Core Dump

How does my computer know about time zones?
Posted on June 7, 2016.

Even simple-looking things can sometimes have a surprising twist in software. And everyone who thinks that programming dates is easy to get right the first time probably hasn’t done much of it.
– Peter van der Linden (Expert C Programming)

Main Problem and Motivation

Recently, a colleague of mine introduced me to World Time Buddy, a convenient world clock and time zone converter used to schedule online meetings. I liked the application and I decided that I would like to develop a clone of it that would work locally with no internet connection and output the specified time zone data as plain text in the command line.

So I made a quick utility that does this exact thing (or at least I have some confidence that it does). The problem was that I did not know how does my computer have all this information about time zones used by different countries. My mind was boggling. I was trying to think of all the different edge cases that can show up on my simple world clock and at the same time I was wondering if my computer can handle them all by default.

What Can Go Wrong

There are many timing problems that can show up within a single computer system (e.g., the RTC and the system clock drifting apart from each other). Fortunately, there exist mechanisms in the OS and hardware that take care of most of them for us. The main thing I was unsure of was the following: how does one go about making an application that maps all (or at least most) locations to a time zone and a Daylight Saving Time (DST) schedule, if applicable, at any moment.

As an example, let’s consider a clock that shows the current time in Chicago(US) and Athens(Greece). Athens is 8 hours ahead of Chicago normally and both cities observe DST. The issue is that their DST schedules are not synchronized. At the year of this writing, Chicago turned their clocks 1 hour forward at March 13th and it will turn them back 1 hour at November 6th. At the same time, Athens turned their clocks forward at March 27th and will turn them back at October 30th. This means that between March 13th and March 27th the time difference was 7 hours instead of 8 and the same will happen again between October 30th and November 6th.

After thinking of the above example, I wanted to search for other edge cases and get a better understanding of the challenges of building a world clock. It turned out that these edge cases are so many I can’t even list them all. For example, several areas recognize partial offsets within a time zone (e.g., 30 and 15-minute offsets). Another example is that there exist places that are on the same time for half a year and an hour apart for the other half because even though they are all in the same time zone, some of them do not adhere to DST. As one reads more examples that differ in nature like the previous two, it becomes obvious that these “irregularities” are not edge cases but the rule.

So the next question is the following: Even if we were to store all these information somewhere, how often would we have to change it? In other words, how often do the different areas around the globe decide to change their time policy?

It turns out to be more often than most people think. To give an example, last year (2015) around 15 different places (countries or areas within countries) changed something about their time policy, either temporarily for that year or “more permanently” for the years to come. Chile abolished DST while Turkey delayed its DST end time by two weeks and North Korea introduced its own time zone.

How Do We Manage

Most operating systems today keep the system time as UTC and instead of having a single time zone set for the whole computer they allow the user to configure the time as they wish. In Unix-like systems, a user generally sets the timezone during the installation of the OS and can later vary the time zone of a process using the TZ environment variable. This means that the OS uses some kind of local database for countries and time zones. On my Linux machine this database exists in /usr/share/zoneinfo/ and gets updated during my system updates by my distributions’s updating mechanism/application.

This local database with all the timezone information generally comes from the IANA time zone database (sometimes called OlsonDB) which is an ongoing collaborative compilation of information about all the timezone and DST cases intended to be used with computer programs and operating systems. This database does not only contain the current time rules in the different areas, but also the history of time changes from decades ago. This can be useful for applications that focus on exact duration of events rather than specific points in time.

If your OS does not ship with any of those databases and has some other mechanism you can still develop applications that use them. Many programming languages in their attempt to be more portable, generally ship with a copy of IANA or similar.

Final Note

You can read more about IANA and the data of its data on its wiki page or learn more about the procedures of adding and changing time rules in it by reading the archives of its submissions mailing list. TimeAndDate is also a useful resource to keep track of the changes in time in the different areas.

The utility mentioned in the introduction that I’ve been trying to write is incredibly simple considering its requirements and therefore it suffices to just plug it into IANA. There exist more complicated groups of programs out there (e.g., vehicle or package tracking) whose requirements may not be that simple. In that case, the programmers should be more aware of the pitfalls of keeping track of time in their software.