[time-nuts] FW: Bulletin C number 30
M. Warner Losh
imp at bsdimp.com
Mon Jul 4 15:16:08 EDT 2005
In message: <31556.1120503755 at phk.freebsd.dk>
"Poul-Henning Kamp" <phk at haven.freebsd.dk> writes:
: The other half is that leap-seconds are just not testable in a computer
: setting, and therefore I am sure that any cost of dropping them will
: be totally offset by the savings in the IT industry.
I can tell you as a writier of time software that we spend a *HUGE*
amount of time on leap seconds, on finding the current offset of leap
seconds, etc. There's so many places where they insinuate themselves
into such simple notions as knowing what time it is. You have to know
them before you can use GPS time to recover UTC, for example, since
GPS doesn't have leap seconds.
Then there's the issue of actually crossing a leap second boundary.
What happens to time? What's the right time to put out? How does
your application deal with time going backwards (it does on unix
steered to ntp)? How does your threading package deal? Is it OK if
your application freezes for a second? etc etc etc.
Leap seconds are a huge pain. I've spent literally hundreds of hours
worrying about these concerns, coding to these concerns and working
around others that couldn't be bothered and got it wrong :-(.
More information about the time-nuts