[fitsbits] Re: leap second alert

[Steve Allen]

On 1999-12-17T18:13:29 -0500, Patrick Wallace <ptw at star.rl.ac.uk> replied:
>something not all data acquisition applications can tolerate

I believe that we are discussing two different things, and not
disagreeing over the usage of the FITS 'DATE' keyword.

The FITS 'DATE' keyword is not necessarily related to data
acquisition.  It is the time of creation of the HDU which contains it.
The observation(s), if any, in the HDU should use other 'DATExxxx'
keywords with values expressed in the appropriate timescale.

My computer does not have a reliable source of time (other than UTC
via NTP), but it also does not perform data acquisition.  It does
regularly create FITS HDUs (mostly consisting of tables) which do
contain DATE keywords for documentary purposes.  There are many other
computers like mine (notably most machines that run IRAF or any other
post-observation reduction tools) which do the same.

Early drafts of the FITS WCS papers had specified TAI as the
timescale, and the earliest draft of the FITS Y2K agreement had
preferred UTC for all 'DATExxxx' keywords.  Both of these were deemed
too restrictive and discussions of the Y2K agreement pretty much broke
down over this and a few like issues.  Agreement did not resurface
until the text was revised to permit other timescales.

Nevertheless, the Y2K agreement singles out the 'DATE' keyword as the
only special case where a timescale was specified.  (This addition was
performed sometime between the 1997-07-03 and 1997-10-21 drafts from
Arnold Rots after discussion moved from the American WFC to the IAU
FWG.)  I've presumed that UTC was chosen because, as it was civil
time, it was already the most widely-used and would continue to be the
most widely-available timescale for the computers which will be
writing FITS HDUs.  (Explicit text was later added to exempt FITS HDUs
written by non-terrestrial computers from the UTC requirement under
the presumption that UTC might be neither available nor relevant.)

But in any case, to get back to the two questions posed
by Rob Seaman <seaman at noao.edu> on Thu 1999-12-16T20:05:37 +0000
> First - what does astronomy need?  (And therefore what must FITS support?)

Whatever the time and frequency community is proposing appears to
presume that something called UTC will persist and continue to be
widely available.  I hope that at this point we can all agree that the
FITS Y2K agreement was crafted carefully enough that such an action
will not break FITS.  If so, we risk straying from the charter of this
discussion list (not that I mind).

As for what "astronomy need"s, and Rob's other question
> Second - what are the resulting implications of this for civil time?
is there anything which hasn't already been covered here?

--
Steve Allen          UCO/Lick Observatory       Santa Cruz, CA 95064
sla at ucolick.borg     Voice: +1 831 459 3046     FAX (don't): +1 831 459 5244
WWW: http://www.ucolick.borg/~sla               PGP public keys:  see WWW
Junk mail is irrelevant -- my return address has been assimilated.