Meet the Guy Whose Software Keeps the World's Digital Clocks In Sync (ieee.org) 78
New submitter Wave723 quotes a story on IEEE: In many cases, the internal clock that ticks away in a laptop or desktop computer is synchronized to an official time service maintained by the U.S. National Institute of Standards and Technology (NIST). This free service shares Coordinated Universal Time with personal devices, web browsers, financial trading software and e-mail programs throughout the world. The service receives 150,000 requests per second (roughly 16 billion a day) from systems that repeatedly ask, 'What time is it?' "If you have a PC, it's probably synchronized to the time service," says Judah Levine, the man who originally built servers and programmed software to send time over the Internet for NIST back in 1993.
Re: (Score:2)
A leap second is really no different from a leap day or the "extra" or "lost" hour due to daylight saving time.
I don't know why people have such a problem with it.
Count your ticks, determine how to show the date and time for a given locale.
Re: (Score:1)
Because the leap second like DST are _artificial_ and adds unnecessary complexity to an otherwise brain-dead simple algorithm. Complexity == Bugs.
The only one who gives a fuck about leap seconds are scientists.
Stop over-engineering a simple concept. Time should be monotonic, and consistent. Not this one-off shenanigans.
Re:Nut in charge of the nut house. (Score:4, Insightful)
The biggest source of problems is POSIX, which some design-by-committee decided should define a day as having a fixed length, ignoring the existence leap seconds.
If you don't care about second accurate time, you don't have to deal with leap seconds, and your complexity problem is solved.
Re: (Score:1)
time zones are artificial, they reflect political borders. And nothing is more artificial than politics.
Re:Nut in charge of the nut house. (Score:5, Insightful)
Re: (Score:2)
Time itself is an illusion. Lunchtime doubly so.
I think I saw that in Reader's Digest.
Re: (Score:1)
That's from the Hitchhiker's Guide to the Galaxy. Hand in your geek card.
Re:Nut in charge of the nut house. (Score:5, Interesting)
The problem is there is a lot of time.
You may remember GMT, nowadays known as UT (not to be confused with UTC). GMT is time based on the Earth's rotation - when the sun is directly overhead, it's noon. There are approximately 86,400 seconds in a day here, but the mean solar day varies because Earth's rotation is not uniform. UT1 is the most common form of this, which is what time is measured at the prime meridian.
UTC is time derived from the atomic clock. It closely approximates UT, but since Earth's rotation is erratic, to keep the UTC day closely aligned with the UT day, leap seconds are sometimes added to ensure the difference between the two times is under 0.9 seconds.
TAI is the time as told by atomic clocks. Here, a day is exactly 86,400 seconds and there is no such thing as leap seconds - this is purely a monotonic clock that ticks away.
The problem is, well, there are a lot of variables. UT is measured generally once a day and clocks set to its time. UTC is a close approximation and generally used as it's easier to obtain without having to have someone observe the Sun every day to calculate when noon is. TAI is just the atomic clock time.
Leap seconds are introduced to keep UTC and UT relatively close to each other. TAI is allowed to drift, and eventually you'll have noon at midnight.
Which you pick is up to your needs. Leap years were created so people in the Northern hemisphere wouldn't be celebrating summer in December as the calendar drifts away from Earth's position in its orbit..
Then there's TAI, which is the true atomic time
Re: (Score:2)
> Leap years were created so people in the Northern hemisphere wouldn't be celebrating summer in December as the calendar drifts away from Earth's position in its orbit..
First World Problems. Someone call the wambulance.
Just use TAI where no fucking hacks are needed, and get rid of this retarded leap second, leap year, and DST.
Re: (Score:1)
Long before we start having winter in July on the northern hemisphere we will start to have to shift store opening hours since 12:00 isn't noon anymore. Given that we want business hours to coincide with daylight this would be a much bigger problem long before we have snow in July. And who cares what season is in what month. That is totally irrelevant. There are places, like Aruba, where there are no discernable seasons. Do they only have June because of this?
TAI is good for scientific purposes, but leap da
Re: (Score:2)
The universe has as much need to coordinate Earth's rotation with the flatulence pulses of the bombardier beetle, as it has to coordinate $ThisLunarSolarEclipse$ with $ThatEuropaIoMutualEclipse$. I.e. none at all.
This sort of adjustment is only an issue if you do it very rarely. If you deal with events more than a coupple of hours different from your local sun ri
Re: (Score:2)
Actually, the leap day is quite different than daylight savings time.
Usually, February has 28 days, except when it has 29 days. This is like a leap second, where once in a great while, a minute has 61 seconds.
daylight savings time doesn't inject any new time, it just retreads old time/skips time.
Practically speaking, a leap second would have been easier to model like daylight savings time (repeat one of the 60 seconds usually found in a minute), except that would mean some portion of indicated time would b
Re: (Score:2)
It doesn't even do that, it just switches between timescales. 3:59 EST = 4:49 EDT. The only thing the government mandated change does is define when the switch occurs. Informally, when someone states the time or sets their clock, they don't consider the xST/xDT part, so the change forward/backward appears to be a 1 hour change. DST doesn't make time discontinuous nor make in non-monotonic.
Re: (Score:2)
Except I can accurately predict centuries in advance when February 29 occurs.
Leap seconds are added whenever measurements cause a committee at the UN non-governmental organization that oversees the standard (International Earth Rotation and Reference Systems Service) to schedule for a leap second to be inserted.
Re:Nut in charge of the nut house. (Score:4, Interesting)
Unfortunately, Unix and NTP treat leap seconds completely different from Daylight Saving. They actually mess with the real measured clock, slowing it down. This is just completely broken of course, suddenly seconds aren't one second long any more. This has lead to system crashes and all sorts of fun.
There are two solutions to this: Either the NTP guys get their heads out of their asses, or leap seconds are abolished.
Unix is mostly ready, there can be 62 seconds in a minute in C, so all you do is switch the timezone files to the "right" directory instead of the "posix" directory (on GNU systems at least). But that cannot be done until NTP works.
Re:Nut in charge of the nut house. (Score:4, Informative)
But yes, ntpd (the reference implementation) is very broken - it doesn't even follow its own RFC with regard to enumerating time. Of course, anyone who is inclined to produce a correct implementation will bump into the fact that the reference implementation is spewing incorrect timestamps everywhere, and systems expect that. It was developed to keep POSIX time, so just like POSIX, doesn't deal well with leap seconds or UTC.
Re: (Score:1)
Huh?
The Impact of Leap Seconds on Digital Time Services
Judah Levine
Time and Frequency Division
NIST, Boulder, Colorado
paper [itu.int]
Re: (Score:2)
Re: (Score:1)
The explanations in that file make it clear that the timestamps are given as NTP timestamps. Unix time and NTP do not count leap seconds. In Unix time and NTP, there are always exactly the same number of seconds from 23:59:30 to 00:00:30, whether a leap second was inserted between those times or not. The extra second is dealt with by stopping the clock for a second or running it slower so that it's back in sync with UTC after a few minutes. It would be a colossal misstatement to say that NTP timestamps repr
receives 150,000 requests per second (Score:1)
Not from me it doesn't. I only ask maybe once a month.
Re: (Score:2)
I'm guessing they meant NTP traffic is estimated to be 150,000 requests per second, globally.
All my machines sync to their domain controllers. The domain controllers sync to a local time server. The time server syncs to the big boy government time servers. They never see the flood of requests from my machines or the other machines at my location. They only see requests from our local time server.
Re: (Score:1)
I'm guessing I need to put in a laugh track. I usually turn off the automatic updates, including the time. *Paranoia strikes deep* You'd be amazed at how sloppy these crystals in our machines are at keeping time. 60 cycle hum is more accurate.
Re: (Score:1)
It may not be that way forever. [msn.com]
Re: (Score:2)
I don't know about America, but over here I have seen the frequency vary from 49Hz to 52Hz as the grid gears up for the 6am winter load(everyone gets up and starts cooking/boiling a kettle etc). On average it is still probably accurate. I am fairly sure the US power companies do the same thing. It is difficult to control the frequency under varying load.
Re: (Score:2)
Back in the Windows 9x days I remember adding a realmode program to autoexec that would compensate for natural drift in the clock. Sync the clock to atomic, sync it again in a week when it was way off, and then it would calculate the error and continuously apply it to the RTC. It worked surprisingly well.
To bad a similar algorithm is not still used.
Re: (Score:2, Informative)
That's essentially what NTP does. It runs in the background as a daemon and disciplines the local clock to keep it in sync with the upstream servers. It won't jump the time unless you tell it to, choosing instead to slow/speed the passage of time to stay in sync.
IIRC, the amount needed to discipline the local clock is stored in the drift file.
This all works very well as long as your clock crystal is mostly rational and doesn't jitter too much based on temperature variation.
Re: (Score:2)
The only tinhg I'd consider adding would be a level-0-or-1 source, such as a GPS/ GLONASS time source (you can argue over whether that is level-zero, or level-one, validly), and some glue so that if your time system and the alternative system differ by more than $SENSITIVITY$, your sysadmins get an ... orange-not-red flag and are aware of the issue.
If you're doing very time sensitive work, that's a different issue. A large bucket of issues.
i'd like to keep the World in Sync (Score:4, Funny)
Re:i'd like to keep the World in Sync (Score:4, Funny)
you'd like to buy the world a clock
and always wind the key?
I'd like to have you come on time (Score:2)
and work a whole damn day
if you don't get your butt in gear
you won't get no more pay!
bzzzzt (Score:5, Interesting)
TFA:
First of all, that's aspirational (or was) in most of the other articles I found.
Contra TFA: NIST Launches NIST-F2 [nist.gov]
Unfortunately, even contra TFA is weak geek tea:
I guess there's a reason why people with tiny UIDs memorize pi to a silly number of places: it helps you not leave off the other five or six significant digits in the rare case where it actually matters. The real frequency standard is only, like, approximately a million times better than that long-assed, dock-tailed string of digits visually implies.
Truly inconceivable—almost—and yet barely able to time slice the total perspective vortex.
Finally, some obligatory geek porn: Atomic fountain [wikipedia.org]
Re: (Score:2)
Pfft... Real geeks know (and understand!) the Bailey, Borwein, and Plouffe Formula and a dozen ways to implement it in C.
Time is a social construct (Score:3, Funny)
I no longer believe in time. I think it's pretty much junk science, and the daylight savings time thing is just an Illuminati plot to keep us subservient to the elite.
You all can do what you want, and spring forward or what not if you need to bend your will to The Man, but I ain't changed my clocks since 2007 and haven't noticed one thing. In fact, I couldn't change them since I threw out my wristwatch, Easy Rider-style, in 2006. Right now, if I look down at the time display on my screen, it's flashing 00:00:00, just like my DVD player and microwave.
Re: (Score:2)
Well, when you're old like us, you just say whatever you damned well feel like saying and damn the consequences or opinions.
Me? I think most people's feet are ugly and there's not a damned thing they can do about it.
Re: (Score:2)
That's not true. I used to have some great shoes that made my feet look great. Of course that was back in the day when they really knew how to make good shoes. Nowadays even Doc Martins have gone downhill and pretty much all shoes are glued rather than stitched.
Its the same with hats. One size fits all? Bollocks more like.
Re: (Score:2)
Honey, is that you?
Re: (Score:2)
I no longer believe you take yourself seriously or ever form coherent sentences at all, instead preferring to go through life just rambling whatever incoherent jumble of words produces itself from the mass of diseased tissue you call a brain.
I initially read this as "deceased". It had the same effect.
Re: (Score:2)
You forgot to mention how we were all educated stupid.
Re: (Score:2)
nist? (Score:2)
Road Trip... (Score:5, Funny)
"What time is it?" 6 Billions times a day...
Sounds like a long road trip with my kids... Are we there yet, How much farther, When will we be there, What time is it
Trading Systems? (Score:2)
I thought trading system typically relied on GPS based NTP servers in their own network?
Oh and if you want to make your one: http://www.satsignal.eu/ntp/Ra... [satsignal.eu] . While probably not as accurate as a commercial version, it is a tad slight cheaper.
Re: (Score:1)
You'd still need a reference time keeper to ensure there's no drift in your (or even their) systems. Everyone synchronises with the official timekeeper.
Note: If there was drift in every other system in the world except yours then that will have the inverse effect on you. In fact the earth's slow degradation as a time keeper is why we have leap seconds.
Does anybody really know what time it is? (Score:1)
Does anybody really care?
Re: (Score:2)
Wow, somebody with mod points really isn't a fan of Chicago!
Horrible mostly wrong article title (Score:5, Informative)
Judah Levine, the gentleman mentioned in the article, built interfaces to existing atomic clocks that allowed other clocks to synchronize with them, which is a worthy achievement.
But today, the vast majority of synchronized clocks are being kept synced by NTP across the Internet, not by radio signals. And although Levine also implemented NTP interfaces at NIST, he didn't invent NTP nor was he responsible for its dominance of Internet timekeeping.
The man who invented NTP and originally wrote the implementation was David L. Mills of the University of Delaware.
Mills is also the man who created the Fuzzballs and EGP, making global-scale internetworking possible.
Re:Horrible mostly wrong article title (Score:4, Interesting)
The man who invented NTP and originally wrote the implementation was David L. Mills of the University of Delaware.
Mills is also the man who created the Fuzzballs and EGP, making global-scale internetworking possible.
I knew Dr. Dave when he was still at the University of Michigan, doing the Data Concentrator and the Language Lab's automation, and I was in high school and hanging around the campus.
Great guy.
Re: (Score:1)
Thank you. I deeply lament David Mills's absence from the original article. He should at least have been mentioned, as he deserves far more credit for innovation in distributing time than does Judah Levine, while still intending no disrespect for Dr. Levine.
Like "man" from Cheech and Chong (Score:1)
I'm not into time...