FITS standard

[Steve Allen]

In article <199804161933.PAA01740 at urania>,
Jonathan McDowell <jcm at urania.harvard.edu> wrote:
>paying attention to the Y2K discussion enough, but I am puzzled by the
>statement in 5.4.2.1 that UTC shall be used for the DATE keyword
>'for all data sets created
>on earth'. Is it felt that there are practical difficulties in realizing
>UTC for computers not on earth? This seems unlikely to the accuracy
>required by the DATE keyword; it is also not a frivolous question;
>there are enough laptops flying on Shuttle and Mir missions which might
>in principle store their data in FITS. I suggest removing the 'on earth'
>qualification (or replacing it by 'in the solar system', since I can
>see there might be simultaneity issues for interstellar probes...)

Note that the "on earth" clause under discussion pertains to the 'DATE    '
keyword, not the 'DATE-OBS' keyword.  The 'DATE    ' keyword is the
time of creation of the FITS HDU, not the time at which the data were
obtained.

At the time of the insertion of this clause the Y2K agreement had
already left the cabal of Rots, Wells, and myself, passed thru the
WGAS vote, and was in the hands of the IAU FWG.  Although I wasn't
officially party to the evolution of the agreement at this phase, I
suspect that my suggestion was responsible for it.

My reasoning follows.  We usually presume that the equipment which
obtains the data has a clock suitable for defining a value of
'DATE-OBS' in whatever is the most convenient time scale.  However,
these data may not be stored immediately as FITS, but may be
transferred to some other computing device which casts them into
FITS.  It is the responsibility of that machine to create the 'DATE    '
keyword, and there need be no requirement that it has a clock which is
set to any particular timescale.  Furthermore, the 'DATE    '
keyword should be rewritten if the original data are reprocessed.

On earth the most available timescale may, without much loss of
generality, be presumed to be UTC.  Note that there is no prohibition
against using UTC for data sets created off earth.

It was my assertion that it was asking too much to presume that a
computer in the data processing center at Farside Lunar Observatory,
or Mars Base 1, or of a L5-based amateur who is simply playing with
the data, should have a clock which is set to UTC.  The clocks of such
non-terrestrial, non-data-acquisition computers should be permitted to
be set to whatever is the most available timescale.  In general, I do
not think that likely to be UTC.

I do not see any significant compromise in permitting the use of other
timescales for 'DATE '.  Indeed, my suggestion to members of the IAU
FWG indicated that I felt that, without this clause, such
non-terrestrial FITS creators would simply curse the Y2K amendment,
and then proceed to disregard it.  With this clause in place, the Y2K
agreement should be able to withstand all solar system use.
(Interstellar applications will beg the question of the use of the
earth-based Gregorian calendar, but that is a matter for a future FITS
working group ;-)
-- 
Steve Allen          UCO/Lick Observatory       Santa Cruz, CA 95064
sla at ucolick.borg     Voice: +1 408 459 3046     FAX (don't): +1 408 454 9863
WWW: http://www.ucolick.borg/~sla               PGP public keys:  see WWW
Junk mail is irrelevant -- my return address has been assimilated.