Feeds

back to article Boffins debate killing leap seconds to help sysadmins

Boffins from the International Telecommunications Union (ITU) and the International Bureau of Weights and Measures ( known as the BIPM for its French name: Bureau International des Poids et Mesures) have met to discuss redefining time. Specifically, they're keen to redefine Coordinated Universal Time (UTC), the international …

COMMENTS

This topic is closed for new posts.

Page:

Silver badge

But… but… That would mean that at the end of my life, UTC time will be a whole minute out of sync with the rotation of the earth!?

4
0
Anonymous Coward

Terrible I say...

What I find funny though is why do we need leap seconds? doesn't that mean when the second was defined they made a mistake and the second is a TEENY bit too slow?

Surely we can easily predict when we need leap seconds, couldn't a non-connected machine be programmed to insert them at the right time?

I'm sure all the 'cloudy' guys will be having chest pains at the idea of something not being connected up!

2
9

Make them predictable

I agree with this thinking. I.e. instead of them deciding each time they need one, if they simply agreed that they would add some at predefined times. (Say 30th June in 2013, 2018, 2025, 2029 etc) to keep time with the predetermined slow down of the earths rotation, then we'd just have a fixed list that lasted 50 years or so. Then we'd all be dead and the children can think about how/ when to add more to that list.

Easy to implement (simply publish them, and then stick to them). Easy to embed in closed systems (just used published seconds - no system is likely to last more than 50 years with no maintenance).

And they could simply decide not to add any more leap seconds at all. That way there is no need to change existing systems (they just read the existing leap seconds from the existing list) and you don't need to change any embedded systems either. The GPS and 'updated' clocks are simply defined as being 18 seconds apart from each other from now on.

Everybody is happy.

0
4
Bronze badge
Happy

What a truly advanced civilization would do

Of course, the proper fix would be to adjust the rotation of the Earth to stay in sync with the atomic clocks!

18
0
Silver badge

> Surely we can easily predict when we need leap seconds, couldn't a non-connected machine be programmed to insert them at the right time?

You wish!

Slowdown however depends on many random effects - tides, air pressure, random displacements of masses within the earth's interior, continental drift etc. So it's not particularly deterministic.

10
0

"What I find funny though is why do we need leap seconds? doesn't that mean when the second was defined they made a mistake and the second is a TEENY bit too slow?"

No.

http://en.wikipedia.org/wiki/Leap_second#Slowing_down_of_the_Earth

#themoreyouknow

3
0
Bronze badge

No, the Earth's rotational period is gradually increasing as angular momentum is transferred to the Moon. However, it doesn't slow down at a constant rate and events such as earthquakes can cause the planet to spin faster.

3
0
JLH

Re: What a truly advanced civilization would do

"Of course, the proper fix would be to adjust the rotation of the Earth to stay in sync with the atomic clocks!"

Easy-peasy. Take some of those surplus Russian atomic bombs the equator, up a high mountain. Maybe Mount Kilimanjaro? Set them off.

Job's a good 'un.

See Project Orion for references.

1
0

Re: What a truly advanced civilization would do

"Of course, the proper fix would be to adjust the rotation of the Earth to stay in sync with the atomic clocks!"

Mmmmyessssss .... Going to be a bit difficult getting the moon to agree to that, isn't it? That sort of energy would require burning everything from fossil fuels to orphans.

0
0
Anonymous Coward

Re: What a truly advanced civilization would do

Well, we can always make more orphans. Not so with the fossil fuels though...

0
0
Bronze badge

I seem to recall that the rotation of the Earth is not quite uniform, perhaps due to sloshing of oceans and the Moon's orbit not being quite a perfect circle. If so, and if that is not fully predictable, it would seem the uniform clock also might need occasional resetting. I'm OK with NTP and a leap second now and then, along with occasional resetting of my pendulum clocks.

0
0
Bronze badge

Re: What a truly advanced civilization would do

That sort of energy would require burning everything from fossil fuels to orphans.

Yes, currently, but I was talking about a advanced civilization. We are not yet even at level I on the Kardashev scale.

0
0
Bronze badge
Pint

What we need . . .

. . . is a leap Monday so that the day starts about 11am

9
0

Re: What we need . . .

. . . is a leap Monday so that the day starts about 11am

...on a Friday. FTFY.

3
0
Silver badge

How about defining a new time zone just for

antediluvian sysops?

5
0
Bronze badge

But don't we all already have a continuous time parameter: The Unix count of number of seconds since 1st Jan 1970?

1
0
Bronze badge

It isn't

GPS time and other atomic clock systems (TAI) are continuous but Unix time isn't. To convert from an atomic clock to Unix UTC, you must shift the epoch and run through the leap second list.

The difference between UTC and Unix time is that UTC will add leap seconds by ending a day at after 23:59:60 than after the usual 23:59:59, thereby maintaining perfectly unique timestamps. Unix time can't do that so it repeats a second.

2
0

Unix time

@ A Non e-mouse: Unix time is the count of non-leap seconds since 1st Jan 1970. That is to say, it is not continuous. Strictly speaking it pauses for the duration of a leap second, practically of course on almost all systems it ploughs on regardless and gets corrected afterwards by NTP.

That said, we do have a continuous time parameter in common use in consumer devices. GPS time is the count of seconds since 6th January 1980.

The "big" difference that this change would make is that devices with GPS but no internet will be able to use GPS time to compute the "user friendly" time. They can't do that at the moment because leap seconds aren't predictable, so without an internet connection (or some other means of update) you can't tell how many have happened since 1980 (currently 16).

I don't own any clocks more accurate than 1 second in 2 years, and I think for the average person GPS is their best chance of having one. That said, IMO 2013 is a bit of a weird year to start worrying about non-connected devices. In terms of capability this measure seems to cover the very narrow gap that starts when GPS becomes easy, and ends when connecting to the internet becomes easy. In terms of cost, I don't know how much leap seconds do cost, but for most uses their effect is negligible compared with the inaccuracy of your on-board time-keeping device. Removing them from the equation doesn't really help because you still need *some* external source of accurate time.

1
0

Re: Unix time

Except the GPS system already broadcasts the offset between GPS time and UTC time (i.e. the leap second count since the GPS epoch)...

0
0
Bronze badge
Facepalm

If only there was a way for computers to send data to each other...

NIST and some other entities provide an online leap second list. There's no need for this to be manual.

4
0
Silver badge

Re: If only there was a way for computers to send data to each other...

So how do you propose to update the clock at the power station, you know, the one you where told to disconnect from the Internet last week.

0
0

Re: If only there was a way for computers to send data to each other...

I think you should go look up GPS clock PCI cards. You'd only need one on a closed network to keep everybody on that network in sync. More if you want security.

For the cost of a £200 PCI card its not worth the risk of changing how we currently work with clocks. We'll have 50 years of legacy data with leap seconds, which some future programmer will forget to cater for.

The risks of the system are being catered for, no need to rock the boat.

6
0

Re: If only there was a way for computers to send data to each other...

The main problem is that GPS satellites don't do leap seconds, so GPS time is already 16 seconds out from UTC. Also you cannot predict when you will need leap seconds as it depends upon the weather on top of the mountains, which affects the amount of ice that accumulates in winter which affects how fast the earth rotates.

1
2
Silver badge

Re: @dkjd

GPS itself uses atomic time starting 1980, so no leap seconds BUT, and this is where you are really wrong, the GPS ephemeris gives you the offset from GPS-UTC, so you can and do get leap-second information that way.

1
0
Silver badge
Boffin

Re: @John G Imrie

Closed network on a power station? Then buy at least two GPS and LW equipped time servers for redundancy.

For example:

http://www.galsys.co.uk/ntp-servers/nts-6001-dual-ntp-server.html

http://www.timetools.co.uk/ppc/ntp-time-server.htm

http://www.meinbergglobal.com/english/products/rail-mount-ntp-server.htm

http://www.spectratime.com/products/ireference/

Simplez!

1
1

Re: If only there was a way for computers to send data to each other...

> I think you should go look up GPS clock PCI cards. You'd only need one on a closed network to keep everybody on that network in sync.

Yes, in sync with a spoofable signal originating from outside the network.

0
0

To (very loosely) cite Franklin

Those who sacrifice precision to gain a little cost saving deserve neither.

...users and beancounters, BAH!

4
0
Silver badge
FAIL

Oh FFS!

So in order to deal with incompetent or poorly tested OS designs that don't actually bother to address the definition of time that has been around for several decades, they want to break compatibility with anything that actually uses that definition by assuming that Earth rotation is never more than +/-1 sec from UTC?

A triumph of the incompetent many :(

Why don't they just tell folk to fix their software, its not a new problem after all?

And for those devices that are not connected to "know" about leap seconds, how exactly would they be keeping accurate time in the first place, and even if they do, how would that matter if they don't interact with systems that are kept in sync?

11
0

surely?

My computer does not need to measure the rotation of the earth. It does however need to agree what the time is with other computers for security reasons. It might also have to tell humans to do stuff at particular times, but that need not be that precise.

0
0
Silver badge

Re: surely?

justincormack: "My computer does not need to measure the rotation of the earth. It does however need to agree what the time is with other computers for security reasons. It might also have to tell humans to do stuff at particular times, but that need not be that precise."

I'm not sure whether you are agreeing with the previous poster or not. But the key point is "agreeing with other computers". This implies a shared authoritative source of time: if not from the internet, then from GPS, and if not from GPS then one of the computers' relatively inaccurate on-board clocks will have to be considered the authorative time on that isolated network.

0
0
LDS
Silver badge

Every system I saw without a time reference to update its clock...

... usually has no really any idea about what time is. There is not only NTP, there are radio and GPS receivers to update time when needed. System employing none of the available systems usually have clocks out of sync of several minutes - can't really see how leap seconds can be an issue for them.

If we are talking about systems who have a time reference but can't handle leap seconds properly it's another matter.

3
0
Silver badge
WTF?

Bollocks presentation

Have you read the linked slide show? Three obvious political-style lies are included:

Page 6 - "Leap seconds interrupt normal operation of timekeeping infrastructures and are costly in staff time to implement" - no, you use NTP and it just happens! Unless the system gets broken due to bad/untested software, you need no interaction whatsoever.

Page 6 - "On June 30, 2012, every clock in the world had to stop for one second" no the fscking did not "stop", they simply stepped one second when needed. If you rely on a basic time-stamp then you might get it repeated, etc, but if monotonic time actually matters deeply for program flow or synchronisation, you use one of the system supplied functions that gives you that. (e.g. clock_gettime() with the CLOCK_MONOTONIC flag) or you implement your code to cope in other ways.

Page 12 - "and significant cost reduction in their implementation" - no, you use a system-supplied library that handles time correctly and then only one competent programmer needs to do it, and everyone else "just works". Having monkey-grade programmers implementing basic time keeping over and over again, and getting it wrong (by not RTFM) is a sign of a far deeper problem in your organisation and choice of staff.

How to we get this joker to correct this and apologies?

15
1
Windows

Re: Stopped clocks

Not only do the clocks not stop, but it's quite simple to make sure they don't repeat (or skip) seconds either. (As you say; thinking this through only needs doing once.)

Interestingly enough, the first I read about this was this post, which talks about how Windows deals with small adjustments coming through NTP (search for "System clock changes"):

http://blogs.msdn.com/b/rxteam/archive/2012/06/20/reactive-extensions-v2-0-release-candidate-available-now.aspx

I swear that it was only a couple of weeks later that there WAS a leap second introduced that caused a bunch of servers somewhere important to get their knickers in a twist, and Google announced that they had a patch in their version of Linux that made it do the same thing. I was surprised at the time that it didn't already.

Ah! Here it is:

http://www.theregister.co.uk/2012/07/02/leap_second_crashes_airlines/

http://googleblog.blogspot.co.uk/2011/09/time-technology-and-leaping-seconds.html

Looks like they adjusted their internal NTP servers. Interesting. That still doesn't describe how a standard linux system will deal with this then. I won't speculate, as I know next to nothing about these things, really.

0
0
Silver badge

Re: Stopped clocks

AFIK the leap second problem that affected Linux last year was down to some timers getting dead-locked, and that was due to a kernel patch that broke the previously correct time-handling for leap seconds. And nobody realised or tested it until the live event:

http://lkml.indiana.edu/hypermail/linux/kernel/1203.1/04598.html

A short check shows a RedHat article including a leap second simulator so you can test a system for its behaviour to debug this predictable event:

https://access.redhat.com/site/articles/199563

While a big event, it just shows the price you pay for not testing something for all expected conditions.

Google slew their machines over 1 day, so no step but also same long-term behaviour. Of course, during that day they are up to 1s out, but clearly that is no big deal for them.

0
0

Re: Stopped clocks

ntpd will make small adjustments by skewing the wall time clock and large adjustments by stepping it.

Following a leap second the system learns from NTP that its clock is suddenly a second fast, then gradually brings it back into synch with the external time. Instead of stopping for a second, or repeating a second, it will in effect stop/repeat a few milliseconds at a time for a while.

I suppose it's possible that this behaviour can be changed, and/or is different by default on some distros. Since I don't actually care I've never verified. I have much worse problems than leap seconds keeping accurate time on the linux boxes I manage, the principle one being VMs whose clock is tied to the host, and whose host clock is incorrect.

0
0
Silver badge

Re: Stopped clocks

In the Linux case it knows (from NTP) that a step is pending, and it jumps accordingly. The ntpd slewing/stepping is for "normal" time errors.

How VMs handle this is another story. From our experience VM timekeeping is pants anyway, so this is just another minor issue. If an OS/application really needs good timekeeping for some task (e.g. audit of network delays for security such as MITM detection), it really has to run on a physical machine.

0
0

Re: Bollocks presentation

> " no, you use a system-supplied library that handles time correctly"

>

Out of curiosity: which system would that be? Surely neither Windows nor Unixes! I hear IBM mainframes run on TAI only, is that what you refer to?

0
0

Re: Bollocks presentation

The main problem with leap seconds is that it becomes possible to have two computer events logged out of sequence. Computers are fast enough that many events can happen during one second - so consider the following:

23:59:59.75 = Event 1

00:00:00.00 = Event 2

00:00:00.25 = Event 3

00:00:00.50 = Event 4

00:00:00.75 = Event 5

00:00:01.00 = Event 6

Now imagine that there is a leap second - in that case the time stamp of event 6 will not be 00:00:01:00 but the leap second puts it back to 00:00:00.00 - which would log Event 6 as occurring before events 3,4 and 5

IOW if an event happens to be logged at 00:00:00.xxx on a leap second day, the software will not be able to determine whether that was the first 00:00:00.xxx or the second (leap) one.

As the article states, you need to represent the leap second with a unique time stamp such as 00:00:60.xxx, which would require a heck of a lot of software to be updated to handle such a time.

0
0

Re: Bollocks presentation

Robert Nelson died early this year, so he won't be correcting the presentation.

0
0
Bronze badge

Re: Bollocks presentation

That's only if your date and time handling code doesn't account for leap seconds. The point of leap seconds is that the system goes 58, 59, 60, 00, 01, etc. instead of 58, 59, 00, 01 as normal.

0
0
Silver badge

The UK's position

Thankfully, it seems the UK's position is sensible, as covered here:

http://www.itu.int/dms_pub/itu-r/oth/0a/0e/R0A0E0000960014PDFE.pdf

Basically they point out that not only would it mean that "1 day" in no longer synchronised to the Earth's rotation as common sense expects, but that you either end up with a long term problem of sun rise/set getting seriously out of sync with our working hours, or you have bigger but less frequent steps which are worse then a 1-2 year leap second in terms of impacting badly designed systems.

Really, why don't they just make proper time-keeping a mandatory requirement for software systems and force vendors to test and demonstrate they can handle it? That is the biggest issues here: most folk don't have (or will pony up for) an NTP simulator to allow them to set up and test the OS/application reaction to these predictable and recurring events, so they simply hope for the best and, surprise, surprise, they get the worst!

9
0

Re: The UK's position

One of the problems is that, over decades, leap seconds will need to be added more often as the earth is slowly slowing, and the standard second is actually derived from the earth rotation rate sometime in the 19th century. However I think it is stupid to address this problem now. We can leave it to the boffins a century or so hence. Meantme there is Y2038 to get through...

1
1
Bronze badge

Re: The UK's position

If the second was still derived from the earth's rotation, we'd be ok! But it's now derived from the speed of light.

0
0
Silver badge

Seconds? Give me hours!

“If leap seconds are eliminated from UTC, there will be no perceptible impact on social activities and conventions,”

Just as the Julian Calendar was just fine ... until it wasn't. I personally would favor omitting the switch to summer time ... to get one hour of extra sleep every year.

0
1

If we drop the leap seconds then astronomical charts will slowly move out of sync as they are all predicated on a noon being noon. That means that the final backup for sea navigation will no longer be accurate.

1
0
Anonymous Coward

Yea

So you need GPS primarily to keep accurate time for navigating by astronomical chart and sextant. Makes sense!

(insert joke icon here).

0
0
Bronze badge

Oh Shit !

Is that the time ? Sorry, I have to go......

2
1

What is the fuss about?

"Sysadmins are among the groups inconvenienced by leap seconds, as while network time protocol (NTP) is aware of them and includes them in its updates, not every device is connected to an NTP server."

If hardware has neither a network time protocol nor an accurate time source, you'll have more than 1 second a year to worry about anyway. If you do have an accurate time source, then it'll be acting as your time server and you'll only have one box to update.

There are very few applications where you have many disconnected nodes which all have their own accurate clocks.

There is the small issue that applications can behave badly if the real time clock is modified underneath them. Normally NTP clients slowly adjust the RTC for any drift, which negates this issues, and even if it is an issue then the application is badly written and should be using the system clock for measuring the passage of time, instead of the RTC.

0
0
FAIL

Keep them

I'm all for keeping leap seconds, and doing so until a significant fraction of the population aren't living on this planet. No matter how you define time, most people prefer to synchronise their day by the rising and setting of the sun.

4
0
Silver badge

What a legacy we will leave

if we remove leap-seconds. OK: it probably will not have any real impact on my kids or grand kids, but as the centuries roll on the clocks will ever get more out of sync with that yellow thing upstairs. When they do come to fix it the problem will be huge: several hours to shift and computer systems which are based on the idea that there are always 86400 seconds in every day without exception.

I am sorry that the real world is more complicated than some would wish - but that is how it is.

Better to get used to it now than have our great, great, ... grandkids curse us for idleness.

5
0

Page:

This topic is closed for new posts.