Hacker News Reader

Comment by SEMW | Hacker News Reader

OK, so what makes you think the next one won't be till 2020? Leap seconds aren't like leap years, they're unpredictable, since the Earth's rotation varies in response to geological events etc. The IERS announces them six months in advance.

View on HN

Admittedly they've slowed down a lot since 2000 (happened at least 7x a decade in 80s and 90s), but I agree, saying 2020 further shows the poster doesn't understand how leap seconds work.

Yikes!! I did a quick look at the periods between 1980 an now and guesstimated that it would be 35 months to the next leap second (2020). As a Ham radio person I get sun spots and solar winds which will make a difference on the rotation.

But as an R programmer I went back and ran the numbers and it's looking like 2020. Happy to see that my Mark 1 Eyeball estimate was close. But I'm happy to see what your numbers are.

With the advent of the Raspberry Pi support for time and the cheap GPS units, I run my time locally. See https://blog.ntpsec.org/ for details.

And yes, I cleared the second change, but I in full disclosure it wasn't until 9 AM before I checked.

Statistically modeling them using the prior dates of leap seconds alone isn't a good method (if that's what you did, not claiming that it is). Knowledge of the past leap seconds doesn't really tell you anything about when future ones will be. A better way would be to model the difference between UTC and UT1 and model the rate of change there. Probably the best would be to model the earth's rotation rate and factor in the likelihood of the events that cause variances. One major source of those events is earthquakes.

Reply on news.ycombinator.com