[LEAPSECS] Following an open source process

Rob Seaman seaman at noao.edu
Mon Mar 7 00:46:48 EST 2011


Hi Tom,

Second try - thanks to Tom for a revised figure sized to fit. I edited a bit in the interim.

Thanks for the opportunity to wrestle over the requirements. By focusing on the definition of the single common shared problem, we naturally constrain the ultimate solution space.


>> - the civil day is the synodic day

>

> Rob, please define "is". Surely you don't mean equality, in a

> mathematical sense. What really is meant by this statement?


What are you offering as an alternative definition?

Just as the rate of TAI is the SI-second, the rate of universal time is the synodic day. Civil timekeeping - time kept for a multitude of purposes for our society's cultural and technical purposes - is obviously connected to the the day-and-night cadence of our calendar.

Timekeeping provokes many subtle discussions. There is nothing subtle about the synodic day. Our calendars don't count fractional or approximate days, our calendars - Gregorian, Chinese, Hebrew, whatever - count integral days. Those days are mean solar days, i.e., synodic days.


> You bring this topic up a lot. But everyone knows civil time is

> now only grossly associated with anything solar; it doesn't match

> at the second level, not even the minute level, though usually near

> an hour or two, the equivalent of hundreds or thousands of miles.


...and I also reject assertions like this paragraph a lot. Apparent solar time is a red herring. The actual day on earth is the sidereal day - the rotation of the Earth relative to the stars. The actual year is 366.25 of these days long. The synodic day is the sidereal rate adjusted to account for lapping the Sun by one integer day per year. One can quibble about the details, but the basic mechanism is clear.

Discrepancies due to the equation of time or an offset from the standard meridian of the local time zone are either static (in the latter case) or periodic (in the former). Rather, the ITU intends to insert a secular tilt to the entire clock-and-calendar mechanism. They are attempting to redefine the day and year on Earth, not just mucking about with some technicality of a leap second mechanism.

Here is a picture of length-of-day, hyper-zoomed to show detail:

-------------- next part --------------
A non-text attachment was scrubbed...
Name: PastedGraphic1.gif
Type: image/gif
Size: 20701 bytes
Desc: not available
Url : <http://six.pairlist.net/pipermail/leapsecs/attachments/20110306/b712b544/attachment-0001.gif>
-------------- next part --------------


(The actual plot reaching up from the origin is flat-topped to within the thickness of the line.)

Length-of-day excursions from the mean solar day are much smaller than you assert. Also, what is a day if not a mean solar day? What kind of days would our calendars count? It might be one thing to assert that the ITU's position is "good enough" in some fashion, but it is simply absurd to assert that the fundamental civil meaning of the word day is anything other than "synodic day". The ITU's redefined length-of-day is tethered to within a few milliseconds of the synodic day. They could not, for instance, assert that a day is 86399 SI-seconds or 86401 SI-seconds. In either case a +/- leap hour would be needed every decade - there is no way the "civil authorities" that folks keep appealing to would tolerate such nonsense. The number they are bracketing is the mean solar day, that is, the synodic day.

The synodic day changes with time. The IERS provides wonderful data sets to characterize the changing length of day. This is simply another requirement - the Earth is spinning down. I'm confident we can identify a consensus mechanism to meet this requirement in some more creative way than the relentless ITU "initiative" has allowed to date.


> Local civil time used to be determine by the astronomer-guy in

> your local village. Now it's not. Get over it.


Again as repeatedly pointed out, astronomers are power users of both universal time and atomic time. These are two different kinds of time scale. Both are important for technical reasons. Both are important for civil reasons. Pretending there is only one kind of time is bad engineering, bad system architecture, just simply wrong.


>> - a non-UT timescale should be called something other than "UTx"

>

> This sounds more like a preference than an absolute requirement,

> but I agree it's a good preference, so leave it on the table.

>

> I notice some model have dates to models; like JD2000. Would it

> help so use that for UT as well, like UT1972 or UT2020? Or make

> things worse.


"UT" means "Universal Time" means "mean solar time" means a rate synchronized with the synodic day. If there are other kinds of UT, by all means subclass them. However, the mess of a draft position that the ITU is going to vote on would no longer constitute a type of universal time.


> Any progress with relaxing decades-old outdated requirements on

> DUT1 would be a good thing.


"Outdated requirements" suggests that anybody is paying attention to collecting the requirements. It is eminently reasonable to reconsider the requirements from time to time. It was the 1999 M&K GPS World article that suggested revisiting the DUT1 tolerances. Isn't it about time that this suggestion actually be pursued?


> Astronomers have UT and no one is touching that.


Rather, if UTC is debauched into something other than universal time, it will call into question the entire universal time "brand name". This would be very bad for astronomers, but will also be very bad for everybody else given how widely used the terminology is. When "users" (people) say "universal time" will they mean "the redefined time scale with a constant offset from TAI" or will they mean what they always have, "time-of-day as an approximation to Greenwich Mean Time"?

Just call any new time scale something other than "UTC". It is the right choice. It is the safest choice.


> Presumably you use UT in your automated systems.


"Presumably" is a word that implies no planning has gone into the ITU machinations. If the ITU believes what you say to be the case, write a white paper telling us how we will use "UT" in our automated systems in the future after they make UTC unsuitable. Currently the vast majority of our systems assume that UTC is "close enough" to the underlying universal time. It will cost the astronomical community a very large amount of money to accommodate a redefined UTC.


> The world of precision timing has TAI and its clones and no one is

> touching that.


Rather, the ITU appears to have an unstated agenda to suppress TAI after UTC is redefined. The precision timekeeping tzars appear to care little for UTC as currently defined, for TAI as it will be in the future, and for GPS as it ever has been.


> Civil time switched from its UT-base in the 60's.


The only reason we've been able fiddle with timekeeping around the edges is that UTC has remained an acceptably close approximation to universal time. On this list we're perpetually fretting about highly technical trivialities of the situation and missing the big obvious issues.


> It seems most of the problem we're running into is keeping DUT1

> within some magical bounds. If you read the old papers it was

> navigators and their sextants they were worried about. First they

> wanted 0.1 second tolerance. Then, what, 0.5? 0.7? 0.9? No one

> worried about astronomers; they already had UT-sub-n for their

> work. It was the sailors that was the problem back then.


I find "founding father" arguments unpersuasive politically and I find appeals to authority unpersuasive technically. Just focus on the system engineering requirements as they currently exist.


> Personally I don't see a problem if DUT1 is allowed to grow to a

> much larger value.


As have folks stated on all sides of the issues on this list. "Much larger" is a topic of debate, but larger than currently is preferable to "forgetting the whole thing". But conversations on this list only matter as far as they may influence actions at the ITU. Presumably this is very little. System engineering best practices support the conversion of personal opinions to shared consensus.

Also, there is a difference between a proposal that the limit of DUT1 = |UTC - UT1| be increased to some value larger than 0.9s and a "proposal" that we ignore the whole damn thing and let DUT1 flop around completely uncontrolled.


>> However, pretending the entire world can ignore the synodic day is not an option.

>

> Please define "ignore".


The Sun rises. The Sun sets. The rate is the synodic day. Tweaking civil timekeeping is a possibility. Changing the entire architecture is not. If timekeeping is redefined for the entire planet, one might expect a white paper or two to accompany the standards process to guide in coherent planning. If timekeeping is redefined worldwide, one might expect consensus to be built in advance.

In short, the world treats days like synodic days while it treats seconds sometimes like SI-seconds and sometimes like "synodic seconds" (1/86400 of a day). For some purposes at some times some people can indeed ignore a precise meaning for the word day. But not all the time, all the people, and for all purposes.

Rip van Winkle doesn't just go to sleep and find out in 500 years that all the clocks are set wrong. The clocks are set wrong every hour of every day in between. During each moment in the interim the error will be asserting itself - very obviously so for astronomers, but in innumerable subtle and not so subtle ways among systems and subsystems worldwide, straining interoperability and logistics.

So:


> Rob, please define "is".


The civil day is not 86,400 SI-seconds. The civil day is not local apparent solar day. The civil day is not a free parameter. The common worldwide civil day "is" the synodic day. The timezone system is a layer above this.

By all means let's discuss tweaking the approximation that is presented by our clocks. But it is simply wrong to suggest that the ITU or any other international standards body can redefine the word "day". "Wrong" not in a moral sense, rather wrong in the sense of ignoring the system requirements. Attempting to pretend otherwise will make unintended consequences and unmanageable risks inevitably reappear in unexpected ways at unpredictable times to unsuspecting projects and people.

The synodic day "is" a requirement of civil timekeeping:

http://en.wikipedia.org/wiki/Requirements_analysis

Requirements describe the common shared problem space. What we do about it - leap seconds or otherwise - are issues for the various solution space trade-offs. The quickest way to reach a consensus on a new international timekeeping solution is to honor the actual system requirements. "Civil day" means "synodic day". Work forward from that point.

Trying to redefine the problem out of existence is the slowest process (as we have seen) to identify actual candidate solutions. It is also just about the riskiest.

Rob


More information about the LEAPSECS mailing list