Comment by avidiax
8 hours ago
An interesting project, that buries the lede to another interesting thing: abolishment of the leap second.
> In November 2022, the General Conference on Weights and Measures decided to abolish the leap second by 2035
I think the story goes that when the leap second was standardized in 1987, the wold was much less connected, there were fewer systems, and hence it was practical to keep astronomical time and civil time coordinated.
But this causes so many bugs and difficulties in practice that companies like Google don't perform leap seconds, they do leap second "smearing". But even smearing has issues, since there are multiple algorithms.
Even worse, the Earth sometimes accelerates, and there was a prospect that we might need a negative leap second, which has never been put into practice.
And as daylight savings makes everyone painfully aware, the public can tolerate even hours of misalignment between civil time and astronomical time, and even an hour adjustment.
So if we decide that we can tolerate a whole minute of skew, we need no leap minute for 50-100 years. And if we could tolerate an hour of skew, we'd not need a leap hour for millennia.
Of course, this is all can-kicking. If a leap second is difficult in practice, a leap hour in UTC (not a flip between a daylight savings and standard time zone) would be even more difficult. I feel the eventual solution would be to very slightly adjust the civil second to be shorter, so that the error begins to reduce.
What deeply confuses me is why we don’t just:
- always use atomic time
- adjust civil timezones as needed to match astronomical time well enough
Somehow, I don’t see most political and contingent timezones lasting long enough to shift by an hour…
That's an interesting idea, but it could be argued that this assumes that all of the world's timezones will agree to observe the same set of leap seconds as they currently do.
Otherwise, it shifts timezones from being 15 minute granularity to 1 second granularity. Not technically impossible, but will definitely make cross-boarder meetings interesting once the discrepancy becomes large enough. I feel the 15 and 30 minute timezones are already bad enough.
Leapseconds don't track solar/astronomical time particularly well-- they track a 'mean time' which you can't even observe (because there are large periodic variations!)-- across the year the span is about 30 minutes.
If you have an application that needs a precise solar or astronomical time, you need to back out the leap seconds and then apply a more accurate correction already.
Because of this, I think it would be completely reasonable to just move timezones by 1 hour once the adjustment would be an improvement in accuracy--- which I seem to recall can be estimated to take about 4500 years. The adjustment is already compatible with existing software, and infrequent planned changes of this sort already happen for political reasons.
Because then one day you wake up early in the morning to find the noonday sun staring at you. It will take a while though. On second [heh] thought, you might not wake up. But some distant relative would.