[ntp:questions] Re: HOWTO prepare ntpd to the leap of a second

David L. Mills mills at udel.edu
Thu Dec 29 22:54:20 UTC 2005


You really should read the commentary in the NIST leapsecond file, which 
explains the rationale for when the file is issued, the epoch of 
insertion and the current TAI offset. The file does not expire when the 
leap is inserted; it remains valid as a record of past TAI offsets until 
the next edition is issued.

It could be that some sites have an old edition and others a newer one. 
The Autokey protocol compares the leapsecond tables for itself and its 
upstream servers and uses only the newest one. If the client has 
dependent clients, it passes only the newest table to them.

When multiple valid sources display conflicting leap bits, the logical 
OR of these bits is used. While a more sophisticated decision algorthm 
could be used, the most common error is when an upstream server does not 
implement or recognize the leap condition. Thus if a server sets the 
bits, it probably knows what it is doing.


Serge Bets wrote:
>  On Thursday, December 29, 2005 at 14:16:09 +0000, David L. Mills wrote:
>>Serge Bets wrote:
>>>The NIST file is absolutely true until expiration. This fact is not
>>>exploited by the NTP daemon, but could perhaps be.
>>I don't understand your concern.
> The NTP daemon will not assert leap=00 against any bogus source wrongly
> saying leap=01 during June 2006.
> The authority of the soon updated NIST leapseconds file, expiring only
> later on 29 December 2006, thus valid, and showing no leap event on
> 30 June 2006, could perhaps be used to force correct leap=00.
> Effectively overruling the less sure leap bits of other sources.
> Serge.

More information about the questions mailing list