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

On 10/18/2016 6:30 PM, John Yeung wrote:

On Tue, Oct 18, 2016 at 4:36 PM, Buck Calabro <kc2hiz@xxxxxxxxx> wrote:

On 10/18/2016 3:47 PM, Charles Wilt wrote:

I find the idea that your DB has to support 2016-12-31-23.59.60 difficult
to accept...

23.59.60 is a real time.

I think you need to clarify. Is February 29 a real date? Sometimes it
is. But day 29 certainly does not occur every February.

The database people seem to have the weird rules for the date part of a
time stamp implemented. This argues that they ought to be able to
implement the weird rules for the time portion as well.

Leap seconds definitely do not happen every day. Since 1972, there
have only been 26 instances when 23.59.60 was a real time.

ISO 8601 allows for leap seconds, just as it allows for leap years.
Does DB2 have a 'we don't need to store legitimate but rare values'
exemption? I didn't see one.

While a leap day is big enough for ordinary people to notice if it's
missing, leap seconds are utterly inconsequential for the vast, vast,
vast majority of people and businesses.

Typically for me, I used a stupid example. Let's say that a trading
partner sends me a transaction and it's stamped 2016-12-31-23.59.60 UTC.
I didn't generate it, I don't have control over the stamps they send
me. What should I do with that transaction?
Alter the incoming timestamp?
Reject the transaction?

Right now, the least worst answer seems to be that I should stop using
TIMESTAMP data types and revert to two separate columns instead: a DATE
and a TIME (which I can NULL out 'rare but legitimate' values of)? At
least then I can store the date part which will (mostly) let me answer
questions like 'how old is this invoice?'