[LEAPSECS] internet drafts about zoneinfo (POSIX Time)

[Ian Batten]

> In practice, if one uses NTP to synchronize to UTC, a 

> POSIX-compliant computer will follow leap seconds more-or-less dutifully, 

> never mind the standard.


Well, only by counting one second fewer than actually elapsed.     At 00:00:01 UTC following a leap second, the kernel will think it is 00:00:02 and will therefore (if externally disciplined) slew the clock in order to match UTC reality once it realises it is a second ahead.  It's can't step, obviously.    If it does this over ten minutes, which would be typical, that will mean that for ten minutes kernel seconds will be one part in six hundred shorter than SI seconds.    If the kernel realises to step backwards, which is unlikely but not unheard of, you'll get stranger effects.


> 

> This was as close to declaring that POSIX seconds are SI seconds (to the 

> accuracy of the local clock) as I could get the committee to go.


But they aren't.  In the aftermath of a leapsecond they are shorter.  This isn't the local seconds being slewed to allow on Wednesday for the thermal or other error drift that happened on Tuesday, this is the local seconds being slewed to match external reality.  Even if the local clock were Caesium-disciplined, you'd still have to do it absent a solid mechanism for handling leap seconds.


> Given that 99% of computer buyers never heard of SI seconds, which is why 

> the clock oscillators in computers often make better thermometers than 

> clocks, this seems to me to be a reasonable position. It is expected that 

> those who do know and care will buy the needed special hardware.


Which special hardware is it that will allow a Unix machine to both tick SI seconds and accurately follow leap seconds?

ian