Why is a leap smear necessary for skipping forwards? Unlike a typical leap second, which goes backwards, a forward jump doesn't repeat timestamps. The clock could literally just jump forwards a second and it would be fine.
I don't really want to get into what code should or shouldn't have been written.
When you are designing infrastructure to support millions of lines of existing code, you figure out what will break the least code, and you do it.
You know what absolutey is inherent to imekeeping?
* Drift between terresterial and sidereal time,
* Drift between terresterial ground level (1G gravity mean 'surface') and orbiting satellite (eg: GPS time)
* Differing frames of time reference best suited for different tasks.
Common computer time server synced time is based on the nominal notion of "mean noontime" UTC .. or the rotation time of the earth to move from Greenwich "pointing at the sun" to again "point at the sun" (perhaps surprising to some NOT a full "360 degree" rotation as the earth has moved forward on its orbit).
Myths about time that programmers should know is a good read .. eventually drift between systems must be addressed ( GPS sats use "epoch time" (local atomic clock ticks) lapsed "per week" and GPS recievers reconcile relativistic drift using broadcast corrections ).
Any significant scientific data collection program (geophysics, astronomy) required the use of some form of "epoch time" with accurate local clocks not synced to UT | GMT | Zulu time servers.
Your apparent desire to not deal with adjustments on the order of a second during your lifetime is simply pushing the problem down the road until it's a case of reconciling hours .. or even 11 days (as has already happened in history).
All those things are definitely important, and leap seconds don't help with any of them.
> Your apparent desire to not deal with adjustments on the order of a second during your lifetime is simply pushing the problem down the road until it's a case of reconciling hours .. or even 11 days (as has already happened in history).
An hour, or fraction of an hour, is significantly easier to reconcile. Even assuming we stick with the same system for that very large number of years.
> to again "point at the sun" (perhaps surprising to some NOT a full "360 degree" rotation
It's actually more than a 360 degree rotation because the earth rotates in the same sense that it revolves. So the sidereal day (1 earth rotation in physics sense) is ~ 23h 56m, and the 24 hour period is over-rotation to bring the sun back into same position, and that's more than 360 degrees of rotation.
Civil time is based on time zones, and it will take thousands of years to have a meaningful drift in comparison to that. If we avoid the hubris of thinking that we know how clocks will be set hundreds of years in the future, then the error from lacking leap seconds is negligible.
And it's not like allowing .9 seconds of drift is objectively correct either.
The boundaries of time zones are defined with reference to a fixed line - the prime meridian. On the prime meridian, "midnight" is an important concept, marking the change of date.
A second of drift is negligible, but a minute of drift pushes the boundaries of time zones sideways by 6086 feet at the equator.
> On the prime meridian, "midnight" is an important concept, marking the change of date.
But having the average solar midnight match that very specific spot is not important.
> A second of drift is negligible, but a minute of drift pushes the boundaries of time zones sideways by 6086 feet at the equator.
Which doesn't matter when they're 60x that size wide. Especially when countries will shift an entire time zone for fun.
A minute of drift is also negligible for civil time, as long as everyone is in sync, and we can't really predict the timekeeping future beyond that first minute of drift.
Civil timekeeping accepts error vs solar time of up to half an hour as table stakes, and routinely raises that by an hour or more. Leap seconds account for half a minute over half a century.
You can get a legislature to change your time zone faster than the problem leap seconds purport to solve can become relevant.
Lack of understanding of the relation between 24 hours and rotational period of Earth.
Earth does not take anywhere close to 24 hours to complete a rotation, regardless of how you measure it.
Earth rotates in 23h 56m 4s with relation to distant stars.
Earth rotates in 24h +- 30s with relation to Sun. The length of day depends on how far Earth is from the Sun, the speed at which Earth rotates the Sun changes throughout the orbit. And so every day the Sun makes a different distance around Earth (as seen from Earth), making the day a little bit longer or shorter.
Imagine the complexities of systems dealing with both perspectives of time, and the drift that can occur over… er… time. Particularly imagine the hypothetical complexities of a system where there’s any N > 1 civil time system. Space colonization is probably what springs to mind, but it wasn’t long ago (this past week if I’m not mistaken) that a civil time system without leap seconds was proposed for Earth and made it to the HN front page.
It’s easy for things to go wrong when basic facts about time don’t line up. Not every set of things and wrong-going will rise to the level of what people imagined about Y2K, but I can personally attest to months of hell untangling and correcting records related to minute differences (sorry, accidental pun but I’m leaving it in) in how two very widely used RDBMSes keep time.
Surprised by this error from time and date .com the earth does not rotate once every 24H of course, that's the time it takes for the sun to be back in the same position (mean solar day) the earth rotates once every sidereal day 23h 56m 4s (minus the 1.55ns)
I picked up on that too. I generally expect that type of simplification in non-technical articles geared toward the general public. But this one was specific enough that I think it deserved a mention.
This is an interesting question. The emission of radiation is proportional to the absolute temperature to the fourth power. So, the hotter something gets the faster it radiates energy out. Its nonlinear so theres definitely some winning strategy.
Energy in is constant, but energy out could vary. I think the way to maximize the energy leaving the planet is to not rotate at all. The energy input would be concentrated and you would have one side very hot — and the hot side would dump the most energy back into space.
The short-term length-of-day chart is very noisy, and has some seasonal components as well. The IERS Bulletin A has a table of DUT1 values for the next year, plus a formula for longer-term projections. (LoD is the rate of change of DUT1.) If you ignore the seasonal components, it is a simple linear formula, multiplying the Julian day number by a LoD factor. Last time I looked this longer-term LoD factor was 320μs short of 24 hours, which suggests the next leap second will be in about 5 years time (2027) and will be a negative leap second.
There was a short period towards the start of this year when it looked like the LoD was increasing, but that was only temporary. The decrease in LoD has been fairly consistent (apart from the occasional pause like earlier this year) for a few years, starting when the LoD was more like 1ms slower than 24h, so it was a while before the wider time-keeping community realised something weird is happening.
From the added momentum from various historical figures whirling in their graves.
See now if we'd paid a bit more attention to orienting them properly (not too late now!) we could use that angular momentum to straighten the earth's axis and thereby eliminate the scourge of winter! Or at least banish it to the benighted polar lands where it belongs.
At least biggest cloud providers have already envisioned this [1].
> A negative leap second, if one were ever to occur, would be smeared by speeding up clocks over the 86,399 SI seconds from noon to noon.
[1] https://developers.google.com/time/smear?hl=en