[ntp:hackers] Re: NTP and leap-seconds

David L. Mills mills at udel.edu
Wed Jul 5 00:31:59 UTC 2006


Harlan,

The only plaint I know about is a suggestion to use a majority test for 
leap indication designed to avoid terrorist servers that show leap in 
error at other times.

Leap seconds don't have timezones. The actual leap is at midnight UTC; 
it will occur at other times in other timezones exactly as intended.
There is no need to support the leap on any timezone other than UTC; 
just add the tz offset and be done with it.

Dave

Harlan Stenn wrote:

> Recent Olson tz library implementations support either/both with(out)
> leap second for timezones. ftp://elsie.nci.nih.gov/pub/ is where to
> find it.
>
> There have been 5 bugs reported with the word "leap" in the summary; 3
> are now closed. (508 and 539 are still open; 521, 525, and 543 are
> closed).
>
> Somebody sent in a request to have ntpd behave differently (properly, in
> his opinion, as I recall) if the base OS was already handling leap
> seconds. I have not been able to find those emails, however. His
> approach was to check to see if the OS was default-configured to use the
> "right" timezone (see the Olson stuff above) and (do something) in that
> case in the low-level time set/get code of ntpd. I think...
>
> H




More information about the hackers mailing list