Hacker News Reader

Comment by usui | Hacker News Reader

Just wondering, is it possible to have a calendar system that automatically adjusts to the Earth's orbit around the sun dynamically with a precision/accuracy as high as possible? I imagine the calendar would need to note how long that calendar year was, how long a day was, and to what precision it is recorded.

It would be interesting if systems never had to account for a leap day every n years. I suppose 0.003% of a day (the error rate each year) each year isn't a big enough concern, otherwise known as 2.592 seconds.

View on HN

Yes. There are two classes of calendar: observational and arithmetic.

Modern calendars tend to be arithmetic, based on fixed rules that approximate astronomy more or less accurately. The Julian calendar is one of the simplest; the Jewish calendar is one of the most complicated and most accurate.

Observational calendars are more directly based on astronomical observations. The Muslim calendar is one example: in some traditions, the start of Ramadan can’t be predicted exactly because it depends on seeing the new moon. UTC is also an observational calendar, since leap seconds are scheduled by observation not by fixed rules. The French revolutionary calendar was also (in its strict form) observational, keeping its years precisely in sync with the equinoxes.

The Iranian calendar defines its new year based on the exact moment of the northern equinox; if the equinox is before noon (Iran standard time, I believe) then the new year is on that day, while if it's after noon the new year is on the next day. As a bonus, the month lengths are designed to be longer near aphelion, when the corresponding solar months (periods between 30-degree solar longitudes) are longer.

The Chinese calendar is defined (since 1644) based on the days of true new moons and solar terms (corresponding to 15-degree solar longitudes), which also makes it purely astronomically defined.

The various Indian calendars are mostly defined using traditional astronomical texts, but I think there have been occasional attempts to update them to modern astronomical knowledge.

I'm not sure where the rate 0.003% of a day per year comes from in your comment; even the error in the Gregorian calendar is longer than this (if measured based on the northern equinox, about 12 seconds; if measured based on the mean tropical year, about 27 seconds).

I do like the idea of variable length 'true' days as a more exact measure of the world around us but the thought terrifies me as a programmer (or even someone who needs to get practical things done). It's so much easier to reason about time in standard units even if they're wrong. Now if we were to eliminate day light savings, that would be a greatly appreciated change.

There’s precedent: the Babylonian day was 12 hours long; the night likewise. They knew sunrise and sunset so could set a given day’s hour to be 1/12th of the difference (and likewise the night’s).

They clearly knew the extreme cases (the solstices and equinoxes) but I don’t know if they actually adjusted each day; I don’t know how they actually market the hours, much less why they would bother.

Even the “minute” (small part) and “second” (the second division) were measured only due to the development of technology that allowed it.

To clarify, daylight was divided in 12 periods of equal length called “hours”, as well as nighttime. This length depended on the day of the year, when measured with the current definition of hour. 24 hours was always a complete day.

Please no. Trying to sync time to Earth rotation is why we have leap seconds, and they are pretty much universally hated.

(They are hated, because AFAIK they are unpredictable, which also answers your question, assuming you were asking about purely algorithmic/mathematical way without any outside inputs.)

Reply on news.ycombinator.com