[LEAPSECS] Crunching Bulletin B numbers (POSIX time)

[Rob Seaman]

On Feb 21, 2011, at 1:14 PM, Poul-Henning Kamp wrote:


> For one thing, please tell why you compare a regular mechanism that is predictable over centuries, to one which is irregular and only announced 6 months in advance.


Because it isn't just a comparison of the distinct deployed solutions, but of the similar underlying problems.

That said, note that one out of four years are leap-years (one leap-day out of 1461 days).  Only one out of every five million seconds have been leap-seconds since this mechanism was introduced.

A six month prediction horizon means that each leap-seconds is known 1.5 million seconds in advance.  By comparison, there have been 100 leap-days in about 150,000 days since Gregory's decree.

In short, leap-seconds are a thousand times more "regular" than leap-days.

By all means let's improve the scheduling even further, but the prior generation of timekeepers did a pretty damn good job with the mechanism we have.

Rob