[fitswcs] forwarded message from Eric Greisen

[Eric Greisen]

Since Don has opened this discussion up without including my reply, I
open it further and include my reply which still seems reasonable to
me.  I do not like private discussions amongst small groups when that
is not essential.

Eric Greisen

------- start of forwarded message (RFC 934 encapsulation) -------
Received: from fits.cv.nrao.edu (root at fits.cv.nrao.edu [192.33.115.8])
	by primate.cv.nrao.edu (8.8.5/8.8.5/CV-2.3) with ESMTP id PAA07097
	for <egreisen at primate.cv.nrao.edu>; Tue, 27 Jul 1999 15:18:19 -0400 (EDT)
Received: from primate.cv.nrao.edu (primate.cv.nrao.edu [192.33.115.67])
	by fits.cv.nrao.edu (8.8.7/8.8.8/CV-2.2) with ESMTP id PAA02800
	for <egreisen at fits.cv.nrao.edu>; Tue, 27 Jul 1999 15:18:19 -0400
Received: (from egreisen at localhost)
	by primate.cv.nrao.edu (8.8.5/8.8.5/CV-2.3) id PAA07084;
	Tue, 27 Jul 1999 15:14:35 -0400 (EDT)
Message-Id: <199907271914.PAA07084 at primate.cv.nrao.edu>
MIME-Version: 1.0
Content-Transfer-Encoding: 7bit
In-Reply-To: <199907271732.DAA24586 at cetus.atnf.CSIRO.AU>
References: <199905171408.KAA21549 at fits.cv.nrao.edu>
	<199907271732.DAA24586 at cetus.atnf.CSIRO.AU>
X-Mailer: VM 6.35 under Emacs 20.2.1
Content-Type: text/plain; charset=us-ascii
From: Eric Greisen <egreisen at nrao.edu>
To: "Wim Brouw" <Wim.Brouw at atnf.CSIRO.AU>
Cc: Don Wells <dwells at cv.nrao.edu>, egreisen at fits.cv.nrao.edu,
        nkilleen at cetus.atnf.CSIRO.AU, mcalabre at cetus.atnf.CSIRO.AU,
        arots at head-cfa.harvard.edu, pgrosbol at eso.org, teuben at astro.umd.edu
Subject: Re: FITS pages 
Date: Tue, 27 Jul 1999 15:14:35 -0400 (EDT)

Wim Brouw writes:

> I had a look at the three A&A submitted articles. I was a bit amazed that it
> was already that far, and not in the form of a discussion paper to be
> considered say in Manchester next year.

    The papers have not been submitted to any journal at this time.
However, this matter has been under review already since the end of
1992 and we cannot afford to prolong it indefinitely for, among other
reasons, the problem you mention at the end.  It has become
traditional to negotiate FITS matters at ADASS meetings which have the
benefit of occurring once a year rather than once every three years.
IAU meetings also have not even been the sort of meeting that those of
us working in the field even attend.  That may have to change and I am
not opposed to discussing matters at Manchester if they are still
under active review at that time.

> 
> A few remarks on the Units side (the CUNIT keywords):
> 1. It is a different 'proposal' from the one I have seen from the High energy
> people.

    It began as a straight copy of their proposal, worded a bit more
cleanly.  Then Francois Ochsenbein pointed out that his data center in
France had a competing proposal.  The result now in the paper is a
merging of the two proposals and has been at some level accepted by
both groups. I had expected units to be a major bone of contention
and was very gratified that the two groups with the most invested were
able to reach a compromise agreement that I believe is really
workable.

> 2. I still wonder whether 'CUNIT' is the correct name for the field. In my
> view a uniot describes the value of a filed, with also the purpose of being
> able to convert between different units (say km and Mm, or s and min). Units
> like 'beam' make that all but impossible.

    Unfortunately, "beam" is one of the units actually used in
astronomy as are all (or most) of the others listed.   They do not all
lend themselves to elegant conversions, but to convert Jy/beam to
Jy/pixel or Jy/asec**2 requires a knowledge of what the beam area
actually is and, as you know, that is problematical in real
interferometers.

> 3. Using SI, even with IAU extensions, could be done by just saying m2 iso
> m**2. Also, I feel that any non-unit factors should be part of a scale and
> offset, not part of the unit.

     I also do not like using a scale in the units.  But the X-ray
folks use that and it was part of the compromise to leave it in.  I
reduced its flexibility to the minimum to leave the capability without
the possibility of multiple scales in one unit.

> 4. I cannot understand why log should be part of the unit definition (and
> sqrt, but no others like sin). It describes maybe something abouit the
> meausrement procedure, but I cannot see that it should be part of the unit
> field -- especially not since they are non-invertable (what about a Complex
> value?) 

    This was much discussed and the trigonometric functions were
dropped from the list.  Log and exp units are used for good and proper
reasons, but we were not confronted with any normal usage of
trigonometric axis units.  Complex values are no longer supported.

> 
> For proper use I would say:
> - units as SI (and IAU); with standard prefices (note that in contradiction to
> what you say in the text, da is two cahracters) 

   An oversight about "da".  I personally do not like any of the 10^1
and 10^2 sorts of units and so ignore them.

> - add a field (fields) for scale and offset

    We already have CRVALj, CDj_i, CRPIXi, etc.  Scale and offset are
not needed to describe axes and that is the use for CUNITj.

> - if necessary make comments about how things are done (function values)

    Comments are by their very nature things for people to read rather
than software.  We are trying to get a solution that might in many
cases be functional without special human intervention.  At the same
time we want it to be human readable, hence the use of connective
symbols such as ^ or **.

> - what about 'once fits always fits' for the existing unit kewords, with
> completely random strings in them?

     "Once FITS always FITS" applies to keywords and other matters for
which there has been a written and generally accepted agreement.  If
my AIPS program were to write some new keyword because I thought it a
good idea, then I cannot assume that any other FITS reader will
support that keyword even on the day I first wrote it, yet say many
years later.  Non-standard keywords of all sorts have been used for
wcs.  It is the intent of these papers to get an agreement.  Then we
will have to support that agreement under the once-FITS rule from then
on.

> 
> Will there be a discussion of this (and other?) FITS issues at Manchester?

It is our intention to discuss and vote on these papers at ADASS early
in October.  If they are generally accepted we will then submit them
to A&A Supplements.  If there is a general consensus to postpone and
modify the papers then they may be discussed in Manchester.  Since I
am not a member of the IAU and cannot even determine how I can go
about becoming one, I cannot speak about the agenda for the IAU FITS
working group.

Eric Greisen

------- end -------