[LEAPSECS] POSIX?

[Rob Seaman]

On Dec 30, 2016, at 10:40 PM, Warner Losh <imp at bsdimp.com> wrote:

> Nothing is ever re-written from scratch.


Which includes vast amounts of scientific literature in which Universal Time corresponds to mean solar time at Greenwich.

Nothing and nobody keep POSIX or ITU or BIPM from defining a new time scale with the features desired. Or simply use the GPS or TAI time scales that already exist and are available from commercial vendors.

UTC should be preserved as a functional representation of Universal Time for backwards compatibility.


> And POSIX time_t can't represent leap seconds any more today than it could in 30 years ago. Until this fundamental issue is resolved, no forward progress is possible.
> 

> The "factual reality" is a construct of man.

It’s a topsy-turvy world if POSIX is regarded as fundamental, and the simple descriptive astronomy of the synodic day that applies to this and every other world is regarded as a construct.


> UTC is arbitrary in its limits of 0.9s.

By all means debate the scheduling rules for leap seconds. There is a six month minimum for announcing leap seconds. Nothing stops IERS from announcing them years in advance, and it is clear that even within the 0.9s limit this could already happen a few years ahead.


> That's an invention of man as well. It’s tradition, sure, I'll grant that, but it isn't a fundamental constant of the universe. Go only as far as Mars and everything changes…


Rather, it is indeed a fundamental observation of the behavior of the universe.

On Mars as well as Earth the day is the mean solar day, and spacecraft operations understand the distinction between time of day on the planet and atomic/interval time ticking away onboard. Day means the same thing on all worlds in this and every planetary system:

	http://hanksville.org/futureofutc/preprints/files/28_AAS_13-515_Seaman.pdf <http://hanksville.org/futureofutc/preprints/files/28_AAS_13-515_Seaman.pdf>

There is one fewer day per year than sidereal rotations on Mars as well as Earth.


> It’s been proposed, and shot down, that we make our best guess 10 years out and adjust
> from there.

It has been raised on this list, and whatever the prior discussions were, they don’t amount to its being shot down.

It has never been debated by the ITU.


> It's been proposed that we announce 2 years out that the current models can predict with 99% certainty, that's been shot down.

Again, it has not been proposed within the IERS.


> GPS time exists because leap seconds were a horrible idea and they had to invent their own thing.

Rather, GPS is how UTC is implemented:

	http://hanksville.org/futureofutc/2011/preprints/32_AAS_11-675_Malys.pdf <http://hanksville.org/futureofutc/2011/preprints/32_AAS_11-675_Malys.pdf>

The GPS timescale exists because its engineers reviewed formally expressed engineering requirements and realized that a leapless timescale was required. One imagines that they first logically considered using International Atomic Time, and that perhaps they were told TAI is a “paper clock” or some such bureaucratic jargon. So they invented their own time scale to meet the engineering and operational requirements. GPS is a reaction against TAI, not UTC.

Applications that require interval time should use GPS or TAI. Applications that require time of day should use Universal Time, often UTC. Those relatively few applications that require both should use both.


> But computers and business don't run on GPS or TAI time like some specialized systems can. They are required in many cases to run on UTC

Systems engineering best practices are the quickest way to resolve the issues in business applications, too.


>> Time of day is a different problem than interval time. Geophysics exists.
> 
> I find that argument unpersuasive. Time of day matters because I need to know when to go to work.

That is one use case. Respectfully, there are many more.

The cadence of our lives ticks by at the synodic rate. That the totally separate phenomena of the SI-second was calibrated against a sexagesimal fraction of the length-of-day circa 1820 is historical happenstance.

We would be better off if there were a round 10,000,000,000 periods of the radiation corresponding to the transition between the two hyperfine levels of the ground state of the caesium 133 atom in an SI-second. This would emphasize the difference between what is ultimately a frequency standard and the unrelated fact of the synodic length of the day on Earth, on Mars, on planets and moons in this and every other solar system.


> The use of time by society is evolving,

Yes. It should evolve in concert with physical reality.


> the close, rigid coupling to the spinning globe may not make sense.

Right, and the way to sever the coupling is to recognize that solar time and atomic time have always been two entirely different kinds of time. We should stop trying to pretend they are the same.


> The de-coupling may not make sense either, but the current unpredictability beyond a half a year is just nuts.

Nothing about the current standard forbids the IERS from announcing leap seconds on a longer schedule. The best change to a standard is no change at all.


Let me know if you’re ever in Arizona and I’ll buy you a margarita ;-)

Happy New Year,

Rob



-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://pairlist6.pair.net/pipermail/leapsecs/attachments/20161231/108ee3ee/attachment-0001.html>