Hacker News Reader

Comment by nousermane | Hacker News Reader

TAI - atomic clock, ignores earth rotation.

UT1 - based on Earth's rotation only, strictly 86400 seconds per day; length of each second varies; takes a heck of a lot of effort (and time) to measure accurately.

UTC - has same length of a second as TAI, but (for now) tracks UT1 to a precision of +/- 1 second. To achieve that, can have days that are 86399, or 86400, or 86401 seconds long.

None of 3 is planned for scrapping. The only change discussed in TFA is to fix UTC day to 86400 seconds (at cost of letting it drift further away from UT1).

View on HN

Oh right, I confused the terminology, but the point still stands mostly.

It's a minor problem. One minute off in a hundred years is something we can all live with.

Can GPS handle this? Or is it under a different time standard?

The article mentioned GPS uses UT1

GPS has its own time thing, kinda fixed offset from TAI iirc

I'm not saying that we can't, but saying that we can seems quite brave (and by "brave" I mean foolishly reckless) without actually trying it. I, for one, am not so sure. There is a lot of stupid stuff about calendars and clocks (of which DST is worst by far), but at least nothing is really drifting.

Is it important that midnight is actually midnight? I don't know. I would prefer it to be, but perhaps I can just update my knowledge on when the midnight actually is once or twice a year. Perhaps it isn't a big deal at all. But I wouldn't dare to claim that "surely there can be no problems with it".

And, yes, that kinda means I don't support the leap second removal. It wasn't great, but eventually fixing the software written by incompetent people sounds more like a solution, than breaking the existing solution and making both competent and incompetent people alike eventually come up with a new one (to abolish it as well after 100 more years).

> Is it important that midnight is actually midnight?

Midnight is almost never at midnight.

Let's talk noon, specifically solar noon, as it is easier to visualize. Solar noon varies throughout the year, and based on the position in the timezone.

Solar noon for two cities in Eastern Time Zone: 11:29:16 - Boston 12:22:36 - Atlanta

Tomorrow, solar noon in Atlanta will be 12:22:50, 14 seconds later than today. No one notices a 14 second shift from one day to the next. As you can see, solar noon is almost never at clock noon, yet no one complains about that. Who would notice a 1 minute shift over 100 years?

The programming problem is made vastly more difficult because the leap seconds are not known in advance. When a new leap second is announced, every dependent system has to be updated.

Reply on news.ycombinator.com