[LEAPSECS] leap smear
stenn at ntp.org
Sun Sep 18 17:32:27 EDT 2011
> Not really. The problem with leap seconds is that they are too rare
> to allow for comprehensive testing of systems, and so such systems
> tend to fail when a leap second comes along. This is true regardless
> of the chosen OS, and the mission software can also screw up.
That doesn't make sense to me. Feb 29th happens within an order of
magnitude as often as leap seconds happen, and we have well-debugged
algorithms to handle this.
We also have plenty of code that ignores leap years, and some that only
applies the "divisible by 4" rule, and then some that only apply the 4
and 100 year corrections.
The problem seems to be that while the leap year algorithm is stable, as
is the leap second handling *code*, the issue of "when is the next leap
second gonna happen" is not so predictable.
More information about the LEAPSECS