[godelski]

A bit off topic but why are calendars so hard? I recently moved from android to apple and I'm just really impressed no one does a dedupe operation on calendars. Seriously, who works on these things? I'll write you the regex if you really really need it but damn, if you're going to push fancy AI on me to make my life easier at least take care of all the annoying trivial bullshit that makes my life harder.

[ttepasse]

My pet theory is that it is the data model: everyone must support ical/Exchange and those only support events with a textual description.

Events may be a part, but not the only part of a good calender. There is a lot of stuff which could be part of a calender but doesn't necessary fit into the event paradigm: Fuzzy events like the car inspection you are putting of. Transit times - not just the fixed time span you get today, but integration of public transport or current car traffic, so that you can plan. If you're menstruating a period tracker. Your health and sleep data. The weather, the day-night-cycle for your location. All background data with time and date, which we have, but not just in our calendars.

And even traditional events are limited: I'd like to have a general repeating workday event but also like events in that workday as extra events. With a normal digital calendar they are clashing. Hierarchical events would be a solution.

Or multi-layered events, some years back there was this blog post on HN which made me rather unhappy with iCal:

https://julian.digital/2023/07/06/multi-layered-calendars/

https://news.ycombinator.com/item?id=36617504

[TeMPOraL]

My pet peeve: buffer time. Or, rather, lack of it. Almost every calendar event requires some extra buffer time before and after, for things like getting to event location and back, and/or preparing, etc. No calendar app I've ever used even supports the concept, much less does anything to help. Right now, I resort to having a separate calendar for "buffer time" events, which I currently fill manually; I'm working on automating this, as almost all events also have fixed buffer time, that's known to me, but again there's no way to record that fact in existing calendar software.

[godelski]

This is a great point! And in the class of problems I was trying to illustrate. There's a lot of low hanging fruit here. We're programmers, we should be able to take our frustrations and turn them into solutions! If we're not, then we're doing something wrong lol.

FWIW, in Apple Calendar there is an option for "Travel Time". This will prepend the event with some buffer value. When looking at the event on my calendar, in my iPhone, and ONLY in my iPhone there is a little : (vertical dots + car) symbol and saying (e.g.) "15 minutes travel time", while the event has a vertical continuous bar next to it. On my Macbook, the whole event is just expanded.[0]

My pet peeve: dramatically differing interfaces between devices. I do understand that at times this must happen. But there's no reason for it to in this case. But worse than that, when CREATING a new event on my Macbook it is just so much clunker. I don't see "Starts" and "Ends" until I click on the time, where it instead expands. Then I can see the "travel time" option (in iPhone it is "Travel Time") and if I click the end time I can only select from a pre-determined time of 30min to 3hrs with 30 minute intervals. If I select "starts" I don't get this. In both cases I can __type__ the actual numbers in.

[0] Update: while writing this comment, my macbook's calendar updated to show the vertical dots but does not include the little car nor the text (again with continuous vertical bar under actual event). Why did this take so long? "Just works" my ass. Arch linux has been more stable than trying to live by the Apple way.

[ivan_gammel]

This is really a good question. Progress with calendars has stopped 10-15 years ago, the only major innovation relevant to it was Calendly, but that’s it. I guess it’s hard to imagine new features or even incremental quality updates as a viable business idea.

[jamalaramala]

> I guess it’s hard to imagine new features or even incremental quality updates as a viable business idea.

There is market for privacy; but the missing piece, after so many years, is how to synchronize calendars in a seamless way.

[ivan_gammel]

It’s not just that. Manual entry is still a thing happening too often. Yes, some apps allow to put your booking in a calendar, but that doesn’t cover all use cases. You call, make an appointment — getting an ical via SMS after the call would be great. You open a website, then check there the opening hours or schedule, taking that information to your calendar app would be great. Etc etc.

[woleium]

do you need that clutter on your calendar though, when a voice query to perplexity (or siri et al when they catch up) will answer your question about opening hours when you need the information?

[ivan_gammel]

Voice queries exist in a small bubble, most users in the world are not used to it. And of course AI may never be able to catch up with hyperlocal information that may not even be available to robots.

As for the clutter, no, there won’t be clutter. There exists a conventional 3-view UI for calendars, but there can be more context-specific variants. E.g. separating awareness from booking on customer journey means that first I add a reminder with the opening hours and later calendar will present an opportunity to book.

[freddie_mercury]

A quick Google tells me Perplexity doesn't support my language, so no a voice query to Perplexity won't work.

"Perplexity is now available in Korean (한국어), Japanese (日本語), German (Deutsch), French (Français), and Spanish (Español)"

It's missing 3 of the top 5 languages spoken in the world today. It only has 3 of the top 10 most spoken languages in the world.

Also everyone in my country drives scooters and I don't see how voice commands are supposed to work with that kind of setup.

[godelski]

  > Perplexity is now available in Korean (한국어)

Wow, usually Korea takes awhile to get things.

  > Also everyone in my country drives scooters and I don't see how voice commands are supposed to work with that kind of setup.

Even if they were supported I would suspect that you would have some issues. My partner is Korean (I'm not) and when she goes back to Korea and we video chat while she's on the subway then her airpods' "voice isolation" systems actually amplify background noise rather, drowning her out, rather than doing the opposite (its intended usage).

Being an ML researcher... I have some suspicions as to what is the root cause. I can't tell you exactly without having access, but I can tell you that it is VERY common for models to be trained without sufficient augmentation, let alone without sufficient diversity in data. So it is fairly common for the system to not be useful in certain environments. Worse, when you hyper focus on maximizing test data results you are extremely likely to perform data leakage (this is quite common and at this point I'd say you should assume it exists) and not enough evaluation on completely held out sets. Especially data which is significantly dissimilar to what you trained on (even when it is "in distribution"[0])

[0] This term can be a lot of things... so it must be interpreted in context.

[TeMPOraL]

That's a filtering issue. For any case less trivial than checking if something is open right now, calendar view is superior. If you want to plan something, calendar view spanning multiple days, combining the thing with your existing plans and events, is pretty much a must.

[thih9]

It’s difficult to maintain basic features in a popular calendar, at least according to the ms exchange team blog:

> calendaring is particularly tricky. Why’s that? Well, consider time zones for a start – a meeting you set up isn’t necessarily in the same time as it is for me, and then you also invited people, from a whole bunch of other time zones (did you know some time zones are 30 mins off, not a full hour?) (…)

https://techcommunity.microsoft.com/blog/exchange/calendarin...

Then there is the “falsehoods programmers believe about time” list: https://gist.github.com/timvisee/fcda9bbdff88d45cc9061606b4b... , with some counter examples: http://www.creativedeletion.com/2015/01/28/falsehoods-progra...

[godelski]

I don't buy it. I do think the problems are not trivial but it's not like there aren't trivial problems within the larger problem.

Let's take a trivial problem: holidays. Every account you have comes with a holiday calendar. So if you got a Microsoft, Google, and Apple account you'll have entries in triplicate. These are all day events with identical names. There's a lot of hints to tell you that these are identical and you only need to display 1.

A bit harder would be birthdays. These might not be in a unique calendar. But the word "birthday" is a huge hint when it's an all day event.

I do agree that defaulting to a position of not destroying data is best (though here we'd never destroy data). But that doesn't mean we can't offer users assistance. We do have the capacity to let users merge events or in some way note that they are identical.

One more issue, why the fuck can't I move an event from my Google calendar to my Apple? In the worst case, just copy the damn thing and then hide the event. You can do this even when you don't have write access to the Google calendar.

So the thing I don't buy is that we can't do more. A programmer's job isn't easy. But it involves problem solving. I think there's not enough grumpy yet motivated people who will voice up such things because management usually says that it's not important. But these non important things add up. Life is complex so the little things actually matter a lot.

[OptionOfT]

Weirdly enough an all day event in Office 365 isn't an all day event on a day.

E.g, I record my wife's birthday in my calendar.

I create an event on April 12th, and mark it as an all day event.

What actually happens is that the calendar records an event that starts at midnight, and ends 24 hours later, and is marked as an all day event.

But, when you change timezones the 24 hours actually shift, which can be very weird when you get notifications that are 6 hours earlier.

[godelski]

Sounds like something that could be solved by a flag.

  if all_day_event:
     show_option("timezone invariant")

If it is timezone invariant, then you can make it the 24hr (or whatever because there might not be 24 hrs in a day...) period relative to the current timezone.

This isn't an unsolvable problem, it is just people not seeking solutions.

I have a theory: a lot of tech products suck today because either:

  - People are not dogfooding[0] (using their own products)
  - People are not raising concerns
  - People raise concerns but issues are de-prioritized in favor of more flashy things. 

None of these things are great and they can only happen while a company has significant marketshare and monopoly like status. In essence, it is internal rot. Someone (plural) is disconnected from the actual product. They are out of touch with what the users want.

And I mean here's the thing. I could go ahead and write a wrapper for all of this and wrap in all my calendars. But that's a ton of work, there's probably no market, but if there was a market then the realistic result is that my success would result in the big players implementing the same feature and thus devaluing my work even though I increased the value of theirs. So I can do this as an open source project, but boy are there a lot of other higher priority issues like this. The real problem is that these issues can't be resolved within these big companies.

For whatever reason that is, it is bad business.

[0] https://en.wikipedia.org/wiki/Eating_your_own_dog_food

[thih9]

> make it the 24hr (...) period relative to the current timezone

What if the event gets shared - should the dates still be relative to the timezone of the creator? Or to the most popular timezone? Or should they be different for everyone? Or something else?

What if the event is recurring and the creator travels between the occurrences, should the times change? When should the times be updated? Should all events change, or only the future events, or only the closest one?

What if a participant uses multiple devices in multiple timezones at the same time?

…etc. Perhaps there are ways to solve all these and more; my point is that in the context of calendars even adding a flag can get complex fast.

[godelski]

  > What if the event gets shared - should the dates still be relative to the timezone of the creator?

You mean

  What if the event gets shared with someone who doesn't have a calendar that implemented this hypothetical yet super basic and highly useful feature.

Well then, it falls back to the standard 24 hr block, right? Yeah, it would be the timezone that it is currently being shared from because our flag was just telling the event to move itself with the timezone instead of being static.

Is that really that bad? This doesn't sound like a reason not to do this. Because the worst case scenario is that... we end up with the current implementation...

  > What if the event is recurring and the creator travels between the occurrences

Here's some shitty pseudocode. You could do this and store as UTC time but I think storing with timezone makes the point clearer.

  class Date(...):
      def __init__(self,
                   year   : int = 1970,
                   month  : int = 1,
                   day    : int = 1,
                   hour   : int = 0,
                   minute : int = 0,
                   second : int = 0,
                   ...
                   tmz    : str = "PST",
                   **kwargs):
          super().__init__()
          ...

      def start_of_day(self):
          self._hr, self._minute, self._second = self._sod

      def end_of_day(self):
          self._hr, self._minute, self._second = self._eod

  class Event(...):
      def __init__(self, 
                   start_time    : Date,         # date event begins
                   end_time      : Date,         # date event ends
                   all_day_event : bool = False, # is all day event 
                   tmz           : str = "PST",  # timezone
                   tmz_invar     : bool = False, # timezone invariant
                   **kwargs):
          super().__init__()
          ...
          self._begin = Date(start_time, tmz=tmz).start_of_day()
          self._end   = Date(end_time, tmz=tmz).end_of_day()
          if all_day_event:
             self._begin = self._begin.start_of_day()
             self._end   = self._end.end_of_day()
          self._ade = all_day_event
          self._tmz_invar = tmz_invar

  class Calendar(...):
      ...
      def display_events(self, ...):
         ...
          for event in self.events:
              if event.timezone_invariant:
                  self.display(event._begin._sod,
                               event._end._eod)
              else:
                  self.display(event._begin.tmz_offset(self.current_tmz),
                               event._end.tmz_offset(self.current_tmz))

All day events are marked as starting at the beginning of a day and ending at the end of a day. That being from our Date structure, which holds Month, Day of month, hour, minute, second, whatever precision, as well as special things like when the beginning of the day is and when the end of the day is. Here the function start_of_day() and end_of_day replaces whatever input values for {hour, minute, second, ...} with the appropriate values and attributes _sod and _eod are those values.

This assumed you stored events as a time with a timezone. If you stored events in UTC you'd need different logic. I mean time is messy, but either way we're moving stuff around. I mean we're just _displaying_ offsets when moving timezones, right? Because it would be wild to update all events every time we moved timezones, right? So we have to do some logic to take the date and display it relative to its time

With a timezone invariant flag all we're doing is saying

  Look at the year, month, and day. Ignore time and instead, use the day's start time and end time. 

  > What if the event is recurring and the creator travels between the occurrences, should the times change?

No, IT IS AN ALL DAY EVENT

  > When should the times be updated?

They don't, IT IS AN ALL DAY EVENT

  > What if a participant uses multiple devices in multiple timezones at the same time?

Doesn't matter, IT IS AN ALL DAY EVENT

I'll raise you a question

  What do you do if this is not an all day event and is in a specific time?

Think about that answer.

It is really not a hard problem. Because you are storing the events in some database. Moving timezones is then a conversion process. The invariant flag is just saying "keep the time relative to the day."

  > my point is that in the context of calendars even adding a flag can get complex fast.

I don't disagree, but I'm failing to see where this specific issue is an actual problem.

[TeMPOraL]

> and then hide the event

Speaking of - hiding events is one of the features of Sectograph[0] I most frequently use. I think local event modifications on shared calendars are another big feature missing almost everywhere. In my case, hiding events lets me just sync the shared family calendar synced directly to my watch, and just hide the events I don't care about, as to not clutter the display, without deleting or otherwise affecting what my wife sees on her devices.

> So the thing I don't buy is that we can't do more. A programmer's job isn't easy. But it involves problem solving. I think there's not enough grumpy yet motivated people who will voice up such things because management usually says that it's not important. But these non important things add up. Life is complex so the little things actually matter a lot.

In Admiral Adama's voice: So say we all.

--

[0] - https://sectograph.com/ - may seem like a bit weird gadget, looking at the app screenshots, but it's true value comes from the same circular view being available as a watch face, giving you always-available view of the next 12 hours on your wrist. This is why I started adding bus/tram departures and event buffer times to my calendars - they become useful for quick checks on the go. IMHO, it's the only actually useful watchface on the Android smartwatch market.

Which is exactly why Google had to fuck it up with their Watch Face Format idea, aka. Manifest V3 For Smartwatches, aka. removing "smart" from "smartwatch". Fortunately, I'm not forced into it until my current watch dies.

[TeMPOraL]

> It’s difficult to maintain basic features in a popular calendar, at least according to the ms exchange team blog:

There's a large amount of things that regular people do with calendars every day, that could be automated or managed through better interaction paradigms, and to which those issues don't apply. That's a lot of low hanging fruits, that everyone's leaving to rot.

Like:

> meeting you set up isn’t necessarily in the same time as it is for me, and then you also invited people, from a whole bunch of other time zones (did you know some time zones are 30 mins off, not a full hour?)

99% of my calendar use involves events happening within days or weeks, and concerning me alone, me and my wife, me and my acquaintances, or me and some random local company. I don't care about timezones for those - they all happen in the same one, with the same DST shift.

(The remaining 1%? Timezone shenanigans happen at few distinct points in a year, and everyone knows to be careful around those and communicate out-of-band if needed. So again, not an issue in practice for users like me.)

Microsoft is thinking about calendars as tools for employees of multinational corporations. But calendars aren't only useful for managers frequently flying intercontinental; there's a lot of regular folks using them for affairs much more localized in time, space, and social graph.

[ivan_gammel]

Timezone stuff is trivial with modern libraries like Java Time API. Event entity and event identity is hard.

[karamanolev]

Eh, I don't buy that. Storage in UTC and converting to a local timezone for display / notification purposes has been the standard practice for a looong time.

What I would believe, I don't know if true, is that different applications support subtly different feature sets, integration has varying levels of support and correctness and so on. Additionally, I feel like vendors are incentivized to offer good support within their ecosystem, but integration with the outside world is a second-class citizen.

[happytoexplain]

"Just store it in UTC" will cause bugs for human time, which calendaring is the most complex example of (because it contains basically every other use case). Don't throw away the date and time the user entered. How you store it is the subjective part, as different choices have different risks. But don't just throw it away.

Also, the existence of this thread should be good evidence that it is indeed not that simple.

[ivan_gammel]

Storage in UTC is generally wrong for future events and doesn’t make sense at all for unbounded sequences.

Edit: better approach will be to store event as it was defined and include processing hints, e.g. when next UTC time of occurrence can be reliably established.

[pluies]

> Storage in UTC and converting to a local timezone for display / notification purposes has been the standard practice for a looong time.

All good until summer time kicks in and your recurring 9am meeting now starts at 8am, oops.

[haakonhr]

But summertime is a (at least here) a separate timezone, e.g., CET vs CEST

[happytoexplain]

The point is that humans want a meeting's clock time to change when they travel to a new time zone, but not when their local time changes (either due to DST or a legal change).

Unless of course not everybody invited shares the same DST patterns, in which case it has to change for some of the participants (how do you choose which?).

Also, the relationship between the datetime the user chose and the UTC time you store can be retroactively changed by legislation.

[ivan_gammel]

Timezones change.

[makeitdouble]

UTC storage is standard practice, but not the best practice nor something that solves all the major issues a good application should be dealing with.

I'd qualify it as "good enough to not get fired" grade of decision, for people who don't care that much about calendaring in the first place or have the luxury ignore complex cases (we're back to "why are calendars so hard?")