Chronology Current Month Current Thread Current Date
[Year List] [Month List (current year)] [Date Index] [Thread Index] [Thread Prev] [Thread Next] [Date Prev] [Date Next]

Re: [Phys-l] leap second limbo



On 01/20/2012 01:49 PM, chuck britton wrote:

Neither International Atomic Time (TAI) nor Global Positioning System
(GPS) time incorporate the Leap Seconds.

The Unix world has something called UnixTime that doesn't use them either.

I wouldn't have said that. In particular, there is not a close
correspondence between TAI and unix time.
-- TAI: Doesn't need leap seconds and doesn't even try to implement them.
-- Unix: Needs leap seconds, but (usually) botches the implementation.

I will guess that the IT world - soft-and-hardware-wise will
eventually come up with some de facto standard long before the UTC
decides to give up these nasty Leap Seconds.

Actually, computer guys in general are *not* known for their deep
understanding of timekeeping. Remember Y2K?

I'm not an expert, but AFAICT it appears that currently-typical IT
practice is *worse* than either UTC or TAI.

Unix time
a) claims to represent consecutive seconds
b) claims to track UTC
c) in most applications, cannot represent a leap second.

Those three things are incompatible, because UTC includes leap
seconds.

Some applications could be easily rewritten to be smarter about
leap seconds, but others not so much ... such as the ones that
depend on the timestamps on files ... to the best of my limit
knowledge. I haven't been able to find a reliable, clear
discussion of the issues.