Thread by @Foone

Stupid retrocomputing fun fact:
For the next three weeks, if you boot up Windows 95, the clock will be wrong. It'll be off by an hour.

(in the US, at least)
This is because it uses the 1995 rules for daylight savings time, and they were changed in 2005. So it changes the clock automatically on the first sunday in april, instead of the second sunday in march.

When this change was made, Windows 95 was already End-of-Life, so it never got an update.
Windows 98 did, however, because it was still supported until 2006.

I was thinking about this because there are again calls to drop daylight savings all together (or rather go on "permanent daylight savings" which is a distinction without a difference)
and it makes me think of systems like this.

if we finally do that, the world will get simpler. no changing clocks... except for all the "smart" ones that do it automatically, for us.

including the ones that now do it automatically and wrongly, because they're using the wrong dates, and can't be updated.

so imagine in 2020 we go permanent daylight savings time. no clock changes! great!

2nd sunday in march comes along, and half our clocks automatically go back anyway, and we have to fix them.

I'm not saying this is a valid reason to not go on permanent daylight savings time, but it's an amusing reality of the world we'll live in for a while after dropping daylight savings time.

BTW, at least three states (florida, california, nevada) are moving towards going on permanent daylight savings time.

Arizona & Hawaii haven't used daylight savings time for decades now.

and you don't want to know what Indiana did.
really, you don't.
in any case, since 2006 they're all using daylight savings time, though they're split across two time zones: central and eastern.

and while Hawaii has never really observed daylight savings, it did temporarily during WW2, as the whole country was on year-round daylight savings time.

BTW, the first time the US tried daylight saving time was in 1918, during WW1.
It ran until the end of the war, when congress abolished it because it was unpopular.
The president (Woodrow Wilson) vetoed the repeal, and congress OVERRODE THE VETO.

for those not familiar with the way vetos work in US politics, the bills congress passes just need a simple majority, but the president can veto them.
congress can then override the veto, but they need 2/3rds to vote to override.

this is not common! it's very rare for that much of congress to agree on anything.
so the WW1 daylight savings time must have been VERY unpopular.

BTW, during WW1 we went on the sort of standard partial daylight savings time where it only happens during summer, whereas in WW2 it was permanent daylight savings.
which arguably isn't daylight savings, it's just moving the whole country one timezone west.

in any case, from 1945 to 1966 it was left up to each state if they wanted to use DST or not and which dates to do it on, so the Uniform Time Act of 1966 tried to make sense of this, sorta.
It just mandated that every state do DST, and on the same dates.

but a few states opted out, like Arizona, and Indiana (which we will not speak of)

The Arizona one is amusing, though, because the terms of the Uniform Time Act say that a state has to opt out COMPLETELY if it opts out.
And Arizona, the state, technically does that, right?

But here's the thing:
On the land that is Arizona, there are sovereign nations that are not run by the US Government, because they're native american nations.
Naturally, they can also decide if they want to do DST or not...

So in Arizona there's the Navajo Nation, covering 27,000 square miles.
It also extends into Utah & New Mexico.
Having those parts of it be on different times would be confusing, right?

So they're all on DST. Even though most of it is in Arizona, they decided to go with the Utah & New Mexico synchronization, and they use DST.

But here's the thing: You see that little hole in the middle of the Navajo Nation?
That's the Hopi reservation! 2,500 square miles, it's a completely separate sovereign nation.

And they don't use DST, because they're entirely in Arizona, which doesn't.

So by starting in the Navajo Nation and travelling west into California, a distance of 300 miles, you can have your clock change four times (if you make the trip in the summer)

1. going from Navajo to Hopi (DST to no DST)
2. Hopi back to Navajo (no DST to DST)
3. Navajo to Arizona (DST to no DST)
4. Arizona to California (Mountain time to Pacific time)

BONUS HARD MODE:
The Hopi Nation is not convex and has exclaves, so by starting at exactly the right point you can add another 4 time changes, bringing it up to 8 time changes in a mere 300 miles.

actually, I'm wrong. Arizona and California are on the same time during summer, so...

https://twitter.com/Sonikku_a2/status/1105202637111062528

That takes away one of our changes but we don't need to go all the way across Arizona now, we can just stop at the edge of the Navajo Nation.
So we're down to like 150 miles, 7 changes. (locations not exact because lazymap)

in any case, before this gets ANY SILLIER, just go watch @tomscott's video about time & timezones and be very glad you don't have to write any of this code.
Time is hard. Timezones are harder. Daylight Savings Time is hardest.

(AND DON'T GET ME STARTED ON LEAP SECONDS)

random fun fact:
when I worked for the US government dealing with old weather records, we had a lot of recordings for what the weather was like on February 29th, 1900.

A perfectly fine day other than it didn't exist.

Because even today 119 years later people still remember leap years as that "every four years thing" but that's not the rule.
It's every 4th year, except every 100th year, except except every 400th year.

so 1900 was NOT a leap year, even though 1896 was, and 1904 was as well.

in 1900 when you're a farmer in the middle of Missouri with no phone, TV, radio, or email, and you see a newspaper once every 3 months, it might take you a while to figure out that it was not, in fact, a leap year.
it might be mid-march or early april before you figure it out.

and a particular type of record that's big in weather recording is the "Record of Climatological Observations Form", also known as COOP data.
It looks like this.
You fill in temperature, rainfall, some other minor info, and mail it in, every month.

so more than a few COOP stations had mailed in their february forms before realizing WHOOPS it's not actually March 16th or whatever, it's March 17th because February 29th didn't exist, as it's not a leap year.

So our database was full of entries which happily told us what the weather was on a day that didn't exist.
BTW, databases & date parsers of all kinds really don't like being asked to parse the date 1900-02-29. It's clearly invalid. So handling these was FUN

the best part is that there was no simple algorithmic fix: you could just assume February 29th was really March 1st, because sometimes it wasn't.
Sometimes they'd be looking at their February form before mailing and go "shit, I forgot to put in February 29th! Lemme fudge it"

And sometimes there'd be a gap mid-march where they resychronized. Sometimes there wouldn't be, which is a scary thought. It means they recorded 366 days of weather in a year 365 days long.

In any case, how we dealt with it depended on what we were doing.
In the long run for most datasets it was "yank that whole month's form (along with march) until someone with a PHD can look at it"

but for some calulations we just pretended not to see February 29th, 1900.
Things like average annual rainfall and low/high temperatures.

disclaimer: I haven't worked for the government for almost a decade now and this is definitely not related to climate change.
we didn't miscalculate global temperatures because one day got screwed up 119 years ago

BTW, fun fact about COOP observations: They started back in the 1800s and they're still going today.
They're now done by something called WeatherCoder3 instead of mailing in paper records.
It's a website where you just type it in instead of putting it on paper.

We were rolling this out back in 2011 and getting lots of pushback because not all the stations had internet access... or computers.

Because of the "no internet access" thing there was an option to print out the forms from an offline version of the site and mail them in.
Some stations are still doing that.

And some stations are still doing the even older paper records, because they were quietly grandfathered in, because it turns out getting people to buy a computer in 2011 is hard.

so basically those stations are in a sort of situation where they're gonna be allowed to do paper for as long as the observer is alive.

and then someday they'll die, and either the station will close or someone else will take up the reins, hopefully someone younger who has a computer.

BTW, the fact that you could print them out and mail them in caused a problem.
cause some stations didn't really get the memo that it was a website that sent us the data directly.

So they'd fill out the website, then print out the website and mail it in

so we had two copies! a digital one and a paper one.
Well, clearly the digital one is better and easier to use, right? we can just toss the paper one?

yeah, about that...

it turned out at least one observer had filled out the form on the website, day by day, then at the end of the month they clicked submit, print, then MADE CORRECTIONS ON THE PAPER COPY AND MAILED THAT IN

so the digital records are, in fact, wrong. the paper copy is the authoritative one.

the final stupid digital-vs-paper wrinkle is the issue of continuity and consistency.
Originally the plan was that we'd have the old website that gave you paper records, and if they had stopped doing paper and moved onto weathercoder3 digital records, it'd point you to that site

this was clearly way better, right? for those records you'd get more data, more accurate data (mostly), and in a nice clean web form instead of an ugly scanned piece of paper that got mailed halfway across the country, twice.

yeah, no.
See, here's the problem: One big use for weather records of this kind is in lawsuits and criminal cases.
There's all sorts of times when you need to show what the weather was like on a certain day, or how it was over a year.

This is handy for things like if you crash into someone and the other guy says "oh I couldn't see you, the sun was in my eyes" but you know it was a cloudy overcast day.
You have your lawyer get a certified copy of the local weather for that day from NCDC, and BAM CASE CLOSED!

or maybe you're filing for some kind of insurance money because your farm had a really bad year, because it didn't rain at all.
but the insurance company says they think it rained constantly, they want to have certified records.

and it's apparently very important when showing evidence to a judge that you have consistency.
like "here's march, here's april, here's may, see how they all show at least 12 inches of rain a month?"

not "here's march and april and ON A COMPLETELY DIFFERENT SITE, may"
the judge is going to ask "why is this one different" and the opposing lawyer is going to raise the question of if you're submitting invalid evidence picked from a different site.

SO, the long and the short of it is: we need the "paper" records, even in the all digital era.
to the end user, it was never paper to begin with. it was PDFs, PDFs of scanned paper, sure, but no paper. We just need something that looks like paper, right?

ENTER THE GENERATED PAPER PDF!
Because this station (Belleview 4 SE, Missouri) had fully moved over to WeatherCoder3 on this date (October 2016), there's no paper form. It's just XML stuffed in a database.
But we can make it look like how a paper form would look!

So now if you take this (and some real scans versions) to a judge, you can show that they're all similar and they're all cerrtified and this definitely tells you that in Belleview 4 SE, Missouri, on October 12th, 2016, it was between 57 and 82 degrees.

and it looks just like (or at least, close enough to) the real paper ones they were sending in a back in 2008. And you can get your money from the insurance and save the farm.

You really don't want to know how long I worked on this project and all the stupid little complications that went into it and how much of your taxpayer money went into not making lawyers skittish about a PDF changing into a HTML page

I don't think about my time working for the government much...
just every morning.
as I'm putting in MY GLASS EYE.

I'm going to stop now because I could talk about the government job all day and timezones and weather and I have other stuff I wanted to do today and also this thread is long enough.

But just remember: Your windows 95 clock is wrong, so don't depend on any for anything this week

Don't be tempted to manually fix it, either.
It'll then get set forward AGAIN in three weeks when win95 thinks DST starts.

Fix your clock, THEN turn off automatic adjustment.
Then you'll be fine.

well, at least until November 3rd.

oh and I promise to shut up after this:
I know it's "Daylight Saving Time". There's no S.
I refuse to call it by the right name, that'd be too much respect for such a terrible idea.