> Whether you choose to consider 23:59:60Z part of 23:59Z or not is
> irrelevant. It is represented as such when using Zulu time, which is
> all that matters. It's on the same shelf as the other seconds in this
There's actually only one very long shelf. On second thought, ISO 8601
mentions at least two: one for UTC and one for UT1. The leap seconds
(and days?) are used in an effort to keep the two in sync.
> UTC was not created by ISO. ITU-R is responsible for the definition
> of UTC. Physically, UTC is maintained by BIPM and leap seconds are
> determined by IERS.
But do any of those bodies document exactly whether or not the second
belongs to the minute? And if so, does ISO 8601 reference such a
document, even if non-normatvely?
> I'm still wondering how negative leap seconds are possible, if hours
> and minutes can never vary.
Yeah, that's pretty much where my arguments hit a wall (so I've avoided
In order for that to work (and it wouldn't) there is no T23:59:59Z that
day (as stated, it's omitted), so for the minute T23:59Z to have sixty
seconds it must borrow T24:00:00Z (T00:00:00Z from the following day).
The T00:00:00Z in question therefore belongs to _both_ minutes...
ummm... which shouldn't be allowed. But at least balance would be
preserved, a positive leap second belongs to no minute and a negative
leap second belongs to two.