from the time-that-takes-survey-of-all-the-world dept.
Arthur T Knackerbracket has processed the following story:
Google, Microsoft, Meta and Amazon launched a public effort Monday to scrap the leap second, an occasional extra tick that keeps clocks in sync with the Earth's actual rotation. US and French timekeeping authorities concur.
Since 1972, the world's timekeeping authorities have added a leap second 27 times to the global clock known as the International Atomic Time (TAI). Instead of 23:59:59 changing to 0:0:0 at midnight, an extra 23:59:60 is tucked in. That causes a lot of indigestion for computers, which rely on a network of precise timekeeping servers to schedule events and to record the exact sequence of activities like adding data to a database.
The temporal tweak causes more problems -- like internet outages -- than benefits, they say. And dealing with leap seconds ultimately is futile, the group argues, since the Earth's rotational speed hasn't actually changed much historically.
"We are predicting that if we just stick to the TAI without leap second observation, we should be good for at least 2,000 years," research scientist Ahmad Byagowi of Facebook parent company Meta said via email. "Perhaps at that point we might need to consider a correction."
The tech giants and two key agencies agree that it's time to ditch the leap second. Those are the US National Institute of Standards and Technology (NIST) and its French equivalent, the Bureau International de Poids et Mesures (BIPM).
This governmental support is critical, given that ultimately it is governments and scientists -- not technology companies -- that are in charge of the world's global clock system.
[...] Adding a leap second causes problems with computers. And at some point, we'd have to subtract one too -- something that's never happened -- and that would likely uncover new problems.
"It could have a devastating effect on the software relying on timers or schedulers," Byagowi and Meta engineer Oleg Obleukhov said in a blog post Monday.
Related Stories
Considering the recent thread on the potential removal of leap seconds, a story in TheAge aussie paper seemed worth adding to the discussion:
Earth had its shortest day since records began last month, with 1.59 milliseconds shaved off the usual 24 hour spin on June 29 - raising the prospect that a negative leap second may soon be needed to keep clocks matched up with the heavens.
The Earth appears to be spinning slightly faster than normal.
Usually, Earth's average rotational speed decreases slightly over time and timekeepers have been forced to add 27 leap seconds to atomic time since the 1970s as the planet slows.
But since 2020, the phenomenon has reversed with records being frequently broken over the last two years. The previous fastest day was -1.47 milliseconds under 24 hours on July 19 2020 and it was almost broken again on July 26, when the day was -1.50 milliseconds shorter. While the effect is too small to be noticeable by humans, it can accumulate over time, potentially impacting modern satellite communication and navigation systems which rely on time being consistent with the conventional positions of the Sun, Moon and stars.
It means that it may soon be necessary to remove time, adding a negative leap second, and speeding up global clocks for the first time ever.
Related stories:
Why One Critical Second Can Wreak Havoc On The Internet
5. 4. 3. 2. 1. 1... An Extra Second to See Out 2016
(Score: 2, Disagree) by bradley13 on Wednesday July 27 2022, @11:49AM (21 children)
They are basically right, except I would go farther and scrap the correction entirely (rather than "smearing" it). There is no reason to tie time so tightly to the rotation of the earth, especially since that rotation is not (quite) constant. If special applications like astronomy need to apply corrections for that, let them. For the rest of us, it absolutely does not matter whether the earth rotated a fraction of a second faster or slower this year.
Managing time is difficult enough without worrying about some minutes having 59 or 61 seconds.
Everyone is somebody else's weirdo.
(Score: 4, Interesting) by shrewdsheep on Wednesday July 27 2022, @12:12PM (20 children)
I tend to disagree. It is a simple engineering taks to get this right. There is one universal clock just counting seconds. Then, we tag certain values of this counter with what we call dates. We already handle leap years, adding leap seconds does not add complexity to a sane design (or minimal at most). Finally, we need to translate the two. All internal scheduling is based on the universal clock, no special cases need to be handled at that level.
(Score: 4, Informative) by maxwell demon on Wednesday July 27 2022, @01:01PM (13 children)
That's exactly how Unix time was originally meant to work, the system just linearly counting the seconds since Jan 1, 1970. Unfortunately POSIX decided to bind it to UTC instead, forcing leap seconds in the time stamp, instead of having the date/time conversion routines handling them.
The Tao of math: The numbers you can count are not the real numbers.
(Score: 2) by ls671 on Wednesday July 27 2022, @01:54PM (12 children)
It wouldn't work with computers going to Mars for example. See my post below about relativity. An atomic clock accelerating and decelerating will run slower. It is even said that astronauts would age slower. A fixed number of seconds since 1970 doesn't mean much in that context since time itself is relative.
Everything I write is lies, read between the lines.
(Score: 2) by acid andy on Wednesday July 27 2022, @03:14PM (1 child)
Could you define time in terms of what an atomic clock would show if it had always been at rest with zero gravitational forces acting upon it? You would need a way to very precisely measure and record the forces that are acting on your atomic clocks, other than by comparing readings on clocks.
Master of the science of the art of the science of art.
(Score: 2) by ls671 on Wednesday July 27 2022, @03:40PM
I wonder where you would then locate that clock, in the center of the universe maybe? Where is the center of the universe exactly?
Atomic clocks on planet Earth are moving along with the planet in space at 30 km/s and a point at the equator moves at 0.46 km/s due to Earth rotation, so the components used on those atomic clocks must have been accelerated at some point. A fixed point that has never been accelerated doesn't seem to exist.
In the end, time will always remain relative so we need a way to sync the clocks from some independent source or event. Things like the summer solstice remain a good way to sync your clocks with, at least for planet Earth :)
Everything I write is lies, read between the lines.
(Score: 1) by aafcac on Wednesday July 27 2022, @03:16PM (9 children)
The initial change is relatively fixed and something that can be calculated. After that it should be a matter of simply keeping up with the changes being made. Again, not that hard. This is mostly an issue while on the way to or from Mars, and it's something that NASA has been dealing with for decades with anything they put into space that needs to be controlled here or send messages back here. The little bit of drift isn't very large.
Chances are that you're not going to be able to run internet stuff directly over a connection to Mars anyways, the latency would be a massive issue if they don't figure out a quantum way of transporting the information over that distance more quickly than photons will allow for.
(Score: 2) by maxwell demon on Wednesday July 27 2022, @03:44PM (2 children)
Quantum mechanics doesn't allow for faster-than-light communication (unless you've got some FTL communication already by other means, then you can also utilize that for quantum communication). Indeed, that's one of the important no-go theorems of quantum mechanics.
The Tao of math: The numbers you can count are not the real numbers.
(Score: 1) by aafcac on Wednesday July 27 2022, @05:35PM (1 child)
It's a theorem, there have been a bunch of theorems over time that had to be changed or discarded. That doesn't guarantee that this is one of those situations, but keep in mind that just because the math says it can't be done now doesn't mean that it's correct and with how far ahead theoretical physics is of the experimental branch, I'd be careful about predicting too far into the future as it's a castle built on sand. We don't yet know if it's in sand form or concrete form.
(Score: 2) by maxwell demon on Wednesday July 27 2022, @05:43PM
The quantum mechanics is the maths. It may of course be that nature deviates from quantum mechanics, but whatever effects do so are, by definition, not quantum effects.
The Tao of math: The numbers you can count are not the real numbers.
(Score: 2) by ls671 on Wednesday July 27 2022, @04:34PM (5 children)
It's not only an initial change. Clocks on Mars and Earth will keep on drifting from each others as Mars and Earth accelerate and decelerate relatively to each other while following their own dedicated orbits. Relative speed between Mars and Earth varies between 6km/s and 54km/s, depending on where the planets are in their own orbits.
https://www.quora.com/What-is-velocity-of-Mars-relative-to-Earth [quora.com]
Again, time is relative and counting seconds since Jan 1 1970 as a solution is as foolish as using the king's foot to measure things! What happens when the king dies?
https://sparkfiles.net/foot-whats-special-12-inches/ [sparkfiles.net]
Everything I write is lies, read between the lines.
(Score: 1) by aafcac on Wednesday July 27 2022, @05:32PM (2 children)
There is no perfect solution. Making the Earth the center only really works if the Earth has everything revolving around it, which is clearly not the case. If this level of precision is needed, either you'd abandon the need to keep it in sync between the two planets, or, you'd likely redefine your reference to be at the sun, then account for the deltas from that. Given enough computational power, you could probably do the same thing with the galaxy and universe if things get that spread out.
As far as the king dying goes, that's not a relevant question. Even while that particular king is alive, you can't exactly expect the king to come out every time something needs to be measured. Eventually some standard for comparison emerges and you stick with that until there's some reason to change. In this case, you'd probably just use whatever you can estimate it to be at the sun using the measurements on both planets to identify the correct time and then use that. That is assuming you even need to have them be consistent.
(Score: 2) by ls671 on Wednesday July 27 2022, @06:19PM (1 child)
Right now I can't think of a single reason why using NTP between Mars and Earth wouldn't work.
NTP tests round trip time before adjusting the clocks, that's all you need to know. Well, maybe some slight modification to the tolerance for very long round trim delays and some loss of precision due to position changes between the round trim tests but I think it might work.
My point is that counting seconds since 1970 with two non-linked atomic clocks, one on Mars and one on Earth wouldn't work because I doubt the clocks would stay in sync even if they kept in sync when side by side on planet Earth.
I haven't done any math on that one although :) The only problem I can see is if the planets move too fast and in a too irregular pattern relatively to each other for NTP to correctly evaluate the delay but even then, ntp is amazing and could maybe provide good enough results by averaging.
If not, setup a NTP server close to the sun :) LOL! enough for today I guess...
Everything I write is lies, read between the lines.
(Score: 2) by ls671 on Wednesday July 27 2022, @06:41PM
Or set up a NTP network across the solar system with a few satellites orbiting each planet or even the Sun then, conquer the galaxy, then the universe :) No way autonomous atomic clocks will stay in sync counting seconds from 1970 even if they stayed in sync while sitting side by side on planet Earth IMHO.
Everything I write is lies, read between the lines.
(Score: 4, Informative) by maxwell demon on Thursday July 28 2022, @05:45AM (1 child)
You only need to define a reference. Then you can correct the clock for different speeds and gravitational potentials (fortunately the orbits in the solar system are highly predictable). It's a problem that had to be solved already here on Earth, in particular (but not only) with GPS.
TAI is based on the time flow at terrestrial sea level. There's also TCG, which is the time shown by a hypothetical clock at the center of the Earth.
For use in the solar system, BCT is probably the best option; it is the time shown by a theoretical clock co-moving with the solar system's center of mass.
The Tao of math: The numbers you can count are not the real numbers.
(Score: 2) by ls671 on Thursday July 28 2022, @06:42AM
Very informative! I wasn't aware of BCT. Date/Time is relative but everybody can indeed adapt to a reference point. See other post where I suggest an ntp network in the solar system, then galaxy, then universe. :)
Everything I write is lies, read between the lines.
(Score: 5, Informative) by ls671 on Wednesday July 27 2022, @01:24PM (2 children)
I disagree as well as you do!
Also, time is not simply "tied to the rotation of the earth" as OP says. One day is 24 hours and 1 rotation of the earth is ~23h56m.
If either the rotation of the earth OR its orbit period around the sun vary slightly, we need to adjust our clocks if we want to keep in sync.
Earth rotation period is almost exactly 23 hours and 56 minutes, ~23h56m04s. Yet, we have a 24 hours day due to the orbital period around the Sun.
Here are some numbers with bc. Background: if the Earth was tidally locked like the moon, it would still rotate on itself once every orbit so we need to take into account 1 rotation every orbit. (For the moon, orbit period == rotation on itself period, ~28 days):
23*3600+56*60+4
86164
(86400-(86164/365.25))/3600
23.9344710624 # 23 hours
0.9344710624*60
56.0682637440 # 56 min
0.0682637440*60
4.0958246400 # 4 sec
In short, the daily noon alignment with the Sun arrives later because the Earth is also moving around the sun so the Earth needs to make a little more than one rotation to align with the Sun again from day to day.
Everything I write is lies, read between the lines.
(Score: 3, Informative) by maxwell demon on Wednesday July 27 2022, @04:15PM (1 child)
Note also that the day (as in time between two consecutive midnights) is not constant length because of the slightly elliptic orbit of the Earth. During Northern hemisphere's summer Earth is farther away from the sun than during Norther's hemisphere's winter, therefore it is moving more slowly along its orbit.
Also, the speed of rotation relative to the stars isn't completely constant either because the mass distribution of Earth changes due to geological processes.
The Tao of math: The numbers you can count are not the real numbers.
(Score: 2) by ls671 on Wednesday July 27 2022, @04:58PM
Duly noted! Thanks!
Everything I write is lies, read between the lines.
(Score: 3, Interesting) by ls671 on Wednesday July 27 2022, @01:46PM
Also, I would be more tempted to agree with "smearing" than simply drifting. Google has been doing just this since 2008 and their systems seems to get along fine with it.
https://developers.google.com/time/smear [google.com]
It might affect some benchmarks but hey, didn't Einstein say that everything is relative? So, we may as well get along with the program!
Einstein must have been right because a guy on the NTP forums owning two atomic clocks that never drifted from each other tested it taking one on a road trip across the USA and when he got back home, the clock had indeed drifted from the one that stayed home. Accelerating and decelerating affects how fast time pass.
The clock which accelerate and decelerate will run slower.
Everything I write is lies, read between the lines.
(Score: 3, Interesting) by hendrikboom on Wednesday July 27 2022, @03:52PM (1 child)
The occurrence of leap years is known for decades in advance of the extra day.
But the insertion of leap seconds is not known much in advance. Keeping the world's clocks synchronised would require a very thorough distribution of leap second information, which may not be easy.
(Score: 3, Informative) by Zinho on Wednesday July 27 2022, @04:50PM
There appears to be plenty of notice in the system we already have in place. Quoting from a leap second reference for Red hat: [redhat.com]
When that announcement gets made, distros push a new version of the tzdata file and NTP servers automatically push leap second data to their clients. If you are running Linux and either keep up to date with patches or are synchronized to a time server then the leap second gets updated automatically for your OS.
"Space Exploration is not endless circles in low earth orbit." -Buzz Aldrin
(Score: 2, Interesting) by exa on Wednesday July 27 2022, @12:10PM (1 child)
...imagine the fireworks that will ensue upon the future "delayed" correction. After you let the poor fortune500 companies and their poor enterprise software happily catch bitrot for 2000 years.
(Score: 2) by bart9h on Wednesday July 27 2022, @01:52PM
in 2000 years the delay would be around 17 minutes, if I did the math right
(Score: 5, Insightful) by pkrasimirov on Wednesday July 27 2022, @12:14PM (3 children)
First handwriting recognition appeared and instead of algorithms improving and adopting to recognise our handwritings better, the people learned how to write so it can be correctly recognised by the software. Then came the speech recognition, we adopted again. It seems they took this process for granted and now want the Earth rotation to adopt to the software... lol. Somebody thinks they are the center of the world, literally.
(Score: 4, Insightful) by acid andy on Wednesday July 27 2022, @12:48PM (2 children)
Yeah I very much agree. At least with the handwriting and speech recognition they've got the excuse that they're tough computing problems for which an ideal solution may not yet be fully understood. Whereas this leap second thing is just like WAAAAAH, "WAAAAAAH, the rules are too
hardexpensive for us to handle properly! Make them go away!"Yup, this is when you know they've become a seriously powerful monopoly. We don't develop the product to fit the market, now we forcibly bend the market to fit our product. I guess part of it comes back to what we always say though that most of the users of their software are really part of the product so seen as fair game to be developed. Although the implication that the planet itself is the one that should change or rather be consigned to irrelevance is going a bit far! God complex?
Master of the science of the art of the science of art.
(Score: 3, Insightful) by coolgopher on Thursday July 28 2022, @01:11AM
I'd much rather ditch DST globally than ditch leap seconds.
(Score: 1) by arcade on Thursday July 28 2022, @05:42PM
The point is that there isn't good rules around when leap seconds are to be inserted or removed, there are only good rules on how it should be done. Rules which have been incompatible with most computer systems for the last 40 years. Furthermore, it's not only Meta, Facebook and big corporations that face the problem of leap seconds.
Inserting leap seconds causes fragility in the world that lead to utterly stupid outages left and right every time they happen. And we have no clue what'll happen the moment we'll remove one, if that happens at some point.
(Score: 2, Disagree) by acid andy on Wednesday July 27 2022, @02:39PM (3 children)
Think about the adoption of SI units for a moment. A big advantage is that you can always derive the quantities used to define the measurements by accurate measurement of physical constants. It strikes me as much more of an open standard than one that can't be determined without consulting the organization that controls it (Remember the official metre bars made in the 1800s).
In the same way I feel like time should be something that someone can calculate precisely through observations of the relative motion of the Earth and Sun. If the governments and standards bodies all go away one day, or even decide to deny access to this information (less likely but times are getting so crazy nothing would surprise me), you should still be able to work out what time it is, down to the second.
From Wikipedia:
Someone needs to invest in better programmers. If they're coding anything that's safety critical, it should be part of the specification and testing.
Master of the science of the art of the science of art.
(Score: 0, Informative) by Anonymous Coward on Wednesday July 27 2022, @05:59PM (2 children)
This is ironically an argument for why Fahrenheit is better than Celcius. Units of temperature are called "degrees" because analog thermometers were calibrated by having the needle be 180 degrees between boiling and freezing. (212-32=180).
Of course you can sort of do that with Celcius, but only if you're using a "metric degree" that divides a circle in to 200 sectors. They even briefly tried to introduce such a unit in to the metric system, but it failed. The SI standard for angles is radians which is cumbersome for every days use.
Hence the misnomer, "degrees Celcius", which is grating on the ears to anybody who knows the real story.
(Score: 2, Informative) by Anonymous Coward on Thursday July 28 2022, @02:25AM (1 child)
Sigh.
No, that's not actually what happened, despite the fact you've been modded "informative."
"Degree" came to be used in English as a term for the measure of heat (and cold -- people used to talk about "degrees of frost") by the 16th century, long before Fahrenheit was born. Fahrenheit got it from the Rømer scale [wikipedia.org], which was one of the first calibrated temperature scales. And the choice of divisions was 60, which was the number of degrees between the lowest attainable temperature of freezing brine and the temperature of boiling water.
The competing temperature scale at that time was Newton's, which set 0 at freezing point of water and 12 at body temperature.
Fahrenheit decided to use the body temperature idea, but Rømer's scale had too many fractions (freezing was 7.5 degrees). So, he basically multiplied Rømer's degrees by 4ish to avoid fractions and give more gradation to the scale, then recalibrated to the lowest 0 temperature he could achieve in brine, with body temperature at 96F (that was his estimate at the time). He then fiddled a bit to make the difference between body temperature and freezing (96-32) to be exactly 64 degrees, so it made it easy to construct thermometers by repeated division in half.
The 180 degrees between 32 and 212 is a coincidence.
The degrees of heat/frost never had anything to do with circles, so it's a bit grating to me to see you getting so worked up about this. Maybe at some point some analog circular thermometers were calibrated in the way you say, but if so that's a really late (20th century) development -- those types of thermometers didn't exist historically, as the standard type historically in the 18th century (when Fahrenheit and Celsius actually designed their temperature scales) were mercury-based with a rising column of liquid.
TL;DR - The idea of "degrees of heat" long predates any association with circular thermometers. And there are no "real degrees" -- just several different arbitrary scales which used different fixed reference points.
(Score: 1, Informative) by Anonymous Coward on Thursday July 28 2022, @02:33AM
Oh, and by the way, Fahrenheit's scale was recalibrated after his death to make the 32F and 212F match up more nicely with the actual temperatures for freezing and boiling (since Fahrenheit's original scale didn't match those numbers closely)... to make it easier to align and calibrate and convert to the already-existing Celsius scale.
(Score: 3, Insightful) by ElizabethGreene on Wednesday July 27 2022, @05:40PM (1 child)
... because programmers make fragile systems.
I saw a mail system, and not a small one, that would hard hang if time moved backwards on a system. That's simply not ok. Similarly, routing systems should not crash, hang, or reset if they get an odd timestamp as input. Blaming the leap second feels like shooting the messenger.
(Score: 0) by Anonymous Coward on Sunday July 31 2022, @04:02PM
(Score: 2) by SomeRandomGeek on Wednesday July 27 2022, @10:10PM
For most purposes, we use time zones anyway. And daylight savings time. Which means that our clocks routinely drift from astronomical time by as much an hour and a half. More in some places. Nobody cares to the nearest hour how the clock time aligns with position of objects in the sky unless they are doing astronomy or navigation. So, why inflict leap seconds on the rest of us? Leave leap seconds to astronomers and navigators, and stop re-defining the clock that everyone else uses.
(Score: 3, Interesting) by jb on Thursday July 28 2022, @03:04AM
If four of the dodgiest companies on earth support something, that's a fairly good sign that it's a bad idea.
Nice cover-up attempt ... but the real cause of those problems is miserly companies (like the same four behind this initiative) hiring inexperienced software engineers and paying them peanuts, so that they don't know how and couldn't be bothered to figure out how to do it properly. Yes, handling time correctly in all cases is non-trivial. So is anything else that's worth doing (in SE or in life generally). The solution is not to redefine time (for goodness sake!). The solution is to fix your broken organisations by instilling a strong engineering culture (which is not possible whilst the marketers remain in charge).
(Score: 2) by crm114 on Thursday July 28 2022, @11:12AM
The government I live under decides to shift time by a full hour twice a year. My $60US GPS receiver lets me know when I'm going to get an extra second when I'm asleep.
Fighting over a single second when people are planning on going to the Moon, or Mars, or beyond. THAT is when you have my attention as to time standards.
https://www.eecis.udel.edu/~mills/ipin.html [udel.edu]