[ntp:questions] The Leap Millisecond

David L. Mills mills at udel.edu
Sun Jun 3 18:38:04 UTC 2007


See the list at http://hpiers.obspm.fr/eoppc/bul/bulc/UTC-TAI.history.

The current and past leapseconds files can be obtained via ftp from 


Judah keeps a current copy for ftp at time.nist.gov.Guy Macon wrote:
> (While researching this, I came across the leap second list 
> at [ http://www.ee.udel.edu/~mills/leap-seconds.3169152000 ],
> which expired in 2004.  Does anyone know of a newer version?)
> In the sci.astro.amateur and sci.astro newsgroups (See "Avoiding 
> the Leap Second" thread) Quadibloc (John Savard) wrote:
>>I've come up with an alternate scheme.
>>Divide the year into ten parts of 37 and 36 days in alternation. Start
>>from March 1 to keep things simple in leap years.
>>For the first 33 1/3 days of each of those parts, sweep increments of
>>100 milliseconds "under the rug" by adding one millisecond to the last
>>second of each eight-hour period. This would allow a time scale to be
>>kept within 0.1 seconds of mean solar time, and it would also mean
>>that, most of the time, a time signal would consist of a steady stream
>>of SI seconds; the long seconds would come at predictable intervals.
> An excellent scheme.  Let me be the first to say that I approve.
> Looking forward to possible objections to certain seconds being
> 1.001 times longer or shorter than most of the other seconds, 
> that seems to me to be far less troublesome than having some 
> minutes be some 1.01666... times longer or shorter than most 
> of the other minutes using leap seconds.
> Here is what I like about this scheme:
> The above scheme and the existing leap second scheme both 
> result in the exact same length of the millisecond, microsecond, 
> nanosecond, etc.  Those are much more commonly used than the 
> second is in the areas of physics and engineering.  In the area 
> of computers, time is typically specified as date and time, which 
> means that the computers already have to work with the occasional 
> minute that is 59 or 61 seconds long. 
> The above scheme results in a length for the minute, hour and day 
> that is no more than a millisecond larger or smaller than  most of 
> the other minutes, hours or days.  This is a thousand times closer
> than under the leap second scheme.  Months and years would be ten 
> times closer.
> References:
> The NTP Timescale and Leap Seconds:
> http://www.ee.udel.edu/~mills/leap.html
> (also touches on GPS)
> The Future of Leap Seconds
> http://www.ucolick.org/~sla/leapsecs/onlinebib.html
> Propagation of a leap second
> http://members.iinet.net.au/~nathanael/ntpd/leap-second.html
> Leap Second Mailing List:
> http://rom.usno.navy.mil/archives/leapsecs.html

More information about the questions mailing list