(Sorry about posting unfinished answer. Please ignore, this is more complete.)
On 2015-05-22T00:38 Chris Barker wrote:
On Thu, May 21, 2015 at 11:36 AM, Jim Biard <jbiard at cicsnc.org<mailto:jbiard at cicsnc.org>> wrote:
The point I'm trying to make about gregorian_nls is that it lacks any mechanism for specifying which time system is being used for the timestamp in the reference date & time.
and that, indeed is the only thing we need the calendar specification for, yes? (after that, seconds are well defined).
I'm starting to be convinced that one might need the leap second specification not only for the reference timestamp, but also for the correct scale of the offset values, i.e. whether the leap seconds should be assumed accounted for (true distance in seconds) or somehow fudged with (as ignored in the Posix time routines). In other words the seconds aren't as well defined as we might want to believe either.
Furthermore, the discussion and proposed solutions have focused a lot on the unit of "seconds since", but what about the others like "minutes since" and "days since"? How would leap seconds need to be interpreted in the context of gregorian_nls vs _utc vs _tai for instance? If they are to be ignored altogether for any lower resolution than seconds, then specifying the added precision for the reference time stamp (instead of just "gregorian") may seem somewhat ambiguous perhaps.
As the leap seconds accumulate closer to the minute between the different "CF calendars", then the implied loss of precision possibly interpreted by using minutes instead of seconds may no longer be ignored.
--
-+-Ben-+-
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mailman.cgd.ucar.edu/pipermail/cf-metadata/attachments/20150522/c4d04b0a/attachment.html>
Received on Fri May 22 2015 - 07:57:36 BST
