On Tue, Feb 07, 2012 at 12:43:11AM +0100, Nicolas Cellier wrote:
> 
> 2012/2/6 Enrico Spinielli <enrico.spinielli at googlemail.com>:
> > For leap seconds and how they are (mis)handled by the different OSs/distros
> > have a look at http://queue.acm.org/detail.cfm?id=1967009
> >

Enrico,

Thanks for this link, it is a good read.

> Thanks Enrico for this instructive and funny article.
> I really doubt that so many airplanes are driven by NTP, except maybe
> your home simulator, and even less so any TCAS.
> And I agree with first comment of this post, if leap second would be
> handled just like another localOffset, coding would be much much
> simpler, because it is already supposed to deal with DST. Then OS
> vendors would just have to routinely update the offset tables, like
> they do when some countries change the law (the main reason why I'm
> not fond of duplicating this administration in image).
> Until nuclear plants are driven by Squeak and NTP, I suggest to not
> bother, a pause of 1 second is just a medium garbage collect anyway ;)
> One last, thing, isn't it a clock synchronization problem that made
> Einstein invent his theory of relativity? Funny that I don't see a
> word about it, either every one consider the problem is solved, or
> they just forgot about it. Good luck though for the definition of
> inter-galactic "true universal time" found in the wish list among
> comments ;) Especially a coordinated one, no matter some galaxy may
> suffer from a few million years delay before they are informed of any
> leap adjustment bound to latest estimation of universe spreading ;)

Nicolas,

Funny you should bring that up. I also made a joke of it when I wrote
the class comment for PointInTime (in the TimeZoneDatabase package) many
years ago:

  "I am based on the naive assumption that time can be represented as
  a single continuum which may be experienced identically at different
  physical locations and velocities, and such that two observers in two
  different locations can identify a single PointInTime as a simultaneous
  event. These assumptions are not valid when relativity is taken into
  account, but are acceptable for use in Newtonian physics and Smalltalk."

;-)

I think that leap seconds are easier to understand when you can look
at them in Smalltalk. If you have TimeZoneDatabase loaded in your
image, try "TimeZoneDatabase systemDatabase explore". You will see
a leapSecondRuleSet that is shared by all time zones that make use
of leap second rules (because leap seconds apply to the rotation of
the earth as a whole, so the same rules apply everywhere). You can
look at the various time zone rule sets by looking into the timeZones
dictionary (which is a dictionary of dictionaries, patterned after
the directory structure of /usr/share/zoneinfo). Time zone names
like "right/America/Detroit" use the leap second rules, while the
"America/Detroit" time zone is exactly the same except that it does
not use the leap second rules (don't ask me why it is called "right",
I have no idea).

Dave