Re: ABSTRACT: Please refer to following table concerning impacts of the leap second change in 2016.

Buck, I'm not arguing against being able store all valid timestamps.
If IBM can and wants to implement it, I'm 100% fine with that.

What I'm arguing is (1) that it's not particularly bad if there isn't
any provision for storing leap seconds as distinct timestamps, and (2)
allowing for storage of distinct leap-second timestamps will either
mean you have to also open the door for a whole lot of invalid
timestamps (vastly more numerous than the valid ones you're adding),
or you have to have a costly, complex, and brittle system to prevent
invalid values and adjust calculations for the new valid ones.

For example, are you proposing that we should be able to enter
1901-06-30-23.59.60 today, as a valid value? How about
2018-12-31-23.59.60? If we enter the latter, and consider it valid
today, what happens in a couple of years if it turns out we never did
take that leap second? Does it then become invalid?

The situation is not like what we have for leap days, which can be
determined through pure calculation. We don't have to wait for someone
to announce that a leap day will be inserted or not.

The situation is also not quite like what we have for daylight saving,
because UTC expressly ignores any notion of daylight saving (it's up
to the individual locales to keep track of stuff like that), while
leap seconds are a mechanism *specifically* for UTC (and for the most
part, everybody else can just ignore them!). What daylight saving and
leap seconds have in common is that they're irregular and depend on
external decisions we're not privy to, and thus implementing them
rigorously involves a look-up table and a lot of ugliness. The look-up
table has to be maintained vigilantly and manually by someone
somewhere. (Most likely not you or me, but someone. Somewhere.)

I'm sympathetic. Of course it's good to be able to store all legal
values. I'm not against being able to store all legal values. I am not
convinced that these particular legal values are very important to
store, and I'm especially wary of all the complexity and/or invalid
values that would necessarily be introduced in order to accommodate
these few special values.

John Y.