Recreating the US/* time zone situation
rachelbythebay.com
rachelbythebay.com
(I should note that in my crowd, references to this movie are always super-negative. "The One" gets damned to hell to eternally fight, but never win his battle.)
P.S. I do distictly remember how a reply on HN to one of her earlier posts on RSS clients mentioned that her laments kinda miguided since her own feed doesn't set one very well-documented and basic cache-controlling HTTP header that most readers actually do respect; but some time later in her later post she described that header as matter-of-fact knowlewdge, no "I've learned about this one recently" remark or anything, and by that time, her RSS feed had started setting it.
While I am personally glad that now there are less infuriatingly stupid network clients around (although, again, those never really amounted to all that much load), and probably adopting a rather caustic attitude at the authors of RSS clients was the only to force them to fix things, but even then: you can do more with a kind word and a gun than with just a gun. Besides, there is some unspoken etiquette between the content servers and content readers; a server that e.g. bans you for exceeding rate-limits when you open the "Full blog archives" page and middle-click at 10 promisingly named links to read them one after another is just rude, personal opinions of the hoster notwithstanding.
P.S. Seriously, max-age=155520? Your server/ISP can't handle serving a ~190 KiB (when gzipped) file even once a day, it has to be almost 2 days? Get it off the public Internet then.
If someone wants to get indignant and broadcast how terrible it is that a project (Debian in this case) is so terrible that they dare have a UX issue that is probably just an oversight, then they're entitled to do that but can also expect pushback where they aren't perfect either.
That's why "US" wins over "America".
This practice is very bug prone, and has lead to high profile failures like goto fail
That particular git repository has history imported from 1996 onward, but Postgres was a very established project by then: https://git.postgresql.org/gitweb/?p=postgresql.git;a=commit...
These days people might not blink an eye at gofmt/rustfmt rewriting the AST to clean it up, but those toolchains were built that way largley because automating anything about large C codebases is so hard.
Let me put this way: If you submit a code prettifying patch to the Linux kernel, it will not be accepted. The risks aren't worth it.
The only real way forward is full migration away from C, for which a better scope is a separate project.
Presumably the last if statement in the diff.
Personally I think the indentation does a good enough job here.
Anyway, poor UX. But of course TZ names could also be argued as poor UX. What if you just did PST/PDT as Los Angeles, CA; Oregon, OR, and Seattle, WA all on separate line items? Sure, it's duplicate data but a backend system (Postgres config files, say) should only store the value of the TZ, i.e. -7 / -8. At least a user could recognize 'oh, I drive to xyz major city occasionally, that's the choice I want'.
To keep ranting, I checked macOS 15 TZ selector for PDT/PST. The selector itself is labeled "Closest city". It has numerous locations in California, a few in Nevada, and a couple in Mexico. No cities in Oregon, Washington, or Idaho (and Hyder, AK... neat [1]).
Closest is a stretch, like I said, over 1K miles from LA. But why several California cities, including minor ones like Oceanside (~175K people), but nothing in Oregon (Eugene, also 175K), Portland (652K), or Washington - Tacoma (220K), Seattle (740K). Note I did not look for the smallest city in the macOS CA list.
It's weird to me. Maybe it's because Oregon == Intel and Washington == Microsoft. ;-)
[0] https://rachelbythebay.com/w/2025/09/11/debtz/
[1] https://en.wikipedia.org/wiki/Pacific_Time_Zone#United_State...
Whether daylight savings time is being used at a given location at a given time of year is a matter of government policy. The city-based timezone selectors should handle that automatically based on the jurisdiction you choose.
> Sure, it's duplicate data but a backend system (Postgres config files, say) should only store the value of the TZ, i.e. -7 / -8.
Then the time may be wrong for half the year depending on where you are.
There's no America/Salt_Lake_City you're recommended to use America/Boise instead. The people in Salt Lake City are about as far away from Boise as you can get and Salt Lake City is more easily recognized as a landmark then Boise. The process of choosing which cities should be landmark cities comes across as faulty and uninformed.
I think it's unlikely they're chosen as landmark cities. More likely, timezones were uniform, then some government likely did their own thing in their jurisdiction. The change was then represented as a new timezone named after the place where the change is centered. IOW, the names have more to do with some random divergence that happened at some point in history, rather than what landmarks are the most recognized for today's timezones. Re: Boise & Salt Lake City, maybe Boise was the first to do their own thing while Salt Lake City had a different timezone. Maybe Salt Lake City later decided to adjust their timezone to fit Boise's to facilitate commerce between their states.
Boise has its own entry because in 1974, the Emergency Daylight Time Act shifted when DST began in Southern Idaho and eastern Oregon. Boise is the largest city in the region.
Technically, if you're in Salt Lake City, you should be using America/Denver, not Boise because of this, otherwise if you say, opened a calendar from 1974, everything will be off by an hour.
If Utah made DST there begin a day earlier this year, Salt Lake City would probably get an entry too.
The issue with this thinking is that all existing datimes referring to that region will continue using the old timezone because the new one didn't exist yet.
You can't really escape the fact Europe/ is going to be after Africa/ and America/ and Antarctica/ and Asia/ and Australia/ and Canada/ and Etc/
It’s common for major cities to be located on rivers that are state boundaries, the area around the city uses the same timezone, and one of the states has a timezone boundary in the middle. Indiana has many tz database entries because of this kind of thing.
There are other fun cases like the Navajo Nation in Arizona.
I live in South Dakota, which is one of them—the Mountain/Central timezone boundary within the state follows the Missouri river. (Locals refer to "East River" and "West River" to refer to the two halves of the state. The capital, Pierre, is technically East River, but is right on the banks, almost dead-center.)
[0]: https://simple.wikipedia.org/wiki/List_of_U.S._states_and_te...
For instance, both Sydney and Melbourne both have entries because Victoria and New South Wales have started/ended daylight savings time in various years since 1970 on different dates like in 1990 when in Victoria it ended on March 18 while in New South Wales it ended on March 4.
It's also why Broken Hill has an entry despite being in New South Wales, they're in their own timezone.
It's very unlikely, but tomorrow some state or major city in PST could decide to add 15 minutes to all their wall clocks. Should your computer's clock change? That depends on what you're using it for...
one can play with timezones all they want, but in the end it's a presentation issue.
Timezone is very much a user interface issue.
What problem are you thinking it would solve?
Traveling.
Again, I don't understand what pain-point or use-case is motivating this "set your laptop OS to UTC" proposal.
... Well, not unless it's some workaround for flawed software running on that machine, perhaps made by developers that misunderstood the business-domain of time zones.
It can't be easily fixed. Imagine I continuously change the OS's setting to always reflect the local time. Now I read a log of events happened last Tuesday. Was it when I was in Europe or when I was in the US? How should I read those timestamps? Then I compare two timestamps from last spring. They differ a few minutes, but one was before the switch to DST the other was after. Which one happened first?
> you want them to be almost always wrong?
I think "wrong" is relative, isn't it?
> Well, not unless it's some workaround for flawed software
Exactly the opposite: there used to be a SO from Redmond that didn't allowed you to set its internal clock to UTC. I think it's been corrected now.
Those two things are unrelated, how frequently you changed timezone settings should have zero impact on your log... unless it means more "settings changed" entries. Changing your current timezone does not move things that already happened.
The event-log entries (A) shall be order-able by actual chronology, excepting relativistic effects, and (B) shall all display under some unified reporting timezone, whether that's hardcoded to be UTC, or the system timezone when exporting/displaying, or a time-zone chosen in an Event Viewer GUI.
If either (A) or (B) are false, that means you're dealing with worryingly flawed software from developers who weren't prepared to implement timezone logic.
> Now I read a log of events happened last Tuesday. Was it when I was in Europe or when I was in the US?
In all cases, your first step is to figure out where those real-world events fall relative to the log entries.
It will always be less work if the event-log display timezone happens to match one of the ones for your departure/arrival times.
> Then I compare two timestamps from last spring. They differ a few minutes, but one was before the switch to DST the other was after. Which one happened first?
Easy: The one that sorts earlier, and if you display logs in UTC they will be visually distinct as well.
It must be possible to sort all system events in an "absolute time line". To do so, the system must know in every moment at what point of this "absolute time line" we are, so that it can stamp the event with such absolute time marker.
Now, we can make it coincide with any known TZ, but once set, we must not change it. I set it to UTC; you might prefer your home TZ, if you have a place you call "home". But then don't change it when you travel or when switching DST on and off!
Of course you can still change the TZ of the time that's displayed to you (so that it's synchronized with the local church's bells, for example, which is convenient and might avoid you a lot of misunderstanding with your local friends), but not the TZ the OS uses for itself.
I find that any choice of home TZ would be arbitrary, so UTC is the best choice.
For many years, when I still had some other clocks besides those included in computers or mobile phones, e.g. wall clocks or wrist watches, those were also set in UTC, thus with no change between winter and DST.
I prefer to keep in mind the current offset of my local time from UTC, and also the offsets of a couple of places where people with whom I communicate frequently are located, and to add those offsets mentally to the displayed UTC time when that happens to be necessary in order to synchronize to some external event, like a meeting or the opening hours of some place. I schedule my own activities, e.g. eating or sleeping, in UTC.
This habit was triggered decades ago by the fact that I found much more annoying the hour change of all clocks to/from DST than changing in my mind the current offset of the local time from UTC, and also by the fact that the local time does not correspond with the solar time anyway, because I an not located on the center of the time zone, so if I want to know when it is noon, I have to also keep in mind the offset of the solar time from local time, which changes when DST applies. At least with UTC, that offset remains constant.
I do not consider myself dumb :-)
On the contrary, I consider that the legal time is designed for people who are so dumb that they cannot remember that during summer they should wake up and go to work earlier than in winter, the same as their ancestors did for many millennia. To be fair, their ancestors did not use a clock for this, but they woke up depending on the rising sun, which took care of this automatically.
Converting timestamps between time zones is perfectly reversible while there is a loss of information when you convert a value to a new currency.
False. This scenario absolutely does happen:
2027-03-08 11:00AM (Local) -> 2027-03-08 2:00PM (UTC)
(tzdata version changes)
2027-03-08 10:00AM (Local) <- 2027-03-08 2:00PM (UTC)Future "time" values are typically not time values. Therefore they must be stored using a different more complex data type, which contains all the information that will be necessary in the future to determine the corresponding time value.
Using the same data type to store both past time values and future time values is in most cases a serious programming mistake, which either wastes resources or is likely to cause bugs.
That added complexity can be much worse than using zoned values for everything.
And that past time value can become a wrong time value if it was converted from a user's local time to UTC before a tzdata version change.
UTC only works for absolute time and accidentally works most of the time for wall time as perceived by humans. It's vaguely similar to how calculating money with floats accidentally works most of the time.
I don't want this scenario:
2027-03-08 11:00AM (Local) -> 2027-03-08 2:00PM (UTC)
(tzdata version changes)
2027-03-08 10:00AM (Local) <- 2027-03-08 2:00PM (UTC)
But the UTC -> UTC conversion is worse. You have to figure out which timezones changed in the new tzdata version, convert them to the new UTC value, then update tzdata. Most libraries seem to only support one tzdata version at any given time, so this is way more complex to me than just having the wall time stored so that it can be converted to the correct UTC value at "decision time".
Well, this isn't unambiguous in all cases either unfortunately: 1:30 AM in Houston, TX (Central Time) on 3-Nov-2024 could be either 0630Z (Central Daylight Time) or 0730Z (Central Standard Time), since the clocks will have fallen back.
Now you can prompt the user for their intent when the time change is known or expected. The ambiguity is now narrowed down to only the situations where the time change could not have been known in advance.
If a historic event occurred at 0300 on Feb 15th, there’s a significant amount of information in knowing the local time. Was it 3am in London, when most people would be asleep. Or was it 7pm Friday night In San Francisco when people were on Valentine's dates, or was it 3pm on Saturday afternoon when people were having a bbq.
If you were to write “the accident occurred at 0300 UTC” then that would imply something very different depending on what the local time was at the time.
How about “the shop opened at 9am every day, without fail”.
Far more useful than “the shop opened at 1600UTC for half the year and 1700UTC for the other half”
In reality that’s not always the case.
Whoah there, no, that's a huge pitfall of sharpened spikes as soon as you deal with events in the future.
If someone proposes an after-work party for "5:30 PM" at the Latverian office in Latverian time, that's not a fixed offset of seconds from now, it's actually a set of triggering conditions.
We can make a decent guess about when those conditions will be satisfied, but don't actually know until it finally happens. At any moment, the administration of Dr. Doom could arbitrarily change the country's clocks. They might skip over that entire hour, or the hour might repeat on that day, or the entire country might cease to exist.
Making a prediction in UTC and storing just that is a very bad idea, because you lost all the original context you need to recalculate a better prediction as things change. Storing the "5PM in Latverian" is how we keep that context.
A past time value (e.g. the times of logged events) actually is a time value that is known and it should be stored as a TAI or UTC value and when desired it can be converted to any date/time format for presentation purposes.
In most cases, a future time value is not a known time value yet and it must be stored as a more complex data type, at the minimum including the local time and the time zone, but for more distant future dates preferably also including the exact geographic location, for the eventuality of changes in the time zone laws.
Therefore, both you and the poster to whom you have replied are right, but you are right only when talking about future time values, while the other poster was right only when talking about past time values.
I think that's essentially what you mean, but it's an important distinction. It's use-case based. Making a data type time-based will not simplify, but only complicate things.
The problem is both the US and Australia have “EST/EDT” - the Australian version sometimes has an A stuck on the front to disambiguate it from the US timezone, but that isn’t always done (especially given some systems insist timezone abbreviations can be max 3 characters). And the problem with disambiguating on the basis of UTC offsets is you’d be surprised by how many people have no clue what any of them are. But “Americas” vs “Australia”, they’ll get that right
And that's just the US, there's almost 200 other countries each with their own laws.
There are guarantees that there will be changes in the future! There are changes regularly, some expected, some not. In some countries the suspension of DST due to Ramadan is decided on the first night of Ramadan itself when a group of elders look at the moon and decide whether Ramadan starts tonight, or tomorrow.
https://www.cnn.com/2025/02/27/travel/ramadan-start-crescent...
Timezones are such a headache. Obviously even UTC for a location varies depending on the time of year.
Even the International Space Station shifted timezones from Houston time to UTC+0.
Curiosity and Perseverance's clocks are UTC but operations run on LMST (local mean solar time) Gale Crater and LMST Jezero Crater- their landing locations. That point is moot until humans start spinning up VMs on Mars which they will one day.
The offset from UTC for a location varies depending on time of year but UTC definitionally has a zero offset throughout the year.
If you’re in the Europe/London time zone your time is equal to GMT/UTC (offset zero) for half the year and BST (offset +1) for the other half.
In other words we have two different types here: Timezones based on location where the UTC offset varies, and the UTC offset itself (like +0100/BST or +0000/GMT/Z.)
And of course, saying 'EST' doesn't actually tell me what fucking time something is happening. Telling me what UTC offset EST is, does.
You can look at the tz data files to see what that looks like.
Choosing a large city you know shares your time zone does make things a bit more “human“.
Because that doesn't tell you when the timezone changes. Two locales can share timezones but start or end daylight savings time at different times.
For instance, Cuba and Florida are both -4 / -5, but Cuba starts and ends daylight savings time 2 hours and 1 hour, respectively, before Florida.
Then there's the fact that locales, once in a while, will change what timezone they're in (like Samoa in 2011) or stop/start observing daylight savings time. Having the timezone set to a place largely solves this problem.
Ideally though, just get rid of DST.
If you want to create an event in a different time zone from your default the select picker it gives you is utterly incomprehensible.
I can't even find the time zone for New York/US eastern in it!
Screenshot here: https://static.simonwillison.net/static/2024/google-calendar...
Typing "New York" gets zero results.
Even knowing the exact name of the timezone I want I still can't find it in that UI!
On android, "eastern" similarly results in a mess of results (but eastern is a bad name anyway...), "new york" does find it.
In the Android app typing "new y" gives
Eastern Daylight Time
18:30 GMT-4
United States, *New Y*orkIf your application can access the current location you don't need to expose a TZ selector to the user. You can figure out what time zone database sector you're in automatically.
https://en.wikipedia.org/wiki/List_of_tz_database_time_zones...
> Sure, it's duplicate data but a backend system (Postgres config files, say) should only store the value of the TZ, i.e. -7 / -8.
Your backend needs to store location because places can switch time zones. The reason for the seemingly arbitrary list of cities is they each define a region where clocks have been synchronized since 1970.
IP geolocation is often wrong and inaccurate. I’ve had a VPS whose IP was geolocated 360 km away, in another country and time zone. But even with residential IPs, they might be pointing to a different time zone in countries with multiple zones.
For the event, your backend only needs to store the timestamp in a timestamptz field and make sure that clients set the correct time zone on session start (this you might want the backend to store in the database too, but probably in the users table).
You can choose that. Set your timezone to Etc/GMT-8. Then, at the exact time your political jurisdiction mandates switching over to PST, go to all your computers and switch them all to Etc/GMT-7. Then do the same thing next year for switching back.
What? That's bad UX as well? Well then, you have to name the correct political jurisdiction that mandates the timezone rules where you live. And that's hard, because so many little tinkerers at state and municipal level decide to change the rules just for their little fiefdom. And they keep changing their minds.
The tz database is looking for the longest-lived identifier that accurately describes that geographic region to which the rules apply. Every time one region diverges from the norm, they need to accommodate the split. They chose continent and city names, because the historical perspective is that city names have remained in use longer than country names.
For your case, however, they have aliases. "US/Pacific" is an alias to America/Los_Angeles", as is "PST8PDT". Set and forget.
In the Debian installer yes, but the stupid Ubuntu installer forces you to pick from a map.
(I don't think the TZ database contains the information to do this though, but maybe it should?)
Except when you can't forget, as in the original case for this blog post in the first place.
I also like that she’s able to reproduce it but I feel like this should have been a test check in the distro to ln the US zones as to not mess up older codebases.
In the end, standardizing is the right way to go. If you’re a maintainer, think about adding tests and checks for files that if moved might break things that depend on it. Maybe a service that watches for fd access to those and gives a warning, I don’t know what the best approach would be but it would be nice to have something at the user land level that said “Ahh ahh ahhh, you didn’t say the magic word”
Debian 13, Postgres, and the US time zones - https://news.ycombinator.com/item?id=45218111 - Sept 2025 (142 comments)
That's what the authority that defines the zone calls it. Using any other name is adding a useless layer of abstraction.