Comment by p_ing
2 days ago
Read the previous blog post [0] -- this one is disjointed without it. I dislike TZ selectors that use locations (cities, countries, etc.). Let PDT be PDT(-8) and PST be PST(-7). Why do I need to choose Cupertino, CA (or LA in the blog post example) -- locations over 1k miles away from me? And while I certainly understand where Cupertino is and how it relates to my TZ, what if someone else doesn't? Cupertino isn't a major population center.
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...
> 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.
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.
Maybe it could be a forcing function for the less-majority DST-rule countries to align to the majority, once they figure out that their IT systems have trouble twice a year they'll be sure to vote on it.
Ideally though, just get rid of DST.
> Let PDT be PDT(-8) and PST be PST(-7). Why do I need to choose Cupertino, CA (or LA in the blog post example)
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.
> The city-based timezone selectors
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.
The reason why certain cities get entries is because sometime after time began (1970-01-01), the region they're in changed their timezone (usually adopting or dropping DST or changing when it began). Those changes need to be recorded in the timezone database. The largest city in the affected area is typically chosen to be representative of the area.
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.
8 replies →
> 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.
Salt Lake City is a 4.5-5 hour drive from Boise. If you want a better example, there are no cities in Texas in the list, and Chicago is a lot further away than Boise is from Salt Lake City!
1 reply →
I run all my clocks and computers in UTC. I hate that I have to select "Reykjavik" in the Ubuntu installer instead of just "UTC".
You can't just go by time zone names because there are weird exceptions, like most of Arizona not doing DST. Then there's Indiana, which didn't do DST until 20 years ago, and there are some counties that switched time zones when the DST law took effect... if you're in one of those counties will you just accept old timestamps being an hour off? Granted, this gradually becomes less of an issue the further we get from the change. But nothing guarantees that there won't be further changes in the future.
And that's just the US, there's almost 200 other countries each with their own laws.
> But nothing guarantees that there won't be further changes in the future.
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...
Even abbreviations have issues. PST (UTC+8) is also Philippine Standard Time. EST could mean Eastern Standard Time in Australia, granted that nowadays is AEST.
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.
> Obviously even UTC for a location varies depending on the time of year.
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.)
A bit of an edge case, but there's also the problem that the time zone you pick today might not be the time zone you have tomorrow: The jurisdiction you're in can change what their clocks use on an institutional whim.
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...
> 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?
If 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.
I’m not letting applications access my location just to let it pick a time zone.
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.
> Your backend needs to store location because places can switch time 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).
> Let PDT be PDT(-8) and PST be PST(-7)
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
I suspect a large number of users might choose PST if in California, when they really mean PST/PDT. Or perhaps in the summer they would semi-correctly choose PDT.
Choosing a large city you know shares your time zone does make things a bit more “human“.
This absolutely would happen. Most Americans I know refer to their timezone as EST/CST/MST/PST even after March.
I’m currently working at one of the world’s largest consumer banks, and for a company that really should care about the accuracy of such things, I see “EST’ frequently all summer long.
1 reply →
> I dislike TZ selectors that use locations (cities, countries, etc.). Let PDT be PDT(-8) and PST be PST(-7).
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.
> You can choose that. Set your timezone to Etc/GMT-8.
In the Debian installer yes, but the stupid Ubuntu installer forces you to pick from a map.
Why would you use Ubuntu in the first place?
1 reply →
If the map let you pick any arbitrary point rather than having to pick a specific city in the TZ database, that would actually be a better way than a list of cities. See the discussion elsewhere in this thread about mapping Salt Lake City -> (Boise vs Denver).
(I don't think the TZ database contains the information to do this though, but maybe it should?)
The good option is not to yell at clouds but to open a bug report, provide a patch, use Debian or change it after install.
1 reply →
Offset alone is not enough because different TZ names also point to different DST schedules (current and historic) and past changes.
You can look at the tz data files to see what that looks like.
I just fired up the selector on macOS Sequoia and I see Seattle, Portland etc. when I type in the combo box (I'm not near there right now). Guessing that you're probably just seeing a ip geolocation that detected your location as further away.
I'm curious as to what people in Phoenix would select as their timezone under your proposed solution?
My prize for worst time zone UI still goes to Google calendar.
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...
isn't it at GMT-4 Eastern Time - New York?
Here's the search results you get if you type "eastern" in that interface: https://gist.github.com/simonw/c2a73e40bbc87b8c25333207c551f...
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!
2 replies →
if one is serious, one just chooses UTC.
one can play with timezones all they want, but in the end it's a presentation issue.
> in the end it's a presentation issue.
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.
Here the problem is that past "time" values and future "time" values should be different data types, e.g. "Time" and "FutureTime".
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.
1 reply →
UTC helps store specific moments in time. Notably it does not solve for "dates" nor recurrence. Many of my hairs have been lost to third parties thinking they've created viable systems simply because they use UTC.
UTC or TAI are the right solution for storing past time values. Any other information is superfluous and it makes things more complicated than they should be.
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.
8 replies →
It's a similar painful insanity as: "Other currencies? Just convert them to US dollars on save, and store that number in the database, nice and easy, no problems."
4 replies →
Running UTC as a clock on an end user workstation is about the dumbest thing you can do (unless they reside in UTC).
I must be the dumbest person in the world, then, because that's exactly what I've been doing on all my computers for 20 years (and I don't reside in Iceland).
Let's hear why you think this.
5 replies →
Thing is... noone resides in UTC at all.
3 replies →
I have set my clock on UTC on all my workstations, desktops and laptops for many decades. This has been particularly convenient on the laptops used in business trips.
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.
I do this on all my workstations, phones, wall clocks, Google calendar, and everything. After a lot of frustration I found it the lesser of evils to just think in UTC regardless of where I am in the world, and convert for local people on the fly.
Doing this at the system level was one of the better ideas to come out of unix.
In an operating system especially good for multi-user remote access, why would you assume all your users are in the same timezone?
Timezone is very much a user interface issue.