Slashdot Mirror
The Future of Leap Seconds
@10u8 writes "Since 1972 precision clocks around the world have ticked using atomic seconds, but earth rotation is slowing down. Leap seconds have been inserted in order to keep noon happening at noon, but they upset some timekeepers. Recent discussions have considered
discontinuing leap seconds in UTC, and a colloquium in Torino next month will present results. It is a matter of international significance."
I don't think that is accurate. the concept of leap years is because the roman calendar sucks. The aztecs had a calendar that is much more precise, so did the Incas. The idea that time ticks at a constant rate is stupid anyways. Did relativity teach us that lesson already?
This site may be more helpful, especially in clearing up some of the problems with leap seconds (and their ultimate creation of an offset from both TAI and GPS time)
This is the link to a summary of the issues involved, written at a slightly less technical level.
(don't have to pay, don't have to register, etc.)
http://www.cl.cam.ac.uk/~mgk25/c-time/metrologia-
Precision: the `depth` to which something is measureed. Ie 1.01023921203 is more `precisely defined` than 1.01024.
Accuracy is what it sounds like - the amount of error in a given reading.
Next?
If I understand what I read correctly, essentially the problem they're trying to solve is this: the Earth's rotation is slowing, but they can't predict exactly how much it's going to slow at any given time. It is a real, physical thing, and while they can model its orbit with extreme and unchanging accuracy (things are widely separated enough that the mathematical abstractions work fine), modeling its rotation isn't really possible. There's all sorts of liquid sloshing around everywhere, both liquid water on the surface and molten rock in the center. All they can do is measure it, and every once in awhile, determine that sunrise is happening just a little late.
:-)
There are two major timekeeping systems: TAI, which is "absolute time" and is never adjusted, and UTC, which is "civilian time". Because UTC is used by normal people, they try to keep it synced to the Earth's rotation, which in theory at least makes it more useful for us mere mortals. (knowing that the sun will rise at exactly X time on X date at sea level, for instance.). So, gradually, UTC diverges from TAI, because one rotation of the Earth is just a little longer than 24 hours, and over time this divergence adds up to be greater than a second. When it's getting close, they add a leap second. These additions are not at regular intervals, because they can't predict exactly when any given second should be added.
There are occasional problems when they add the leap seconds (programs that don't expect 61 seconds in a minute, for example), or programs that don't realize that there are X number of seconds (15 or so?) that simply didn't exist since 1970. (sometimes this stuff matters).
Thus, they're debating about doing away with leap seconds altogether. One possible substitute is a 'leap hour' every thousand years.
It seems like a rather anal-retentive thing to argue about, but these people are paid to be precise to a degree we can't even imagine.
A worthy slashdot story. This is serious geekery.
The problem with leap days has nothing to do with the Roman calendar. It is because the time it takes the Earth to revolve around the Sun is not an integer multiple of the time it takes the Earth to rotate on its axis. The Lunar calendars you mention have leap-months.
Jason
ProfQuotes
requires a lookup table and regular (like every
two weeks) network connections to the Navy's
leap second table server to detect updates,
and the software needs to parse the table and
account for the update if and when it occurs.
Since we have not had a leap second update since
1999, it has meant there has been lots of time
for folks to get complacent and ignore the update
checks, so most recent code that handles leap
seconds is trouble waiting to happen.
I will be very happy when leap seconds are put
to bed.
time "ticks" at a constant rate in any particular reference frame (SR), which is how time standards are defined, anyway.
Satellite clocks have to be relativistically corrected, especially for applications like GPS.
Because time is crucial to all sorts of physical and scientific endeavors, such as planetary motion, navigation, GPS, etc. We need an accurate standard, or stuff quits working.
Here's an earlier (1994) proposal to do away with leap seconds and suggests among other things, leap hours every few centuries.
Leap hour proposal
It looks like the day is getting an average of 2ms longer per century, but it fluctuates 4-5ms away from that on a decade timescale plus some shorter-term noise.
A really interesting guy on this topic is Tom Van Baak, the fellow that runs leapsecond.com. As a measure of the level of obsession a person can obtain, this guy has multiple cesium frequency standards, but he had to go out and buy a crazy russian hydrogen maser so he could get better than a microsecond a year accuracy. He's also got some interesting information about the leapsecond debate on his website.
Me, I'm a simple guy, I just need to keep NTP locked to a couple of microseconds to sleep well.
jeff
> Perhaps at some time the Earth orbitted the sun once ever 365.0000 days instead of once every 365.2425 days.
Sometime between 7 and 20 million years ago the Earth was in fact orbitting the sun in exactly 365 days. But, what's really changing here is the length of a day and not the orbital period.
IIRC, all leap seconds when inserted or deleted, is well planned in advance. IIRC, GPS is prepared for these events also. Here's a link.
You are being MICROattacked, from various angles, in a SOFT manner.
The amount of time it takes is not constant. All we can do is mesure the rotation and let everyone know when it's changed.
autopr0n is like, down and stuff.
Right. The real problem is that lots of people who write code dealing with times and dates do a really crappy job. Some piece of software breaks every leap year, every time we change to daylight savings time, every time the dates of daylight savings time change, every time there's a leap second, every time you move your computer across time zones, every time a year divisible by 4 isn't a leap year... Just last month I reported a bug in a library function in a well-known software product, and had to explain to the developer that no, we weren't on DST yet, but Australia was.
So this leap second debate is really just a cunningly disguised way of talking about the crisis of software quality. The fact that a lot of software can't deal with 23:59:60 is the same problem that causes a lot of software to fall over when there's a February 29th.
GCHQ Quantum Insert installed. If only our tongues were made of glass, how much more careful we would be when we speak
This has to be the most tired, rediculous argument I continue to hear on a regular basis. No, the results of economic policy don't occur overnight. However, in the 12 years covered by the Reagan and senior Bush administrations, there was a general economic lull and by time Bush left office in 1992, the federal debt was over FOUR TIMES what it was when Reagan entered office in 1980. While government spending does encourage private spending, letting the federal government bleed cash profuciously IS foolish in the long run.
When Bush entered office in Jan 2001, the federal government was running a $127 billion surplus. The projected figure for the 2003 fiscal year is a $300+ billion DEFICIT. Simply put, the federal government will lose more money this year than it payed back off under the 4 years of surplus combined. How is this wise economic policy? The downturn may have started before Bush, but he hasn't done anything of significance to reverse it. And of course, Clinton had absolutely nothing to do whatsoever with the economic prosperity of the mid 90's. All those silly trade deals and the callaborative work he did with other nations most certainly did nothing to help at all. International agreements, who needs'em?
Not true. GPS receivers get all the information they need directly from the GPS satellites - which track their own "GPS Time" that dispenses with the leap-seconds.
You're right that having an accurate astronomically-relevant time is important for navigation - if you are determining your position with a sextant. It's the decreasing relevance of sextants to the world of navigation, and the increasing need to keep electronic equipment of all sorts in lock-step, that is driving this movement away from the leap-seconds.
See a summary of the issues from one of the US Naval Observatory scientists in charge of this stuff: PDF, Postscript.
-renard
Initially the precise measurement of time was the province of astronomers and ship navigators. Time was fundamentally the measurement of the orientation of the Earth. Time was a function of location. Noon was when the sun was at zenith. If you could know the difference in time measurements at two locations, you could determine the difference in longitudes of the two locations. In order to determine the differences in time systems, mankind developed precise mechanical time measuring systems. The new time measuring systems allowed man to measure the durations between events very precisely.
Eventually man developed atomic clocks that could use the decay of atoms to provide an incredibly stable time reference. However, some time ago, we reached a point where the mechanical time measuring systems became more stable than the Earth's rotation. So the atomic clocks which were counting down seconds very accurately were now getting out of synch with the Earth's rotation which was slowing down (and not smoothly slowing down, either).
Since no one who was concerned with the durations between events wanted seconds that varied in length, which is what would happen if you fixed the varying length of the day at 86400 seconds, the concept of using seconds of fixed duration (based on an atomic standard) was developed. The ever accumulating count of these seconds is TAI (Time Atomique Internationale aka International Atomic Time). The time which represents the orientation of the Earth is Universal Time (UT). (This is a simplification, there are a number of subtle variations on UT that I'm not going to go into, but which aren't important for the purposes of this discussion.)
If left alone, the difference between UT and TAI would grow. So, many years ago the concept of UTC (Universal Time Coordinated) was invented. This time standard uses the standard TAI second, but at irregular intervals, an additional second may be added (on either June 30th or Dec. 31st) to always keep UTC and UT to within half a second of each other.
The bottom line is that for people who have to deal with durations, especially long durations, having those irregular additional seconds is a bookkeeping pain and for those who need to be very concerned about the orientation of the Earth, a half second isn't nearly accurate enough. The latter group are undoubtedly using much higher resolution correction data that is produced by the IERS (International Earth Rotation Service). For most civilians, the fact that noon is shifting off by a second every couple of years just doesn't matter. (Especially since the railroads introduced the concept of time zones a little over a hundred years ago, which means that the sun is rarely at zenith when the clock says its noon.)
A lot of people in the field have questioned for some time whether in the era of modern computers where using the higher resolution IERS corrections is trivial, the leap second has any use. Now it may finally be going away.
Now if you want to get really esoteric, here is something to ponder: For astronomy and celestial mechanics, time is defined as the independent variable in the equations of motion of the universe. For physicists and those who use atomic time standards, time is defined as the independent variable in the decay of atomic particles. Noone, to my knowledge, has ever been able to detect a difference in these two independent variables, but it is not a given that they are the same.
For those who'ld like to know more, the University of Texas teaches a graduate level course in the Aerospace Eng. Dept. on the "Determination of Time".
This is not a legal opinion, no representation is expressed or implied.
There are several timescales. There is already one that does not have leap seconds, and one that does. What is important for the average person is that when the beep, beep, beeeeeeeeep goes beeeeeeeep it's the same for *all* people. While 22 seconds isn't a big deal for most people, it's a huge difference in a lot of other areas from financial trading to shipping. There's a hint of the fact that a USBN (submarine) hit something because a leap second got inserted in to a clock that no one was prepared to handle and they went a second to far.
.9 seconds of cosmic time.
The leap second reconizes the fact that the "second" is defined in terms of particle physics (a quantity of state changes) which is very stable (it's always going to take the same amount of time for the same quantity of state changes), where as the idea of time really comes from the cosmos. When the sun is directly overhead it's 12:00.
Where the earths orbit around the sun is very stable, 265.24 days, the rotation of the earth is very unstable. In fact, there's a provision (though never used) to remove a second from the day! The speed of the rotation is constantly changing. Over the long term it's pretty stable with a stable decay, but in the short it could be necessary to add a second rather quickly to keep the civil time within
The long term average is that we need to add a second to the day about every 18 months, but we haven't needed a leap second since the end of 1998 (over four years!) so in the short term the stability of the earths rotation is low compared to the order of magnitude we measure.
In order to handle this a desicion is made every six months as to a new leap second at the end of June or December (or to remove a second). This is a problem because some systems can't handle the addition of a second on six months notice such as the submarine!
One proposed solution is to allow UT1 (cosmic time) and UTC (civil time) to be out of sync by as many as 10 seconds. This would allow for ample time for warnings to be produced and everyone to know exactly what is going to happen and how to handle it. I don't know if the protocol would add 10 seconds at once, or warn everyone a few years in advance that a second is going to be added at several different points in time.
One interesting side note. Most computer systems don't handle leap seconds. Time keeping software slows the computers clock down (since it's important not to have events which have happened (past) in the future (future). This means that if your measuring anything else based on time that measurement is going to be wrong. The theory being that the accuracy in what time it *is* is more important than what time it *was*. The reason I bring this up is that time is something that can be measured with amazing percision, where as other things can't be measured as well. If you can convert one measurement to time you can measure it more percisely. For example, how fast does the ISS move? If you know it's altitude by measuring how long it takes to bounce a light off of it, and you know how long it takes to get from A to B (or from A to A again), you know how far it moved and how long it took to move and voila, speed, all by measuring time. If a leap second got thrown in while you weren't paying attetion during your measurement, your speed will be wrong.
Darthtuttle
Thought Architect
Because while the day is 86400.002 seconds long (on average) it's getting longer. About 170 years ago the day averaged 86400.000 seconds long. In 170 years it shoudl be about 86400.004, though the slowing down is caused by the moon, which is moving away from us, so it's effect will get less and less (though probably not that fast) but it could be as little as 86400.003. The point is, the second is defined by one measure and the day another, and we are trying to cram the two measures together when they don't have a linear relationship.
Darthtuttle
Thought Architect
Too late. In most places, local solar noon hasn't been used as a time standard for more than a century. Depending upon where you live within your time zone, the local solar noon can be different from standard time by a half hour or worse--and I'm not going to mention the impact of Daylight Saving Time.
Correcting--or not correcting--the time through use of leap seconds makes a difference of less than half a minute per century. The leap second correction is too coarse for almost any scientific work, and much too fine for the average person on the street.
Why not have a leap minute, where necessary, once every two or three centuries? It will still be dark at midnight, and we reduce the hassle of dealing with time discontinuities by a couple orders of magnitude.
~Idarubicin