607: 2038
- Not to be confused with 2038: Hazard Symbol.
| 2038 |
![]() Title text: If only we'd chosen 1944-12-02 08:45:52 as the Unix epoch, we could've combined two doomsday scenarios into one and added a really boring scene to that Roland Emmerich movie. |
Explanation[edit]
The 2038 problem is a well-known problem with 32-bit Unix-based operating systems. Unix time is stored as a 32-bit signed integer on these systems, counting the number of seconds since 1970. In 2038, we overflow the highest number we can store in signed 32-bit integers, leading to unexpected behavior. The switch to 64-bit operating systems will most likely be complete by the year 2038, which is why Randall is relieved. The reference to Y2K is a throwback to the year 2000 problem, in which people were concerned that computers storing years as two digits (e.g.: 99 to represent 1999) would cause problems when the year 2000 began because 00 could have been interpreted as 1900 by error. That Y2K issue was covered widely β with only some small mishaps β but calculating dates beyond 2038 is still not solved on many 32-bit UNIX based systems today. The "even WORSE" is possibly referring to how our increased reliance on computers means the bug could affect many more vital systems, but with Y2K passing by relatively uneventfully especially in light of the hype that preceded it caused people to take this sort of problem less seriously.
The title text is a reference to the film 2012 which is about the world ending in December of 2012, at the end of the Mayan calendar. If the designers of the UNIX operating system had used 1944 as their epoch instead of 1970, then the UNIX crash due to a variable overflow would coincide with the end of the Mayan calendar. Thus, the implication is that there could have been a boring scene in the movie where the UNIX time rolls over and nothing happens and no one cares β because the world doesn't exist any more.
Transcript[edit]
- [Text above the panel:]
- I'm glad we're switching to 64-bit, because I wasn't looking forward to convincing people to care about the Unix 2038 problem.
- [Cueball's Cueball-like friends asks him a question. Cueball raises his arm above his head while answering.]
- Friend: What's that?
- Cueball: Remember Y2K? This could be even worse!
Trivia[edit]
- This is the first numbered comic whose title contains no letters.
Discussion
- Can anyone explain the mouse-over text? Saibot84 (talk) 23:02, 7 May 2013 (UTC)
- Good thing it's explained now, because I was relating 1944 and apocalypse with WW2. 108.162.212.196 21:57, 3 January 2014 (UTC)
"calculating dates beyond 2032 is still not solved on many 32-bit UNIX based systems today". Is the year 2032 a typo, should be 2038? If not, what is the relevance of 2032, should be explained. --Pudder (talk) 07:30, 12 September 2014 (UTC)
Woah, I learned about the 2038 problem yesterday, and I clicked "Random page" today and got this comic! Anyone remember what that phenomenon is called? LuigiBrick (talk) 13:57, 6 January 2017 (UTC)
- The phenomenon might be called apopheny. 172.68.213.181 07:59, 13 July 2024 (UTC)
It's the Baader-Meinhoff Phenomenon 162.158.78.22 14:39, 28 August 2018 (UTC)
The phenomenon is called "coincidence." -- Davidh (talk) (please sign your comments with ~~~~)
This wiki will have it's own 2038 problem, as when we get the 2038th comic (assuming both explain xkcd and the comic itself are still ongoing), http://www.explainxkcd.com/wiki/index.php/2038 will have to be shared by two pages (currently, this link redirects to this comic, as does this one) 172.68.54.10 17:03, 1 September 2017 (UTC)
- Writing to you from the day of comic no. 2038, I can reassure you that our admins solved the problem. --172.68.150.64 19:55, 27 August 2018 (UTC)
Maybe the "even WORSE" part was a pun because things will roll over 136 years instead of 100 as Y2K did. I don't want to add it without discussion first 162.158.75.136 16:32, 25 March 2019 (UTC)
- it's possible, but I think it's mostly in reference to how much more we use computers now than we did then.--Twisted Code (talk) 15:59, 24 November 2022 (UTC)
I never really understood why people were worried about the Y2K bug. At worst, surely it would just interpret it as 1900 instead of 2000? Beanie (talk) 13:58, 19 April 2021 (UTC)
- despite its simplicity, don't think you want to have a bill that the computer system thinks is overdue by 100 years. Same problem here.--Twisted Code (talk) 15:59, 24 November 2022 (UTC)
- Y2K was understood to be a big issue, so a lot of work was done to prevent it. So "normal" people might think that the "Y2K prophecy" turned out to be about nothing. Which is what can make 2038 worse, as people won't understand it as an issue, won't do work to prevent it, so Y2K might happen with delay but in full force. Maur11cio (talk) 08:58, 17 March 2026 (UTC)
- As a person who went through Y2K prep, even the work that was done that didn't prevent anything, let us know for sure (by checking) what would still work, and/or give impetus to another change (e.g. how many machines were still Win95/98, or even Win3.x, that could be usefully upinstalled to be (or replaced by) 2K...). And, you, know, just a general excuse for a more thorough audit. Like making sure that no-one has plugged in anything to the network that they really shouldn't have, or working out what things didn't go through the Change Control procedure and should have (and making sure everyone knows how to go through that, both end-users and the people administring it).
- It absolutely definitely caught not-so-compliant 'Y2K fails', a handful of years ahead of time. Those that might have caused anything from no effective issues, or only minor disruption, or could have been really problematic. And such effects could have been a few years or more down the line (when things actually ticked over, upon the 1999-2000 'false milleniu') or even immediately (if due-dates "in a few years" started to be generated that were treated as 90-odd years overdue). But it also had great associated benefits for general 'housekeeping', doing things that we might not have done.
- My company at the time had started in the '80s, at least my bit of the larger company that it now was. Other fragments might have been older, or newer, or created by the larger company to complement or revamp the other bits it had drawn in. There were so many different approaches to doing certain things. Our local Mail Handling System was a Netware-based thing, it had to be screw-plated onto the global Microsoft Exchange one. I think others had a Lotus Notes-based system that also had had to be jacked in accordingly. I can't remember whether or not the MHS was failed with Y2K (I think it was Ok, and was kept on as a peripheral legacy system for dealing with certain legacy contacts that might occasionally crop up further down the line), but it was a good reason to rationalise the corporate-wide system to be far less patchwork. And implement a useful biennal re-review cycle that picked up future problems with 'legacy' resources that might have just stayed stuck in the corner until someone wanted to rearrange the office (or the 'magic blue smoke' escaped from it, and nobody who knew what it was for(/what it was actually doing/how to now quickly do it with something else) was still working there.
- And, usefully, the processes that Y2K implemented should be useful for 2038 (not in my Y2k-era company, it's no longer trading, but that wasn't due to Y2K or from Y2K-mitigation), because all the old problems have been 'solved' (people should know how to manage such things). Completeky new problems (just letting AI control everything, perhaps, might be capable of flawlessly dealing with 2K38 issues, but have never experienced them so could be an out-of-context issue that they can miss until they're struck by it, rather than Y2K-tested human anticipating things and thinking ahead of time/telling their successors to do so) are another. 81.179.199.253 14:47, 17 March 2026 (UTC)
Add comment

