1. 38
bipm.org

Introducing ronna- (10^27), ronto- (10^-27), quetta (10^30), and quecto (10^-30).

1.

2. 15

What does it say about me that I immediately dropped everything else I was doing and read the announcement.

Some fun stuff in here:

• The motivation for new prefixes was “the needs of data science in the near future to express quantities of digital information using orders of magnitude in excess of 10²⁴”. Right now there’s an estimated 4e22 bytes bytes of data in the world, and we might hit the yottabyte era in the next 50 years.
• Everybody’s sick of leap seconds. They’re accepting proposals to “ensure the continuity of UTC for at least a century” that would decouple UTC from UT1.
• They’re also accepting proposals to redefine the second! Right now it’s based on the frequency of a caesium atom, but nowadays we can get much more precise measurements with other atoms, like ytterbium. The eventual redefinition will probably try to find a balance between precision, robustness, and simplicity.
1. 8

Everybody’s sick of leap seconds. They’re accepting proposals to “ensure the continuity of UTC for at least a century” that would decouple UTC from UT1.

Kinda disappointing that we’re preparing to change (well, lie about) reality in order to make programmers’ oversimplifications correct.

I’m not just trying to be quippy here. Unix time originated in the 60s and we’re still trying to make it work by any means other than fixing it. How much longer has it got? In another 60 years when we all live on Mars will we still be assuming there are exactly 86400 seconds in a day?

Edit: Reading that back, I realise that “we’re still trying to make it work by any means other than fixing it” is a pretty good summary of my feelings about computing in general.

1. 5

It’s not really about programmers. UTC just formalizes “1 day = 24 hours of 60 minutes of 60 seconds”, and was first established in 1963. The leap seconds are to keep it aligned with UT1 (1 day = 1 complete earth rotation), which is used almost-exclusively by astronomers. So for the past 60 years all of society was bending over backwards to make the astronomers happy.

(I think UT1 is also used for navigation? Probably not anymore because everybody uses GPS now, and GPS uses atomic clocks, so it’s better off without leap seconds too)

1. 2

I don’t think it’s only about astronomy though. If you drop leap seconds entirely, you’ll let clocks and observed time (ie the day/night cycle) slowly drift apart. This is definitely important for everyone. Obviously it would take a very long time for this to become a problem, but that’s not really an excuse.

1. 10

Eh, solar noon is already a bit before 11:30 am my time, and I’m not even at the easternmost end of my time zone. Leap seconds are not the low hanging fruit here.

1. 4

Sounds like a problem you should fix by adding leap seconds to the timezones, not to UTC itself.

2. 2

It’s not really about programmers. The only people who would be affected for around the next thousand years[1] if we stopped doing leap seconds would be astronomers, the people who are affected with leap systems are any people who depend on systems that have to keep accurate time. We are easing complexity of safety-critical systems to avoid the need for astronomers to have their own time units. That does not seem like the right trade.

The day is calibrated so that clock noon (12:00) and high noon will be at about the same time and that is the only calibration that anyone who is not an astronomer cares about. Within a time zone, there is at least half an hour’s error in this. It’s impossible to predict leap seconds (this is why they exist and why they’re so annoying), but rough estimates suggest that it will take about a thousand years for this to be the largest source of error in that calibration. At that point, if the majority of humans both have a functional civilisation and are still all living on this planet then either having a leap hour or just moving everyone’s time zone (we already have infrastructure in time zone handling for this) would be vastly simpler.

1. 2

Within a time zone, there is at least half an hour’s error in this.

Even more, depending on the time of the year. Right now in Chicago, solar noon is 11:35 AM. During the summer Detroit’s solar noon can be as late as 1:40 PM.

2. 1

In another 60 years when we all live on Mars will we still be assuming there are exactly 86400 seconds in a day?

I propose Martian days, based on Julian days, which are already supported in SQLite’s `julianday` function.

1. 0

I agree, I don’t think changing UTC is the solution here.

We do need to stop touching the system clock, which should be a strictly monotonic counter of seconds (or more granular if necessary) since epoch (basically TAI). That’s the way computers understand time, and I don’t think there’s anything really wrong with it.

Instead, ship a table of leap seconds with the tzdata and handle them when converting the time for display (or converting human input to computer time). We already have to do a bunch of math for that anyway, so what’s a simple extra LUT fetch?

This should avoid synchronization issues related to magically changing the time while keeping the leap second’s meaning, which is to keep UTC and UT1 not in perfect sync, but at least close enough.

1. 2

I may be reading the resolution wrong but I don’t think anyone’s thinking about changing UTC. The proposal is to accept a wider magnitude for the UT1-UTC difference. That would basically just amount to making the LUT smaller and, notably, ensuring that we can continue to have only positive entries in it (i.e. that we only have to insert leap seconds, rather than add them).

FWIW, the problem of universal time references is pretty old and definitely predates programmers making wrong assumptions about UTC and leap seconds. Adding (or removing, assuming we can do that without blowing up ships and whatnot) leap seconds poses real traceability problems, and traceability is one of the cornerstones of the modern metrological system.

2. 1

They’re also accepting proposals to redefine the second! Right now it’s based on the frequency of a caesium atom, but nowadays we can get much more precise measurements with other atoms, like ytterbium. The eventual redefinition will probably try to find a balance between precision, robustness, and simplicity.

The resolution is pretty understated, and the primary motivation is definitely technical, but I have a feeling this is going to cause some interesting ripples. (Okay, well, I’m using “interesting” loosely here). The fact that the definition of both meter and second were largely uncontroversial made it a lot easier to reach a consensus regarding the definition of the kilogram, not that it didn’t take, like, forever to finally reach it back in 2019. Even that consensus was somewhat uneasy for the metrological community, which despite its occasional tug-of-wars is remarkably consensus-driven. If the future redefinition will simply involve another atom species, that’s probably going to be fine, but a significantly different realization, involving a different mechanism and, thus, a different enough definition, could well end up muddying some waters.

I am going to be so fun at parties from 2027 onwards!

3. 20

I guess that means hella- didn’t become official.

1. 1

Pour one out :(

2. 5

It took me a bit to notice that the “ronna-” and “quetta” prefixes are for extremely large values, while the “ronto-” and “quecto-” prefixes are for extremely small values.

1. 9

It’s like that for everything larger/smaller than exa/atto; we already have zetta and zepto for 10^21 and 10^-12, and yotta and yocto for 10^24 and 10^-24.

Would’ve been kinda cool if it applied to everything maybe? kila/kilo, mega/mecto, giga/gecto, tera/tetro, peta/pento, exa/exo, zetta/zepto, yotta/yocto, ronna/ronto, quetta/quecto.

I noticed that “tetro”, the 4th power of 1/1000, looks suspiciously like a word based on the “tetra-” prefix meaning four, so I looked at the other prefixes. We have penta- meaning five, which looks a lot like “peta”, hexa- meaning six, which looks a lot like “exa”, “octa-” which looks like “yocto” if you squint, and “nona-” which looks a little like “ronna”. I wonder how much of this is coincidence and how much is on purpose.

1. 1

gecto- would also need to drop hecto-, though — although of course it would be dropped just because it ends in -o-.

Femto- already sounds not so far from pento-, and the comments cited seem to confirm it is intentional: https://phys.org/news/2022-11-earth-ronnagrams-metric-prefixes-voted.html

1. 1

I wonder how much of this is coincidence and how much is on purpose.

It’s intentional. From https://phys.org/news/2022-11-earth-ronnagrams-metric-prefixes-voted.html:

“The only letters that were not used for other units or other symbols were R and Q,” Brown said.

Convention dictates that the larger prefixes end in an A, and the smaller ones in an O.

And “the middle of the words are very, very loosely based on the Greek and Latin for nine and 10,” Brown said.

2. 4

-chico -harpo -groucho -gummo -zeppo

1. 2

So does this mean we now have robbibytes and quebbibytes?

1. 3

I am not completely sure, but to my understanding the IEC would first need to put out a new version of IEC 60027. Not completely sure though and I don’t think one can legally access that document without paying.

2. 1

Coincidentally, those are the names of my daughters.

1. 1

Setting up a very nerdy yo momma joke.