[LEAPSECS] Fractured fairy tale

Rob Seaman seaman at noao.edu
Fri Dec 24 02:02:12 EST 2010

Rob Seaman wrote:

>> Which is to say that mean solar time is a requirement. Leap seconds as we currently know them are one possible way to implement that requirement. The latter is negotiable. The former is not.

Warner Losh wrote:

> Mean solar time is not a requirement. It is merely a tradition.

No, it is a system requirement in the engineering sense of the word.

What are requirements? Requirements are not "specifications", rather they are physical, functional, and logistical conditions implicit in a problem's "concept of operations". In this case to provide a timescale suitable for the civil purposes of our culture and its technology. The assertion that mean solar time is a requirement for civil timekeeping derives from two more fundamental concepts:

1) Many or most human activities (and the technology in support of these) obey a diurnal scheduling frequency.

2) Mean solar time is the Earth's diurnal period averaged over a year to remove periodic effects.

Civil time must closely track mean solar time. Closely enough (a few hundredths of a second out of 86400 seconds) that it is obvious:


that the meaning of the word "day" is "one mean solar day". Mean solar time is simply the sidereal rotation period of the Earth adjusted by subtracting one (the integer "1") day per year due to the Earth lapping the Sun. Mean solar time is not a free parameter (unlike the original definition of the SI second).

The ITU is proposing a cheat. Another word for cheat in systems engineering is a "lien" against the requirements. Which is to say that it might be recognized that a requirement is not satisfied by some deployed implementation - but that most certainly does not make the requirement go away. Another word for lien is kludge.

> As has been noted before, there has been a steady march away from mean solar time. Time originally was governed by the local solar time. Then, it was governed, for those in cities, by whatever the clock in the center of town said. Then timezones of the 1850's expanded the notion nations, and then internationally. We moved further away from the sun with the adoption of the second as based on the mean second of 1900, away from the second of the current era. Then we moved to atomic time kept in sync at first by its varying frequency. The most recent change has been to tick at a constant frequency but insert leap seconds into the labels we put on the seconds that tick by to keep things more or less in alignment to the earth.

And has been noted time and again before, this mythical recounting confuses secular trends with periodic effects, for instance see the discussion and links within:


Is it the word "mean" that perpetuates the confusion? Lots of physical effects have the word "mean" in them, e.g., "mean free path" or "mean value theorem". In this case, "mean solar time" defines the natural synodic day of the Earth. If "mean solar time" seems somehow deficient as terminology, how about the word "day"?

> Based on this history, it is far from clear that mean solar time is an absolute requirement.

The phrasing "absolute requirement" suggests a fundamental misunderstanding of the concept of an engineering requirement. A requirement is just that - required. Even the least, most obscure requirement must be met by a conforming system. But the concept of the "day" is not obscure; it is embedded throughout our cultural and technological artifacts. Much more fundamentally than the murky concept of the "second".

Mean solar time is that timescale that defines our civilization's calendar and clock. It is special. It is unique. That it may be regarded as a pain in the ass by some, doesn't make it optional.

There are other timescales which don't face the same requirements as MST. If you don't like leap seconds, either propose a different mechanism for MST that continues to satisfy its requirements, or use one of the other timescales and trace its different requirements through to your own system design efforts.

> Mean solar time is another way of saying that time must measure the angle of rotation of the earth.

"Measuring the angle of rotation of the earth" is another way of saying "time-of-day".

> Time keeping has moved beyond that one property. It is merely tradition at this point as many of the motivating factors for keeping UTC and UT1 in close harmony have changed.

As factors change it is as likely that prior requirements become more stringent as otherwise. On Earth innumerable systems and processes obey diurnal constraints. On the Moon, the Apollo astronauts most certainly paid attention to lunar daytime. On Mars, rovers operate on the local diurnal cadence. "Diurnality" is an important requirement. Even in Asimov's Caves of Steel, the unseen Sun was a critical influence.

Length-of-day is the fundamental harmonic of our lives. Mean Solar Time is the musical scale within which that note is sounded. It is a scale which like a fine piano requires occasional retuning. That is not a justification for allowing our clocks to play flat.

> While you make many good points, this point I think is the fundamental source of disagreement in this list...

No, it is the fundamental source of misunderstanding. The word "disagreement" implies that more than one point of view is valid. If the ITU plans to cheat the system, the corresponding system engineering plan needs to describe in gory detail how this lien is justified and what mitigation is required.

The ITU has no authority over the definition of the word "day". If they are going to redefine UTC to break its internal foundation layered on this concept, they have a responsibility to document the contingent issues. Hard work? Yes. Necessary work? Yes.

Due diligence is laughably absent in the ITU proposal.


More information about the LEAPSECS mailing list