The one second war: Finding a lasting solution to the leap seconds problem | Hacker News Reader

The one second war: Finding a lasting solution to the leap seconds problem

cacm.acm.org

81 pointswgrover27 comments

Maybe the most awesome part of being a human being today is recycling solutions from the past to the solutions of today.

Once upon a time, written language was represented in computers by arrays of characters. Manipulating non-english user content in this model was big pain. Today, effective software dealing with written language uses UTF-8 Characters and the Grapheme abstraction. The problem was solved by separating the concepts of Bytes, Characters and Graphemes, which were previously conflated.

Clearly, it's ineffective to use the same abstraction for "elapsed time" and "time of day" due to the variable length of the day. Representing both concepts with one value leads to the problems described in this article. The concepts should be separated and a new abstraction created: Earth Position (EP).

Statements about the time of day, day of the week or lunar month of the year are really statements about the Earth's position relative to some other entity: the sun in most cases, but sometimes the moon in the case of lunar dates.

Given that "Unix time" is widely understood as "the number of seconds elapsed since Jan 1, 1970", it would be convenient to let it only represent Elapsed Time and define Earth Positional values in terms of Elapsed Time.

The leap second concept should only affect applications interested in Earth Position queries (time of day, etc) and not have any representation at the Elapsed Time level.

You mean TAI vs UTC?

TAI: http://en.wikipedia.org/wiki/International_Atomic_Time

UTC: http://en.wikipedia.org/wiki/Coordinated_Universal_Time

"As of 1 January 2011, 00:00:00 UTC, TAI was exactly 34 seconds ahead of UTC"

Unfortunately, TAI (Atomic Time based on radioactive decay) may not be as consistent as we originally thought it was according to recent observations...

http://news.discovery.com/space/is-the-sun-emitting-a-myster...

No. The link you gives mentions that the decay rates of radioactive particles are changing, this saying the half life is changing. Atomic clocks are based on the frequency of light emitted from an electronic transition. Ie, an electron makes a transition from one state to another and emits a photon, then we compare the frequency of that photon to another photon which has a nearby frequency, we measure the difference to like 8 digits, then use the frequency of the light produced from the transition as the inverrse unit of time.

One is based on nuclear decay, the other an induced electronic transition and not affected by unknown variations in the half life.

More like TAI vs UT1, if I read jtwb correctly.

Within 20 or 30 years, we may need Luna Position and Mars Position queries to cater for many users on the network. Of course, we'd also need to know the distances between them, Earth-Mars Distance, Earth-Luna Distance, and Luna-Mars Distance queries, for calculation adjustments.

You really nail it. I just want to add that all of this applies to DST too, although it’s a little different because it’s artificial and well-known.

Speaking as a mad scientist, it's clear to me that the best solution is just to adjust the Earth's rotation to be exactly 24 hours.

Maybe 24 and a half, so we can all have a bit of a sleep in every morning.

No it should be 24.0159254 old-hours, so that the orbital period is an integral number of earth rotations. Assuming you want to waste your mad-scientist powers on silly bureaucratic nonsense, and not on, say, spinning the earth fast enough to lower surface gravity (leap tall buildings in a single bound to get to Starbucks).

No, speed up the orbital period too so we can have a 360-day year.

Quarterly reports coming sooner? No thanks. Let's slow down the orbital period to 400 days. Nice, 100 day quarters. And, as that change leads to other changes in our daily life, maybe I'll finally eventually use that "grad" button on my calculator.

The article discussed that problem, indirectly. The earth is always slowing down, so even if you fixed it perfectly it would become broken with time.

So we keep on fixing it. Big ol' railguns firing mass out into space. (If we run out, borrow some from the moon.)

If we run out, we have problems much worse than varying day length.

It would be easier to carefully collide asteroids at an oblique angle.

Wouldn't that make a mess though?

Put the impact zone in your favorite geopolitical enemy's backyard. Kills two birds with one, err, stone.

If you live in Britain and you ever hear seven pips on BBC radio instead of six, there's just been a leap second. That's the Greenwich Time Signal computer compensating for the extra second.

Dup: http://news.ycombinator.com/item?id=2418234

Rats, on my first submission no less! I knew this was too interesting to be original. I'll search harder next time. Thanks for catching it.

Don't worry. I missed it the first time and enjoyed it tremendously.

I've always loved the brown M&M anecdote, and consequently it always bothers me when it becomes a cliche sign of a diva in movies. Then again, lots of things bother me in movies (i.e. most things here: http://en.wikipedia.org/wiki/List_of_common_misconceptions).

From my understanding, setting leap seconds 20 years in advance is not a good idea either because large earthquakes (such as the recent one in Japan) can change the Earth's rotation speed enough to require an update.

According to my calculations, the Japan earthquake will not require us to adjust a leap second - at least not within the next ~761 years.

The Japan earthquake shortened the day length by 1.8 microseconds. [1]

1.8 microseconds * 365.25 days * 20 years =~ 0.01 seconds off. Not enough to require an update, even less an urgent update before the 20 years.

1.8 microseconds * 365.25 days * 761 years =~ 0.50 seconds off.

[1] http://www.space.com/11115-japan-earthquake-shortened-earth-...

Ignore it until it's one hour off, then do a "leap hour". We do that twice a year anyway.

PHK's argument is that leap seconds don't come frequently enough to ensure that software is tested in their presence. Replacing them with a 3600-times rarer event would make the problem much, much worse (but on the plus side, it'll be our descendents' problem, not ours ;)

I think he talks about just adjusting timezones to match.

Reply on news.ycombinator.com