Adding a leap second this year is expected to cause Internet woe

[HEXUS]

IER announced that it is adding a second to the clock on June 30, 2015.

Read more.

[ZaO]

The linked article says "Nowadays, however, adding a leap second—which is scheduled to occur at 23:59:59 on June 30 (the halfway point of the year)—can cause computer systems to become confused when their clock shows 60 seconds, rather than rolling over after 59—and we all know what that generally means—outages. Other computers will show the 59th second for two seconds in a row, which can also cause problems." Instead, why not just set the clock back by 1 second once we're a a couple seconds into July? You know, like we do with daylight savings when we're an hour in or whatever? Websites seem to deal with that ok don't they? I don't get why this is a big deal...

[kingpotnoodle]

The linked article says "Nowadays, however, adding a leap second—which is scheduled to occur at 23:59:59 on June 30 (the halfway point of the year)—can cause computer systems to become confused when their clock shows 60 seconds, rather than rolling over after 59—and we all know what that generally means—outages. Other computers will show the 59th second for two seconds in a row, which can also cause problems." Instead, why not just set the clock back by 1 second once we're a a couple seconds into July? You know, like we do with daylight savings when we're an hour in or whatever? Websites seem to deal with that ok don't they? I don't get why this is a big deal...

Setting the clock back a second is what causes it to show 59 for 2s, which to some software might be confusing, say an action requires something to happen for 10s between 50-59 secs, stopping when it flips to 00, when there is a leap second that task runs for 11 seconds.

[TheAnimus]

There are two kinds of problems, the COBOL style problem, and the duration problem.

The COBOL one, only effects really dumb programmers, so brace yourselves if you've got JS or PHP. Here someone treats a date, not as a number, but as a series of symbols (called a 'string' as they are symbols strung together). They then trigger something happening when the last 2 symbols equal something else. This is bad. Those responsible should be shot and their children shot.

You have the duration one, which is a bit more tricky, most systems measure time in the number of seconds (or fractions of a second) elapsed since a specific point in time (called an epoch). This is just a big number, computers like big numbers, and this of course doesn't change, as it's still 100 seconds ellapsed, regardless if you show 59 or 60, because the number ellapsed is obviously true.

However, sometimes people will calculate a duration, durations mean these seconds ellapsed have to be mapped to the time as we know it. If ever you needed any kind of proof that god isn't real, work with date and time, so one person might have the leap second in their lib, someone else, not. Will one second matter in this case? For most things, no. For things that really care about one second, they use time since an epoch anyway.

[DanceswithUnix]

Setting the clock back a second is what causes it to show 59 for 2s, which to some software might be confusing, say an action requires something to happen for 10s between 50-59 secs, stopping when it flips to 00, when there is a leap second that task runs for 11 seconds.

Timing a duration you would use a monotonic clock not wall clock time, or at least you should. And there lies the problem, summer time hits us twice a year, leap seconds are not common enough to be well tested against.

I would have thought that most modern code runs on virtual machines where the clocks are so wonky that an extra second turning up would be casually accepted.

[Jace007]

There will be an implosion, reality will cease to exist and be destroyed by the software erratum of 0 !!!

Listening to Bob marley until the End of...

[edzieba]

Handling time is hard.

[peterb]

It isn't the first time a leap second has been added, neither will it be the last. For the majority of systems it won't make any difference, and for those systems that synchronise over the net it won't matter at all, systems will sync real time plus whatever latency is there anyway. Those that synchronise over GPS will not be affected.

Where it would matter is in systems like GPS satellites, and that will be covered anyway, as will organisations that operate time standard clocks.

Personally, I shall enjoy my extra second of sleep!

[shaithis]

We survived Y2K, I think we will survive this.

[ZaO]

It isn't the first time a leap second has been added, neither will it be the last. For the majority of systems it won't make any difference, and for those systems that synchronise over the net it won't matter at all, systems will sync real time plus whatever latency is there anyway. Those that synchronise over GPS will not be affected.

Where it would matter is in systems like GPS satellites, and that will be covered anyway, as will organisations that operate time standard clocks.

Personally, I shall enjoy my extra second of sleep!

Aren't satellites experiencing time differently anyway? Or are they not quite far enough out for that?

Some interesting posts here. But I bet the majority of us won't notice anything at all. My motherboard seems to keep falling behind with the time, then catching back up when it syncs with ms servers. It's nothing more than annoying. It's not messing with any programs as far as I can tell. But I do now understand that it could be messing with some computers/programs. I suppose it might cause some big problems for computers running time precise experiments or something..