Time is Hard

A look at why it's such a pain to deal with Date/Time in code, and some interesting corner cases. Given at IPRUG.

Transcript

1. Time is
   Hard
   @ChrisSinjo
   IPRUG – 4th June 2013
   

   View Slide

2. What’s so
   interesting?
   

   View Slide

3. Subtlety
   

   View Slide

4. Time.now is easy
   

   View Slide

5. Time.now is easy
   (We’ll see about this)
   

   View Slide

6. Most apps care
   (Conjecture)
   

   View Slide

7. We won’t cover…
   Inconsistent servers
   Unreliable clocks
   Clocks in space
   

   View Slide

8. So what can go
   wrong?
   

   View Slide

9. Server timezone
   

   View Slide

10. sinjo@widdershins:~$ irb
    > Time.now
    => 2013-05-22 22:10:53 +0100
    sinjo@hydrophobe:~$ irb
    > Time.now
    => 2013-05-22 21:11:07 +0000
    

    View Slide

11. sinjo@widdershins:~$ irb
    > Time.now
    => 2013-05-22 22:10:53 +0100
    sinjo@hydrophobe:~$ TZ=UTC irb
    > Time.now
    => 2013-05-22 21:11:07 +0000
    

    View Slide

12. Time.now is easy
    

    View Slide

13. Time.now.utc is easier
    

    View Slide

14. User timezone
    (browser)
    

    View Slide

15. var offset = new Date().getTimezoneOffset();
    offset = -offset;
    var time = moment("2010-10-20 14:30 +0000",
    "YYYY-MM-DD HH:mm Z");
    var localTime = time.add("minutes", offset);
    

    View Slide

16. var offset = new Date().getTimezoneOffset();
    offset = -offset;
    var time = moment("2010-10-20 14:30 +0000",
    "YYYY-MM-DD HH:mm Z");
    var localTime = time.add("minutes", offset);
    Changes with DST
    

    View Slide

17. Trust the
    user’s machine?
    

    View Slide

18. DST
    

    View Slide

19. HTTP Logs
    [27/Oct/2013:01:30:00] "GET /some/page"
    [27/Oct/2013:01:59:00] "GET /another/page"
    [27/Oct/2013:01:30:00] "GET /some/page"
    

    View Slide

20. HTTP Logs
    [27/Oct/2013:01:30:00 +0100] "GET /some/page"
    [27/Oct/2013:01:59:00 +0100] "GET /another/page"
    [27/Oct/2013:01:30:00 +0000] "GET /some/page"
    

    View Slide

21. HTTP Logs
    [27/Oct/2013:01:30:00 +0100] "GET /some/page"
    [27/Oct/2013:01:59:00 +0100] "GET /another/page"
    [27/Oct/2013:01:30:00 +0000] "GET /some/page"
    [27/Oct/2013:00:30:00 +0000] "GET /some/page"
    [27/Oct/2013:00:59:00 +0000] "GET /another/page"
    [27/Oct/2013:01:30:00 +0000] "GET /some/page"
    

    View Slide

22. Changes to DST
    

    View Slide

23. 2011: Russia
    abandons
    DST
    

    View Slide

24. Timezone hopping
    

    View Slide

25. 2011: Samoa
    skips a day
    

    View Slide

26. Offsets are not timezones
    Tip:
    

    View Slide

27. Offset: +0000
    Zone: "Europe/London"
    

    View Slide

28. SQL doesn’t
    get this
    

    View Slide

29. So how do we
    get it right?
    

    View Slide

30. UTC everywhere
    

    View Slide

31. Localise late
    

    View Slide

32. Use the
    libraries
    

    View Slide

33. Trust IANA
    data
    

    View Slide

34. Some fun
    

    View Slide

35. Days that
    aren’t
    

    View Slide

36. 1752: England
    11 days gone
    

    View Slide

37. 1752: England
    A 9 month year
    

    View Slide

38. Leap seconds
    

    View Slide

39. 86,400
    seconds in a day
    

    View Slide

40. 86,400 86,401
    seconds in a day
    

    View Slide

41. 2012: Everywhere
    The end of
    the world
    

    View Slide

42. 2012: Everywhere
    The end of
    the world
    

    View Slide

43. 2012: Everywhere
    Leap-second
    kernel bug
    

    View Slide

44. 2012: Everywhere
    Leap-second
    kernel bug
    

    View Slide

45. 2012: Everywhere but
    Google
    Leap-second
    kernel bug
    

    View Slide

46. But I digress…
    (Where were we?)
    

    View Slide

47. The main things
    UTC rocks
    IANA tzdata rocks
    You need to think
    

    View Slide

48. Questions?
    @ChrisSinjo
    

    View Slide