[LEAPSECS] any other parties?

[Warner Losh]

On Jul 9, 2012, at 6:10 AM, Nero Imhard wrote:


> Zefram wrote:

>> isn't very celebratory.  A friend of mine has suggested that one should

>> celebrate a leap second the NTP/Unix way, by repeating whatever one did

>> in the previous second

> 

> Is that really the Unix way? I always thought that "proper"

> implementations didn't step back, but paused the time for a second at

> :59:59 (while still having subsequent calls of gettimeofday return

> increasing values by adding microseconds as needed). Do I recall

> incorrectly?


There's four different Unix implementations of leap seconds

(1) Repeat the first second of the next day.
(2) Repeat the last second of leap day.
(3) Freeze time
(4) slew it in over many hours.

Most opt for 1 or 2 so that smart programs can see the leap second through non-standard interfaces.  3 is actually quite rare. 4 is used to paper over the leap second, but the number of stupid apps out there is quite large, so this actually is viable for a shockingly large class of users.

Warner