[ntp:questions] A Suggestion For Abolishing the Leap Second

Richard B. Gilbert rgilbert88 at comcast.net
Sat Jun 2 21:54:52 UTC 2007

Danny Mayer wrote:
> Quadibloc wrote:
>>Some time back, controversy was raised by a suggestion that, due to
>>the problems in Internet timekeeping caused by leap seconds, that we
>>just go to straight atomic time.
> What problems? Leap second implementation is a mature and well
> understood technique and noone has any problems. The average person
> isn't even aware that a leap second took place.
>>If people really don't like local 12 noon happening at a time other
>>than lunchtime, they can always have a "leap hour", in effect
>>switching time zones, eventually.
> The person who suggested this has no clue about how or why leap seconds
> are inserted nor what the consequences would be.
>>Obviously, this was rejected. For one thing, even if it could be
>>argued that people would put up with such shifts in time - after all,
>>they accept daylight savings time - it would mean that while waiting
>>for the leap hour, the scheme of daylight savings time would keep
>>having to be changed to maintain the proper effects on energy use.
>>I suppose one possibility would be to allow UT1 minus UTC to get lots
>>bigger than 0.9 second, and just insert a *leap-minute* with *lots* of
>>advance notice. After all, while people notice daylight savings time,
>>nobody complains about the +/- 15 minutes of daylight savings time.
> Nobody but the people who need it even notice the leap second.
>>But I have another suggestion.
>>In a year where we would have had a leap second, why not just make the
>>second *of civil time* longer by a fixed amount, somewhere in the
>>neighborhood of 31.7 nanoseconds?
>>This wouldn't affect the SI second, so TV stations wouldn't have to
>>change frequencies slightly. Except for them, everyone else in the
>>private sector uses quartz crystal clocks, accurate to perhaps 5
>>seconds a year, so a 1 second per year change in the length of the
>>second shouldn't affect them.
>>This way, the exact length of the second is stipulated in advance, and
>>it just has a limited set of values, one SI second, plus or minus some
>>multiple of 30 nanoseconds.
>>One way to make the change to the second a "round" number, and make
>>the shift in time an integer number of seconds over a year, so as to
>>maximize compatibility with the present leap-second scheme, would be
>>to lengthen the second by 1/(86,400*360) of a second, but only for the
>>first 360 days of the year, going back to the SI second for the last 5
>>or 6 days.
>>For now, only the national time standards would have to update their
>>This would seem to the general public as if we went back to the old
>>mean solar day system, yet it would have the same exact precision as
>>leap seconds and atomic time had provided.
> Obviously you have no clue as to how or why a leap second is inserted.
> In the first place making the length of a second changeable is
> unworkable in practice and would break just about everything. In the
> second place you seem to be unaware that the frequency of insertion of
> leap seconds is irregular and has to do with both the rotation of the
> earth about it's axis as well as the rotation of the earth around the
> sun not to mention the effects of the moon on the earth's rotation. Even
> changing the length of a second permanently by a fixed amount  would
> break so many different things it would cost trillions in any currency
> that you care to name to make such a change. What do you think it would
> cost to change just the clocks within the existing orbiting satellites?

We would have to replace, or adjust, every clock and wrist watch on the 
planet!  We would have to do this within 24 hours or less and those 24 
hours would be pure hell for people who really need to know what time it is!

Outside of the NTP community, hardly anybody notices the leap second. I 
suspect that most of the clocks in the world are off by at least one 
second anyway.

As I recall, we didn't handle the last leap second any too well; at 
least half the NTP world appeared not to handle it correctly.  Hopefully 
we learned something and will do better next time.

More information about the questions mailing list