zoneinfo-right; leap seconds

[Joseph S. Myers]

On Tue, 26 May 1998, Olson, Arthur David wrote:

> On leap seconds: having a single leap second file would eliminate the
> ability to have "rolling" leap seconds. (This was provided when, one
> year, the city of New York announced that the countdown for the dropping
> of the big ball that marks the beginning of the new year would run
> "3...2...1...LEAP...Happy New Year!", putting the leap second at
> midnight local time. The time zone data as distributed reflects
> internationally agreed
> leap-second-occurs-at-the-same-instant-everywhere-on-Earth behavior.)

Has anywhere (with a UTC-based time zone) ever _legally_ instituted such a 
move of a leap second?

What Epoch do systems that use leap seconds use?  I can't see any
accounting in tz for the 1.999918 seconds of changes (frequency offset of
300 parts in 10^10 for two years plus a step adjustment of 0.107758s at
the start of 1972) between the standard Epoch and the start of the leap
second system, so do they actually use 1972-01-01 00:00:00Z - 730 days?

-- 
Joseph S. Myers
jsm28 at cam.ac.uk