[LEAPSECS] No leapseconds on trains

[Rob Seaman]

On Nov 17, 2011, at 9:57 AM, Warner Losh wrote:


> It is a problem with the current system.


No, it is a feature of the current solution.  The problem is civil timekeeping.  The current solution is Coordinated Universal TIme, that is, mean solar time (and the details we're all familiar with).  A feature is leap seconds.  That feature has positive and negative attributes for different communities.


> No solution is offered to this problem in this context.


I woke up and PHK was on another hobby-horse.  If the Danish railways want to hire him (or me, for that matter :-) to work on mitigation (assuming any is needed), I'm sure he'll do a competent and professional job for them.  Introducing the hobby-horse as a project management scandal doesn't provide much confidence, however :-)


> It is a problem that nobody can deny exists.


Like I said, in an engineering sense it is a feature that reveals various issues with diverse systems.  I don't deny that some issues exist, but they don't exist independently of the systems in question.  That civil timekeeping is derived from mean solar time does have an independent existence.  The inherent engineering requirement to keep civil time synchronized with time-of-day will automatically assert its existence no matter what the ITU decides.  There is plenty of flexibility in these facts to pursue different solutions with different features.  A timescale equivalent to TAI does not solve the problem on the table, however  - without some additional features.  The ITU proposal is bad engineering.  The ITU process is a bad process.


>> call a leap-less timescale "TI" as was decided in 2003.

> 

> Was that official?  No.  Do people actually know about it? No.  Heck, I knew that was a recommendation, but didn't know it was formally and finally decided.  :)


It is more official than any other action that has been taken in the intervening 8 years.


> No, they've come up with a different solution: ignore the problem and someone/something else will reset the clocks if there's an error.  That's the predominate solution that exists in the marketplace today: leap seconds are hard/obscure/not worth the cost: f* em.  If time gets out of sync, the magic time fairies will set things right.


As we've discussed here innumerable times, all clocks must expect resets to occur from time to time.  During those resets leap seconds are automatically accommodated for many systems.  This is not magic.  And whether ignoring features of a standard is good engineering is not something that can be decided in a blanket fashion.  And whether or not fairies are involved, the reality is that the marketplace is currently layered on mean solar time.  Asserting that changing UTC to implement something other than mean solar will have no deleterious effect does not make it so.  Due diligence has not been served.


>> And of course the ITU process has spurned the participation of railroads as it has everybody else.

> 

> Perhaps that indicates that the ITU process isn't serving the needs of the people its standards are supposed to support.


Perhaps it does.

Rob