[LEAPSECS] Short notice for DST changes

[Rob Seaman]

On Feb 24, 2014, at 3:23 PM, Clive D.W. Feather <clive at davros.org> wrote:


> Rob Seaman said:

>> Chile's rules are familiar because many observatories are located in that timezone, but presumably the same shenanigans play out worldwide.  It is not obvious why a couple of weeks notice is acceptable for hour amplitude clock adjustments, but six months is not acceptable for leap seconds.

> 

> Leap seconds affect the base time that we're all working to. Summer time

> changes only affect the presentation to one area, and the locals there will

> know things get screwed up because of the short notice.


And different areas will get screwed up in different ways at different times and there will be no underlying solar time scale to appeal to.  We’re being told that the entire issue is about time intervals.  What about comparing time intervals between different timezones - or merely between differing epochs in one location that used to be UTC+8 and is now UTC+7.  (Quick!  Did I get the sign wrong?)  You’ll need to update and access a heck of a lot more peculiar database of worldwide timezones to compute intervals than a single monotonic leap second table.


>> elsewhere they are still talking about accumulating the embargoed leap seconds into leap minutes.

> 

> Who is?


Others with influence over the ITU.


> Who, apart from a few astronomers, actually needs a reference - as opposed

> to local - time that is roughly synchronized with the sun rather than one

> that is easy to work with?


I reject your multiple premises.  As in innumerable previous messages, “rough synchronization” is your talking point, not mine.  Apparent solar time, the local offset from the standard meridian, and political DST adjustments have nothing to do with anything.  Mean solar time is easy to work with.  And atomic clocks simply keep a different timescale, readily accessible via GPS.

It isn’t a question of “who needs” but rather “what is”.  Day means synodic day, on Earth as elsewhere.  It is inevitable that this fact will make itself known to the systems we deploy.  If you believe otherwise, then presumably you believe that we could call 25 hours of 3600 SI-seconds a “day”?  Why is it not being proposed to rationalize time by getting rid of those wacky 86,400 seconds per day in favor of a nice round 90,000 seconds per day?  Or if it’s really all about POSIX, brain damaged computers and incompetent programmers, why not make it easier for them and have a day be a nice round 2^16 SI-seconds?  Wouldn’t they all be happier with 65,536 seconds per day fitting cosily in a unsigned short int?

Rather, there have to be very close to 86,400 SI-seconds per day, because that how long a day is.

Rob