On Aug 20, 2014, at 9:28 PM, Brooks Harris <bro...@edlmax.com> wrote:
> On 2014-08-20 10:23 PM, Warner Losh wrote: >> On Aug 20, 2014, at 7:40 PM, Brooks Harris <bro...@edlmax.com> >> wrote: >> >> >>> Since the beginning of civilization society has pursued the goal of perfect >>> timekeeping. UTC is one of the great intellectual achievements of mankind. >>> >>> The difficulties with Leap Seconds are rooted in computer standards and >>> implementations, not theory. We must strive to advance the state of the >>> art, not abandon 4500 years of timekeeping history just because its >>> difficult. Ceaser didn't quit. Pope Gregory didn't quit. Harrison didn't >>> quit. Newcomb didn't quit. >>> >>> "Dropping Leap Seconds" is something like burning great libraries. >>> >> Not sure I buy this hyperbole. > > That's the fun of it! > >> Leap seconds suffer from many minor issues that are only going to increase >> as we get more and more connected. > > That's where an effort to consolidate the specifications and remove the flaws > in computer standards could lead to uniform implementations. Absolutely. We get leap days right because we don’t have to hear from the pope’s astronomers every year to know if it will be a leap year or not. We know for thousands of years. >> The stubborn refusal for any changes in the current implementation has me >> thinking of them as less than great libraries, and more as pebbles in my >> shoes. > I see the "pebbles in my shoes" as the flaws in standards and > implementations. They are sharp and painful, to be sure, but abandoning Leap > Seconds dismisses the history of timekeeping and, by itself, won't eliminate > the pain anyway - you'll still need to reform the computer standards. Reform > them to support Leap Seconds and we've advanced the state of the art rather > than returning to flat land. Yes. We need to reform leap seconds if there’s any hope of them remaining viable for the long haul. One way to reform them is to toss our hands up in the air, because we know eventually leap seconds are doomed to die due to the quadratic acceleration of the accumulation of the differences (wow, why didn’t I just say eventually, they will happen so fast that once a day won’t be enough). One way to help get rid of the pebbles is to make them more regular so people know they have to shake their shoes out more often. IERS likely can predict out to about 3 years a 90% or 95% confidence interval on DUT1. Publishing longer range leap seconds would be one way to make it more regular, but that’s still an observational calendar (people have to measure and correct often). Another thing we could do is say “we’d like to keep things mostly in sync” over the next hundred years. We could say “There will be a leap second every 18 months for the next 15 years.” and reevaluate 10 years out to correct any oversteering. With any PLL (and UTC is a PLL since it is steered to an unpredictable wobbly rock), the steering constant is important, as is the amount of error to steer out. With the current system, we know and measure the error to predict the best time for a steer. >> There are many better ways to implement them, and “perfect timekeeping” >> always begs the question of “perfect for whom.” De we align them with the >> day or the year? If we pick one over the other, the other gets out of sync. > The day, as has always been the traditional goal. That's why leap years and > UTC with Leap Seconds in the first place, right? Has it? Time of day precision smaller than hours has only been important since 1500 or so when clocks became accurate enough where minutes mattered and people started moving into the cities and working for others. This accelerated with the industrial revolution around 1800 as factories grew more numerous. Prior to that, they didn’t care what any clock said: it was light out, so you needed to be working the fields, it was dark, you should sleep. Most of the advances in time keeping were astronomers of various flavors understanding the heavenly bodies so they could predict the next eclipse or solstice. These events were much more important than time of day, since they governed when you’d plant your crops, harvest your crops, etc. All these events were annual in nature, hence the length of the year was more important until he calendar rules were standardized enough for people to forget about it as a solved problem. There are two different clocks regulating human activity that are running at different rates. Optimizing one will pessimize the other. Warner > -Brooks > >> >> Warner >> >> >>> -Brooks >>> >>> On 2014-08-20 04:02 PM, Preben Nørager wrote: >>> >>>> OMG its 290091200278565000. With THAT my proposal still stands :-) >>>> >>>> >>>> 2014-08-20 21:48 GMT+02:00 Keith Winstein >>>> <kei...@cs.stanford.edu> >>>> : >>>> Check that multiplication... :-) >>>> >>>> On Wed, Aug 20, 2014 at 2:42 PM, Preben Nørager >>>> <samp5...@gmail.com> >>>> wrote: >>>> >>>>> I can not edit the numbers in my initial post, but I can do it here, and >>>>> with that my proposel still stands: Drop the leap second, and continue UTC >>>>> without leap >>>>> seconds, so that 1 mean solar year is defined as the >>>>> duration of 290091175979732 [31556925,9747x9192631770] periods of >>>>> >>>>> radiation in the caesium atom >>>>> >>>>> >>>>> >>>>> 2014-08-20 16:43 GMT+02:00 Keith Winstein >>>>> <kei...@mit.edu> >>>>> : >>>>> >>>>> >>>>>> To be a pedant [but if you can't be one on the leapsecs mailing >>>>>> list...], the SI second is *9192631770* periods of the radiation etc. >>>>>> Your figure is high by 1000. >>>>>> >>>>>> On Wed, Aug 20, 2014 at 9:22 AM, Preben Nørager >>>>>> <samp5...@gmail.com> >>>>>> >>>>>> wrote: >>>>>> >>>>>>> In the discussion about whether or not to drop the leap second, I think >>>>>>> it >>>>>>> is not a question about solar time or not solar time. It is in other >>>>>>> words >>>>>>> not a question about either solar time or atomic time. >>>>>>> >>>>>>> >>>>>>> >>>>>>> If we drop the leap second it will be in favour of another timescale, >>>>>>> which >>>>>>> uses only atomic clocks to tell the time, but the time in that other >>>>>>> timescale will still be based upon a kind of solar time. >>>>>>> >>>>>>> >>>>>>> >>>>>>> About a hundred years ago it was decided, that the mean solar year, and >>>>>>> not >>>>>>> the mean solar day, should be the unit of international time. >>>>>>> >>>>>>> >>>>>>> >>>>>>> In 1960 the second was defined as 1/31556925,9747 of the mean solar >>>>>>> year, >>>>>>> and in 1967 the second was redefined [equally in length to the >>>>>>> previously >>>>>>> defined second] as the duration of 9192632770 periods of radiation. >>>>>>> >>>>>>> >>>>>>> >>>>>>> When the second was defined in 1960 it was defined as a fraction of the >>>>>>> so-called tropical year. That was a mistake of wording. The tropical >>>>>>> year is >>>>>>> a measurement of the solar longitude on the ecliptic, but the >>>>>>> international >>>>>>> definition of the second is not based upon measurement of the solar >>>>>>> longitude on the ecliptic. >>>>>>> >>>>>>> >>>>>>> >>>>>>> The definition of the second is based upon Newcomb's theory of the solar >>>>>>> system, and in that theory it is the barycenter of the solar system, and >>>>>>> not >>>>>>> the center of the sun, which defines the length of the solar year. >>>>>>> >>>>>>> >>>>>>> >>>>>>> The length of the solar year, according to Newcomb’s theory, is the time >>>>>>> for >>>>>>> the longitude of the barycenter of the solar system to increase 360 >>>>>>> decrees. >>>>>>> >>>>>>> >>>>>>> >>>>>>> The solar year, thus defined, can be measured either for one year, or >>>>>>> for an >>>>>>> average of years. >>>>>>> >>>>>>> >>>>>>> But the 1960 and the 1967 definition of the second can also be used as >>>>>>> an >>>>>>> international definition of the mean solar year. >>>>>>> >>>>>>> >>>>>>> >>>>>>> I think we should drop the leap second, and continue UTC without leap >>>>>>> seconds as TI [International Time], so that 1 mean solar year is the >>>>>>> duration of 290091231835491000 [31556925,9747x9192632770] periods of >>>>>>> radiation in the caesium atom. >>>>>>> >>>>>>> _______________________________________________ >>>>>>> LEAPSECS mailing list >>>>>>> >>>>>>> LEAPSECS@leapsecond.com >>>>>>> http://six.pairlist.net/mailman/listinfo/leapsecs >>>>>>> >>>>>>> >>>>>>> >>>>>> _______________________________________________ >>>>>> LEAPSECS mailing list >>>>>> >>>>>> LEAPSECS@leapsecond.com >>>>>> http://six.pairlist.net/mailman/listinfo/leapsecs >>>>> >>>> >>>> >>>> _______________________________________________ >>>> LEAPSECS mailing list >>>> >>>> >>>> LEAPSECS@leapsecond.com >>>> http://six.pairlist.net/mailman/listinfo/leapsecs >>> _______________________________________________ >>> LEAPSECS mailing list >>> >>> LEAPSECS@leapsecond.com >>> http://six.pairlist.net/mailman/listinfo/leapsecs >> >> >> _______________________________________________ >> LEAPSECS mailing list >> >> LEAPSECS@leapsecond.com >> http://six.pairlist.net/mailman/listinfo/leapsecs > > _______________________________________________ > LEAPSECS mailing list > LEAPSECS@leapsecond.com > http://six.pairlist.net/mailman/listinfo/leapsecs
signature.asc
Description: Message signed with OpenPGP using GPGMail
_______________________________________________ LEAPSECS mailing list LEAPSECS@leapsecond.com http://six.pairlist.net/mailman/listinfo/leapsecs
