Re: [LEAPSECS] decision tree for civil time

From: Rob Seaman <seaman_at_noao.edu>
Date: Sat, 13 Aug 2005 21:55:12 -0700

On Aug 13, 2005, at 5:00 PM, Greg Hennessy wrote:

> Well, even though they are predictable, lots of computer programs
> still got them wrong.

I'm not interested in disputing this statement, rather, we've often
heard about various problems implementing and interpreting temporal
algorithms and interfaces. Assume this is true. What is different
about the handling of time from all other algorithms and interfaces
imposed on computing? Certainly in scientific computing there is
nothing particularly special about time handling compared to lots of
other pretty fancy algorithms and complex interfaces.

But surely this applies to general purpose computing as well - and
not just the subtleties of operating systems, either. If a typical
programmer can be expected to master hash tables, sorting algorithms
(or anything else out of Knuth), the ridiculously "rich" class
libraries of Java, public key cryptography, pixel and other spatial
transforms, bit and byte operators - why not leap seconds?

I guess I'm a little unclear. Nobody suggests that we dumb down J2EE
or pretend that race conditions don't exist. So why are leap seconds
poison?

Rob Seaman
National Optical Astronomy Observatory
Received on Sat Aug 13 2005 - 21:55:19 PDT