Leapseconds are inserted to keep UTC within ±0.9s of UT1, a measure of earth's rotation.
Due to historical reasons the length of a second was defined somewhat shorter than 1/86400 of a mean solar day at the time our current definition was established. Because of this the trend between UT1 and UTC is largely in one direction and all leap seconds, so far, have been positive. But leap seconds can be negative if earth's rotation speeds up enough.
As things are leapseconds are a barely mitigated disaster: They, like their distant cousin daylight savings time, cause widespread disruption in time sensitive electronic systems for very little benefit: Without leapseconds it would take thousands of years for UT1 and UTC to diverge by an hour-- and at that point, if we really cared, we could update timezones to move noon closer to astronomical noon.
Leapseconds even cause outages and system failures when they're not even happening due to false leapsecond events and the difficulty of validating leap-seconds (https://news.ycombinator.com/item?id=28046997). Though the offsets caused by leapseconds are small they're worse than DST because DST is a timezone effect while leapseconds make the UTC timebase our systems count on unreliable.
Recently earth's rotation has been speeding up, causing the UT1-UTC offset tend to stall and begin moving in the opposite direction. At some point this may result in the issuance of a negative leapsecond. Since ordinary positive leapseconds cause widespread issues and there has never been a negative leapsecond it's reasonable to assume a negative leapsecond would be much more disruptive.
There have been formal proposals to abolish the leapsecond in the past. It's due for discussion at ITU-R World Radio Conference in 2023 after being postponed back in 2015. There is more background on the regulatory process here: https://www.ucolick.org/~sla/leapsecs/onlinebib.html
An alternative to eliminating UTC would be to move systems over to TAI (or, TAI plus the current 37 second offset) but this change would be costly and disruptive while the same effect for most systems could be achieved by simply not issuing any more leapseconds.
and here is the above with the sloped removed, e.g. if the second had been defined to match the mean solar day in those 22 years: https://nt4tn.net/random/UT1UTC2000secondfixed.png -- this cheating a bit since obviously using the slope generated from the same data is going to fit pretty well.
Due to historical reasons the length of a second was defined somewhat shorter than 1/86400 of a mean solar day at the time our current definition was established. Because of this the trend between UT1 and UTC is largely in one direction and all leap seconds, so far, have been positive. But leap seconds can be negative if earth's rotation speeds up enough.
As things are leapseconds are a barely mitigated disaster: They, like their distant cousin daylight savings time, cause widespread disruption in time sensitive electronic systems for very little benefit: Without leapseconds it would take thousands of years for UT1 and UTC to diverge by an hour-- and at that point, if we really cared, we could update timezones to move noon closer to astronomical noon.
Leapseconds even cause outages and system failures when they're not even happening due to false leapsecond events and the difficulty of validating leap-seconds (https://news.ycombinator.com/item?id=28046997). Though the offsets caused by leapseconds are small they're worse than DST because DST is a timezone effect while leapseconds make the UTC timebase our systems count on unreliable.
Recently earth's rotation has been speeding up, causing the UT1-UTC offset tend to stall and begin moving in the opposite direction. At some point this may result in the issuance of a negative leapsecond. Since ordinary positive leapseconds cause widespread issues and there has never been a negative leapsecond it's reasonable to assume a negative leapsecond would be much more disruptive.
There have been formal proposals to abolish the leapsecond in the past. It's due for discussion at ITU-R World Radio Conference in 2023 after being postponed back in 2015. There is more background on the regulatory process here: https://www.ucolick.org/~sla/leapsecs/onlinebib.html
An alternative to eliminating UTC would be to move systems over to TAI (or, TAI plus the current 37 second offset) but this change would be costly and disruptive while the same effect for most systems could be achieved by simply not issuing any more leapseconds.
Past HN discussions: https://news.ycombinator.com/item?id=29924020 https://news.ycombinator.com/item?id=25658733