Earth Rotation Records Spur October Vote to Avert Negative Leap Second
5 points by fanf
5 points by fanf
This article is a mess. Every time it tries to get technical, it makes odd & confusing mistakes. Maybe it's a bad copy of someone else's real article, or maybe it's just slop.
Either way, don't waste your time.
A leap hour 50 years from now? I don’t think that’s correct.
Also:
Transaction ordering in distributed databases depends on timestamps; skipping a second can make a later transaction appear earlier than an earlier one. No production system has ever encountered this scenario
Literally every distributed system has dealt with mismatched clocks.
I came here to say this, non-monotone clocks are nothing new. The one thing that folks should be worried about is that recently in-DC clock skew has gotten very low, and there may be some modern distributed systems that don't do cross-DC work that assume monotone system time. My guess is that there will be real bugs. But it probably won't be that bad.
Clearly wildly wrong. Even if you were getting a leap second every single month, it would take 300 years to get to an hour. There’s no point even thinking about such a time scale (let alone several thousand years) on such a topic.
The article is somewhat littered with bizarre errors. Another of them has a negative leap second subtract two seconds for some reason:
this means the clock would jump from 23:59:57 directly to 00:00:00, with 23:59:58 and 23:59:59 simply not existing.
Smearing a leap hour sounds like something that would take a long time. For countries that observe daylight savings, it would surely make more sense to skip changing the clocks back or forth, dependent on the direction of the correction. Given there would be decades (or centuries?) to work that out, it seems like that should be possible?
Leap hours don’t have to be implemented as part of UTC: instead the difference between UT1 and UTC could be allowed to grow without bound, and (as you imply) any country that decides that it doesn’t like where noon has drifted to can change its time zone. The big advantage is that this doesn’t require any global coordination.
The weird thing is that DST is likely to be abolished soon: there were serious moves in that direction on both sides of the Atlantic in about 2018-2019 which subsequently stalled, but maybe abolition in British Columbia will get it moving again. One consequence of abolishing DST is that time zones will almost certainly ossify, so it won’t be as easy to use them to adjust the clocks as we are accustomed to.
And a leap hour will not be needed for several centuries. The drift is so gradual that people are likely to think it isn’t worth the effort to move noon from 13:00 (normal) to 12:00 (pedantic).
<hollow laugh>can change its time zone. The big advantage is that this doesn’t require any global coordination
This is a fairly good summary of the situation, but I think it overstates how much this vote is spurred by the risk of a negative leap second.
The process to abolish leap seconds has been going on for over ten years (at least this round of the effort; there was an earlier attempt around 25 years ago that failed). It involves two major treaty organizations, the CGPM, which governs the BIPM that is responsible for the SI and implements UTC; and the WRC / ITU-R, which defined UTC. And there are scientific coordination bodies that must be consulted, primarily the IERS and IAU. So there’s a lot of international politics and bureaucracy, and it has taken several rounds of the respective organizations’ triennial / quadrennial treaty update conferences.
It managed to get on the agenda after the CGPM approved the new SI in 2018, so a big chunk of work left its agenda making space for updating UTC. There’s a timeline and more context in a blog post I wrote in 2022 https://dotat.at/@/2022-12-04-leap-seconds.html
Concurrently with this is the weirdness of the earth spinning faster. There’s some similarity with the previous big gap between leap seconds, 1998-2006, but this time the earth was starting from a slightly faster speed and managed to drop below 24h per day. A few months ago I drew a chart of the length of day and UT1-UTC over the last 55 years https://mendeddrum.org/@fanf/116229600425864518
There’s some extra weirdness that the earth’s nutation changed behaviour abruptly in 2020, losing an annual periodic component. Dunno how significant it is (the parameters are measured in microarcseconds which is frankly tiny) https://mendeddrum.org/@fanf/116234523185471864
What I didn’t notice when drawing those charts, and what the article doesn’t mention, is that the earth is showing signs of slowing down again: the length of day is now about 200us longer than 24h whereas for the last few years it has averaged around 100us less than 24h. The IERS weekly Bulletin A, which promulgates measurements of earth rotation with predictions for the next year, shows UT1-UTC trending down again https://mendeddrum.org/@fanf/116913457313127408
Even so, as the article says, the leap second abolitionists are using this unusually fast rotation to try to persuade the Russians to accept a faster timeline, so that ideally there will never be another leap second either positive or negative. (Even with the 2035 date proposed 4 years ago, there’s a fair chance we might not need one!)
The discussion of leap hours in the article is weird. The proposal 4 years ago was that the permitted divergence between UT1 and UTC should be increased from 0.9s to something large enough that there would be no need for any leaping before 2100 or within 100 years or something like that. Based on recent decades in which there were a few years between leap seconds, that plan suggested a leap minute, not a leap hour. A leap hour will not be needed until nearly the year 3000 by which time (if they were not being abolished) we would need several leap seconds each year. Steve Allen has lots of charts and tables that he prepared during the previous failed attempt to abolish leap seconds https://www.ucolick.org/~sla/leapsecs/dutc.html