well. You'd be hard pressed to find a worse example of how to handle dates than Excel. -- Jan