The rotation of the Earth is a terrible timekeeper. We used to have 6 hour days. Now it wobbles. Major earthquakes throw it off. On the other hand, ignoring this leads to weird results. In Rome, the priests used to control the Calendar, and they had to decide how many leap days to put into the cycle. Senators served through the end of the year, and they had influence over the priests. By the time Julius Caesar implemented a new calendar, this sort of political drift in the calendar had January in the middle of summer.
I guess people don't like mid-summer January because it makes historical comparisons more difficult. Not impossible, just awkward.
It's a horrible messy dilemma. Leap seconds don't seem as dramatic as the Roman intercalary days, but there's the same battle between synchronicity and simplicity (where simplicity = something short of insane perturbations).
Maybe none of this matters for radio broadcast time signals. Maybe you just want a fixed counter everyone's synchronizing to, and people should convert to the Earth's messy system at the last possible moment.
Keep in mind that leap days serve a different purpose than leap seconds. Leap days exist because the number of rotations of the earth per one revolution around the sun is not a whole number. Leap seconds exist because the duration of one rotation is not constant.
I have no idea how often leap seconds occur, but if we were to say that there should be one leap second every year and we instead ignored them, it would take 43,200 years before that happened.
It would take 60 years of skipping yearly leap seconds before we saw even a 1 minute offset.
TAI (weighted average of atomic clocks, correcting for relativistic effects due to altitude) is the basis for UTC, and is currently 35 seconds ahead of UTC.
Prior to 1972, UTC and TAI used different definitions of a second (UTC used a definition based on positions of celestial bodies while TAI used a definition based on the hyperfine transition of cesium, and the two were periodically manually "stepped" back into sync). As of 1972 they use the same definition of the second (TAI's definition, which is also the SI definition), and a final "step" was applied to cause the difference in the two times to be an integer number of TAI seconds. That created an offset of 10 seconds, which has since grown with the addition of 25 (and soon to be 26) positive and zero negative leap seconds to UTC.
The rotation of the Earth is slowing down, and this slowing down is expected to accelerate in the future. So in several thousand years, we will need multiple leap seconds per day to keep up. Whether we should take this into account now is the question, though.
> [...] it would take 43,200 years before that happened.
That's how long for noon to end up at midnight. Noon would occur at night after only around 22000 years (I'm being fuzzy because people may not all agree on what night means).
Do earthquakes also change how long it takes for the earth to rotate around the sun? Because it seems like the earth's orbit should be more or less constant (ignoring relativistic effects).
Earth's orbit is probably variable, but not because of earthquakes; rather, when the gravitational pulls of other Solar System bodies are consisdered, there are forces other than the Sun's gravity acting on the Earth. The most notable of these is the moon, but in terms of tangible orbital disruption, other planets (particularly Jupiter, which is massive enough to pull the Sun-Jupiter barycenter outside the sun, meaning that Jupiter orbits the sun much in the same way that Charon orbits Pluto) can and do disturb the orbits of their fellow planets, often pulling them whichever way.
Earthquakes are also not the only means by which the Earth's rotation can be affected. The moon's gravity creates all sorts of tidal forces on Earth; such forces don't just create tidal waves, but can influence everything from Earth's rotation to continental drift.
Of course, hypothetically, a sufficiently-violent volcanic eruption could end up working like a giant rocket engine and thus push the Earth in some direction or other (in much the same way as normal rocket engines do; in this case, magma/lava/steam would be the propellant(s)). That's the only way I can think of off the top of my head for earthquakes to affect Earth's orbit, and should that happen, I highly doubt UTC being synchronized with Earth would be particularly relevant at that point, since something of that magnitude would probably significantly damage (if not destroy) Earth, and would almost certainly exterminate anything unfortunate enough to still be living on the planet.
No it does not. Earthquake changes the internal structure of the earth, which offset its rotational rate (as angular momentum must be conserved and the moment of inertia changed).
It does not change the angular momentum of any system, nor does it change the moment of inertial of the sun-earth system.
> the awesome 13 month Calendar that we could have adopted
That calendar would have never worked. You would have to keep changing when the weekend is each year as the real Saturday/Sunday would shift by a day each year relative to the calendar. Some people would not care others would, and it would be a mess.
And don't just flippantly suggest "so don't shift", the 7 day pattern of weeks has been kept by humanity for thousands of years by cultures all over the world. It's not something so easily dropped.
If I'm reading the linked projections correctly, even the worst predicitions put us at less then three minutes off from the earths rotation by the year 2100, and it will take somewhere in the area a thousand years to be off by an hour.
I'll be frank: if we haven't progressed to the point that those offsets generally don't matter by those points in time, we've probably already failed as a species.
Offsets "generally don't matter" now, which is why this is an obscure bit of minutiae you're reading about on HN instead of NYT.
But they matter to some applications, and those people have arguments trying to decide which to use.
But yes: the ultimate solution if the world were to move to atomic/absolute clock synchronization instead of referencing the Earth's rotation is that every few hundred years we would need to adjust all the time zones by an hour to keep "noon" in the middle of the day. Given that we do this every decade or two anyway chasing DST vogues, I think this would be acceptable.
It is the heavens that set the time for humanity, all else must either imitate or pick another name. Abandoning the leap seconds means breaking the connection between utc and what the people consider time to mean.
It is not so much that the actual times drift apart very fast, it is that the clock tries to corrupt the meaning of the word "day".
If leap seconds cause problems for your application, you probably should find another source of synchronisation signal.
Plenty of applications need to know a precise number of seconds (or milliseconds or nanoseconds) between the present and a past or future date/time. Leap seconds throw a wrench in that. Whether UTC should serve the role of an unaltered time source is indeed a matter of debate (and for good reason), but it's pretty clear that we do need an unaltered stream of seconds (upon which UTC - or an intermediate between UTC and various timezones should UTC fill this role as advocated - would base itself, as would any other applications that require said unaltered stream of seconds).
This isn't even mentioning more far-fetched things like space travel (at which point the non-constant rotational parameters of an arbitrary planetary body are neither relevant nor appropriate for system-wide/galaxy-wide/universal/multiversal timekeeping).
The deal is the people chartered with the task of providing an accurate timebase are failing to do so because they are adding and subtracting arbitrary units of time. So if you actually need a monotonic timebase, you can't get it without having a human diligently subtracting leap seconds at random times. The proper thing to do is to generate a proper monotonic timebase and then publish a separate offset with microsecond or better resolution. Thus people that need monotonic time are happy and people that want to set the wall clocks are happy.
Also seriously, leap seconds suck as an offset because a second is a huge amount of time for many applications.
At root, this is all just a quibble over terminology. The fundamental problem is that there are two reasonable time standards for different applications: mean solar seconds for ordinary day-to-day timekeeping, and cesium seconds for applications that require a precision of more than a second per year. These two standards drift, so you have two choices:
1. Let them drift
2. Introduce a computational complication (leap seconds) to synchronize them
But the dispute is not over which of these is "right", the dispute is over which of these two options gets to go by the label "UTC". This could be easily resolved by simply defining two UTCs, one with leap seconds, and the other without. You could call them UTC-L and UTC-N (or UTC-S and UTC-N if you wanted to mollify the French ;-).
...and just to add that this enshrining in law was most likely done with the general expectation that UTC would keep a close relationship with solar time.
I should imagine it would be easier to use the cesium seconds for UTC and just bundle everything else into the time zone offsets, which everyone is already using to turn scientific timekeeping into something "meaningful" for human consumption.
Would that actually alter the spin of the Earth? I assumed that you'd be more likely to have more success by making a few passes past the planet with a sizable asteroid.
I don't know enough about any of this to judge which one is more feasible. Detonating nukes on the surface is "easy", but would it be effective? How much energy and money would it take to divert some asteroids? I haven't done the research, and almost certainly won't :P
I can't speak for the rovers, but in my experience you start with the UTC time and the convert to a UT1 for space launches. We have to correct for the difference, which is called DUT1: https://en.wikipedia.org/wiki/DUT1
Because that's everywhere. Whether everywhere should "just use TAI" is exactly the question (only they're not talking about TAI as far as I can see, but a new standard which would I think map 1:1 to it).
TAI currently differs from UTC by 35 seconds, so if you literally switched to it, every clock in the world would jump 35 seconds. If leap seconds are abolished, UTC and TAI will remain 35 seconds apart forever.
And then that time will be neither UTC not TAI. Either we'll have to call it something different (like GPS time: always offset from TAI by certain number of seconds), or a different name will have to be made for what used to be called UTC (time tied to earth rotations, from which zoned wall time is to be delivered). Great. That's exactly what https://xkcd.com/927/ is about.
I don't know if that's true. We already have a name for the time that is tied to Earth rotation, "UT1". Users who care about positions of astronomical objects can refer to that. UTC was always a bit of a compromise, which partly tracked UT1, but only up to the nearest second.
On the other hand, I think the idea of abolishing leap seconds is to no longer derive wall time from earth rotation, and to derive it from atomic time instead. So "UTC" would still refer to "the source of wall time". As I understand it, this is the point of changing the derivation of UTC: there are lots of laws in different countries that say that wall clock time is tied to UTC, so if we want to switch from earth rotation to atomic time, it is easier to change the definition of UTC than to change the laws in all those countries.
I see that the solution the author proposes on the "right+gps" subpage is basically the same as what Dan Bernstein has proposed for many years now, except using GPS time rather than TAI (which only differs in the seconds offset). Sounds like the sanest approach to me.
Dan Bernstein proposed it and Bradley White wrote the code in the time zone package used by Unix-like systems. The point of "right+gps" is that all of the pieces needed to implement such a scheme are already deployed and tested and with appropriate caveats anyone can decide to use that to avoid leap second issues.
I guess people don't like mid-summer January because it makes historical comparisons more difficult. Not impossible, just awkward.
It's a horrible messy dilemma. Leap seconds don't seem as dramatic as the Roman intercalary days, but there's the same battle between synchronicity and simplicity (where simplicity = something short of insane perturbations).
Maybe none of this matters for radio broadcast time signals. Maybe you just want a fixed counter everyone's synchronizing to, and people should convert to the Earth's messy system at the last possible moment.
Fun story about the awesome 13 month Calendar that we could have adopted (except for Hitler): http://99percentinvisible.org/episode/the-calendar/
http://www.livescience.com/38083-earth-core-day-length-patte...
http://www.scientificamerican.com/article/earth-rotation-sum...
http://news.nationalgeographic.com/news/2011/03/110316-japan...
https://www.uwgb.edu/dutchs/PLATETEC/RotationQk2004.HTM
https://en.wikipedia.org/wiki/Julian_calendar#Motivation