[LEAPSECS] stale leap second information
Tom Van Baak
tvb at LeapSecond.com
Fri Jan 16 10:20:06 EST 2015
> Presumably a negative leap would be denoted ending :58?
That's a good question, one that most leap second tables (or leap second code, for that matter) tend to ignore. Your :58 suggestion means his table is not really a table of leap seconds anymore. Instead its a table of names of the last second in the month. Although you and I know the difference, I think confusion like this is asking for trouble:
In the positive leap second case both the name of the leap second inserted and the name of the last second is :59:60.
In the negative leap second case the name of the leap second deleted is :59:59 and the name of the last second is :59:58.
There has to be a better way to symmetrically represent both positive and negative leap seconds in a single table.
Perhaps the BIPM favored (?) method of naming the first date of the month after the leap second is the right way to handle it. So instead of :59:58 or :59:60 for 2015-06-30 you simply use "2015-07-01 x" where x is pos or neg or +1s or -1s or something like that. This gives equal treatment to both type of leap seconds, as well as avoids use of number 58, 59, and 60, which are prone to off-by-one errors.
More information about the LEAPSECS