The earth turns as it pleases.
And computers don't give a damn. But the question, looking at the technicality of the thing, is much more complex, even if in the end it is reassuring. Since META has also been put into it, I would say that it is appropriate to clarify a few things before examining the difference between "clock" and "calendar". Because the rotation of the planet does not change the time at all: it changes the calendar.
Let's get rid of META's charlatans first. If the system you have designed depends on external physics so much that an event of 0.0001% of magnitude can cause very serious damage, you have designed it badly, your architects are incompetent, and you demonstrate the same cialtroneria that you demonstrate when facebook is down for HOURS , which should NEVER happen to such a replicated system. Have you ever seen the internet, or the cellular network, or the telephone network, go down for hours for two billion people? No.
Having said that I don't care what the amateurs of META (1) say, let's explain for a moment the difference between time and calendar, in which computers care a lot about time but little about the calendar, while systems like GPS rotate the earth, that is the calendar, matters much more.
If we refer to non-relativistic physics (so let's exclude GPS from the discourse for the moment), that is a relatively local physics, computers need the time to make sure that the events are in order. It means that if I close a file BEFORE saving the last piece I wrote, the piece I wrote will not be saved.
If I have a queuing system, where events are processed in order, and my computer clock goes backwards, some of these events will seem to have started in the future, and my queue will have to do its work in figuring out which order to put them. .
If we think about a transaction, it goes without saying that it starts BEFORE it ends, and if we turn back the clock, the problem is that it could (apparently) end before it starts. Which would make it rather difficult to interpret.
So, if two computers think they live in two different times because their clock is out of phase, and the first starts a transaction that ends in the second, this can cause problems. Usually the solution is that those who have the "journal", that is, those who take note of the local time, always use their own clock: this however means that two computers involved in the same transaction but out of phase in time can have two different journals, and this means that in the event of a forensic investigation, the thing must be evaluated.
But this no longer happens today, because time (not the calendar, attention) is spread on the net. Cell phones were once used, and although they have GPS today (almost all, or at least many) cell phones also use NTP (bizarre choice, I must say), and here we have to stop to understand how NTP works.
The first concept to understand is the "stratum". Let's say there are very precise watches in the world. The atomic ones. And suppose these watches make their ticks available on the internet. Is this a correct assumption? It is a correct assumption. Here are some of the watches, and here is their conventional denomination.
(as you can see there is also Galileo and there is GPS, which I will talk about later because in the discourse "time vs date" they occupy a special case.)
The other sources distribute the time, the simple pulse per second, with different modalities. For example, DCF is used both via radio (they also sell radio clocks based on this signal) and on coaxial 1588 by major German telcos. The 1588 coaxial is then connected, in the Central Office of the telco, to the service edge switches, where the LIB (Lawful Interception Box) is usually attached, so that the police, to produce evidence of a telematic crime, always have the exact time. (as far as the impact of nanoseconds escapes me, in the logic of human actions, but so be it).
But we are talking about a specific thing.
The weather. NOT THE FUCKING CALENDAR. Not THE DATE. They are two different things.
However, we call that layer “stratum 0”, because it is physical and at least originally has no latency problems.
If I attach my computer to these clocks, I attach myself to the more accurate clocks, but we forget that the signal takes time to get from my computer to the device network on stratum zero, and then come back. Therefore stratum 1 is LESS precise than stratum zero, even if by very little because that time (even if small) can distort very precise measurements. (for example, someone in a physics laboratory, in measuring very rapid events, could also conclude that they have seen a phenomenon go back in time).
However, apart from these physicists who need very precise measurements, stratum 1 is considered accurate. Also why, otherwise you would have to bring an atomic clock on the spot.
Now imagine attaching yourself to computers on stratum 1. You are stratum 2. So you are even more inaccurate. If you stick to stratum 3, you add more latency, and you are even more inaccurate yet.
On Linux you can visualize this situation very simply:
If you see the “st” column, you get that the “clitosaurus” computer is connected to two stratum 1 clocks which are “PZF” and “PPS”. PPS is the pulse per second, while PZF is this gentleman here:
https://www.meinbergglobal.com/english/products/rack-mount-1u-ntp-server.htm
the others, as you can see, have a higher stratum and as you can see they have different offsets.
The system makes 16 surveys, measuring the average phase shifts, and chooses the clock that it finds best. In the case above, he chose PPS. Foodie.
This is how computers see time, and they synchronize with each other over the network. You will say: and what does the way the earth turns? This seems an "absolute" time (ARGH!), And in any case it seems disconnected to the rotation of the earth, and connected to atomic clocks.
That's right, the rotation of the earth has nothing to do with a pipe.
However, you will tell me, if I go to write "date", which uses the PC clock, then I get this:
Apparently, there is a big dependence between the calendar and the time, but it is only apparent. I don't want to go into the meaning of CEST: it would be confusing and just a bulgar display of power, as the Pantera overdosed on goulash said.
The problem instead is that "August" and "13", as well as "13:25:15" are linked to the clock only through another physical entity: the planet earth. When we say that it is "August", a summer month, we say that the earth is at a certain inclination and in a certain stage of its orbit around the sun. When we say 13:25:15 ″ we refer to the rotation of the planet.
So we have TWO distinct concepts, both equally "physical events":
- time as an ordering of events
- the calendar or "time" if we consider "13:25:15"
The calendar (or date) is nothing more than a way to indicate the state of the planet over time. And there can be many ways: we have the Jewish calendar, the Orthodox one, and so on. For instance:
Both calendars are correct, as long as the necessary conventions are activated. 3188 is as accurate as 2022, which is based on Christian mythology , while 3188 is based on Greek mythology, of the goddess Eris.
But time, at least locally, passes by one second every second. It is the computer that transforms the time it receives from its sources into a calendar date.
But the clock is the clock. The date is the date. (I will not rage with the timezones, because they have nothing to do with the speech).
So what's the problem for computers if planet earth goes out of business? None, per se. None because the computers on the network do not align the calendar, but rather the time. At best, the problem comes when the planet is assumed to interpret time on a regular basis ( and now we know it does NOT) and sometimes one day lasts less than another. End.
Of course, if you want a computer to also take into account the fact that today the midnight of the planet has moved and REPRESENTS time more accurately, then you should download the accurate conversion tables and do your conversion with different tables from moment to moment. moment.
Or base your computer's timestamps on unix.nano () time, so as to depend as little as possible on physical events and external fluctuations.
In summary:
- you measure local time with atomic clocks
- you measure the date by estimating the state of the planet at a certain moment.
both (atomic clocks and planet state) are physical events . The fact is that these two events are not regular in relation to each other , and therefore the conversion can fail.
What's the problem'? NONE: the difference in duration between time-planet and time-atomic clock is absolutely absorbed by the delay and jitter you have when you connect to an NTP server. Therefore, for the computer it is imperceptible, except for computers connected directly (and with little latency) to stratum 0, for example via coaxial 1588.
But… there is a "but".
You will have seen that in the list of sources (stratum zero) of stratum 1 there is the GPS, and also "GAL", Galileo. (in reality you can also use the Russian Glonass network or the Chinese one).
The problem with the time taken by satellite networks is that on the one hand it is very precise, if we talk about time, but since satellite networks are affected by the gravity of the planet (the field is not perfectly spherical because the earth has no homogeneous or symmetrical density) and the satellites are built to continuously make small changes of course to compensate for the variation of the field, in fact they recalibrate continuously.
Obviously, when it happens that the planet thickens in the center and turns faster, the satellites recalibrate the course, and this could also cause some slight delays in your GPS, GAL, GLONASS; etc.
But is GPS used in computer science? Well, of course you know that each of your cell phones has a GPS, but on cell phones it is not used to receive the time signal (generally, although it would be rational).
For work computers there are GPS / GALINEO cards, which you attach to the USB and can align the computer extremely well, on stratum 1. (due to relativity the stratum zero is on the satellite. Otherwise the GPS would not work).
Since most computers are indoors and have no view of the sky, they are normally only used by people who really need them, or for outdoor applications, when you don't want to bring a coax cable from the nearest atomic clock. (which can be uncomfortable).
Perhaps these applications would be affected, of course. But generally, those who make gps-based applications and need precision, know well what the limits of accuracy are.
So why do META's mediocre trombones complain?
The only possible explanation for this complaint is that META makes REALLY extravagant use of cell phone location data. Or incompetence.
Why do I say "incompetence"? Aren't they the best minds on the planet? No.
The point is, they used to be. They only hired to be a janitor if you had a PhD. The problem is that there is no infinite supply of PhDs. And when on your CV, after the PhD, it remains only that you have been a porter at Google for three years (the average turnover time at most GAFAMs), PhDs have begun to avoid GAFAMs.
GAFAM asked you for immense academic knowledge, and then bumped into doing javascript on some frontend, or concierge, or human resources support.
Over time, therefore, it happened that the "best minds" began to avoid them, and those who had (due to their turnover mechanisms) went elsewhere.
The result is that they specialized in looking for "talent" even where there was no academic qualification. But in reality they have simply fragmented the work so as to require only monkey-employees, who on average receive instructions from automatic systems, and work in an alienating way. They don't have the best brains anymore. If you have an IQ of 190, don't work for Google: consult with Gartner. And sell slides to google monkeys.
So I would say calm down: the earth has always been asymmetrical, those who work in satellite control stations know it since Sputnik, the gravitational field is not perfectly spherical, the rotation has never been perfectly constant, and the only The problem is that Facebook's "minds" are no longer the best in the world.
Amen.
(1) I call them amateurs because they have an amateur SLA. If we take such a thing as the emergency call, it is guaranteed to work all over Europe with 100% uptime. Because it goes down for a minute, between calls to the police, firefighters and ambulance, count the dead by the thousands. Facebook has also been down for HOURS. When it comes to uptime, they are amateurs in trouble. Point. If the emergency call for two billion people were to fail for six hours, you would count hundreds of thousands of deaths.