Re: [LEAPSECS] decision tree for civil time

From: Rob Seaman <seaman_at_noao.edu>
Date: Sun, 14 Aug 2005 07:30:08 -0700

On Aug 14, 2005, at 4:11 AM, Ed Davies wrote:

> Could come under "other" perhaps but millisecond steps are worth
> listing explicitly. At least, that's how I assume the epsilon
> scheme would be implemented.

These options are meant to describe civil time, not timekeeping in
general. To permit fixed size millisecond leaps would require the
flexibility to schedule a leap throughout the course of a day. Just
divide the typical annual or 18 month leap second latency by a
thousand. Many days would have three or more milliseconds to
accommodate. I was assuming (which is where we all have gotten into
trouble, no matter what scheme we're favoring), that the daily
epsilon would either be issued as a arbitrary (albeit currently
small) height leap at UT midnight or other daily schedule - or more
likely, that this would be handled like ntp-style time slices. In
fact, just input the epsilon as a non-zero clock rate to be ironed
out using the current algorithms. Over the centuries, of course, the
daily millisecond loads would accelerate - unless Calabretta had some
more subtle idea in mind.

That said, we're just trying to list the options now, so I'll add a
millisecond option.

Rob Seaman
National Optical Astronomy Observatory
Received on Sun Aug 14 2005 - 07:30:28 PDT