[LEAPSECS] leapseconds on trains

Rob Seaman seaman at noao.edu
Thu Nov 17 15:00:49 EST 2011

On Nov 17, 2011, at 11:57 AM, Ask Bjørn Hansen wrote:

> I've been following the list for a while and seen many examples of communities and significant industries where leap seconds are causing problems

But you've seen no investigations of the significance of the problems, and only anecdotal evidence that they indeed actually exist.

> (and just stopping them with cause no new friction).

And no evidence whatsoever that redefining UTC will not cause even bigger problems for these same communities and industries. Nobody has looked. No Y2K-like inventories have been performed. No tests have been carried out.

> Count me in with the uninformed then and please help inform us.

See presentations at:


The proceedings are just being finished up and should be available in the next week or so.

> (Just to be clear here; I'm completely serious - I understand the general concept of course, but I genuinely don't understand the specific use cases for having leap seconds).

System engineering sermon again. Use cases are tools for discovering engineering requirements. They operate in the problem space. Leap seconds are a feature of UTC. UTC is the current solution to the problem of civil timekeeping (or one aspect of the larger solution). The phrase "use cases for having leap seconds" is not a coherent sentence in systems engineering since characterizing the problem and proposing and evaluating candidate solutions are completely different parts of the process.

An engineering requirement for civil timekeeping might be phrased something like: (A) "human civilization is dependent on diurnal phenomena". (This is phrased in a rather schematic way to simplify the discussion.) The use cases that led to discovering this are many and detailed. They would include lots and lots of diurnal phenomena and more importantly the way human society depends on them. Asserting that requirement (A) is not actually a requirement is equivalent to finding ways to undermine the big long list of use cases, i.e., that trans-Atlantic flights don't follow a daily schedule, that electrical loads don't exhibit one characteristic 24-hour response in the Summer and another in the Winter, that rush-hour traffic doesn't rise and fall one way on week days and another on weekends, etc.

Requirement (A) and other requirements have led to mean solar time being adopted and maintained as a standard over several centuries. Mean solar time remains stationary with respect to time-of-day and thus satisfies the diurnal requirement. Leap seconds have been the price to pay to implement mean solar time over the past few decades. Other mechanisms are possible - we know this because mean solar time was successfully implemented before leap seconds.

> Other than telescopes and some related equipment, I have yet to see an example of industries or communities that'll have problems with no new leap seconds being introduced.

Nobody has looked. The aerospace community seems quite vociferous. Not sure why the blatantly obvious issues this would cause immediately for astronomers are not taken as a warning - as a canary in the coal mine. It will be an interesting experience to fly in a few years as the vast number of systems internal to the aircraft interact with the vast number of systems on the ground. Will they all understand that time no longer means time-of-day? Will they implement corrections in the same direction? Will Airbus and Boeing behave the same way?

> In my experience the vast panoply of systems is coordinated based on NTP or other similar equipment. Most of which does a terrible job handling leap seconds.

NTP distributes UTC. That it does not do it perfectly is not the issue.

> Making UTC not have leap seconds would make it more "Universal" in that the number of systems telling the same time would go up.

No. "Universal Time" is an atomic concept meaning "mean solar time". One supposes it might make it more "coordinated", but as with everything else nobody has actually investigated the details of how well NTP works now versus how well it would work in the future.

> Can you (or another participant) give me some concrete examples of stuff that needs leap seconds and don't already have well established mechanisms for adjusting the time output from their GPS or other time keeping equipment appropriately?

See http://www.cacr.caltech.edu/futureofutc/program/presentations/AAS_11-674_Storz.pptx.pdf

Even very very high tech equipment has been implemented under the quite reasonable assumption that "Coordinated Universal Time" is a kind of "Universal Time", meaning (as it always has) mean solar time.


More information about the LEAPSECS mailing list