[fitswcs] Re: WCS activity

[Mark Calabretta]

On Thu 2001/06/14 16:08:52 -0400, Don Wells wrote
in a message to: Mark Calabretta <Mark.Calabretta at atnf.csiro.au>
and copied to: Don Wells <dwells at NRAO.EDU>, Steve Allen <sla at ucolick.org>,
      Eric Greisen <egreisen at NRAO.EDU>, Ernst Raimond <exr at NFRA.NL>,
      Peter Teuben <teuben at astro.umd.edu>, Preben Grosbol <pgrosbol at eso.org>,
      Osamu Kanamitsu <kanamitu at fukuoka-edu.ac.jp>,
      Patrick Wallace <ptw at star.rl.ac.uk>,
      William Pence <pence at tetra.gsfc.nasa.gov>, "Doug Tody" <tody at noao.edu>,
      "Lindsey Davis" <davis at noao.edu>, "Frank Valdes" <valdes at noao.edu>,
      fitswcs at NRAO.EDU

>There are two models which could be used to produce FITS standards:

Dear Don,

As you intimated, there are more than just two models to choose from.  I'd
like to talk more about that, but later.

My present point is that, whatever model you're using, it is clearly not
working.

May I remind you of what you said in your first exhortation to the fitswcs
list three years ago (the full text is appended):

   ...but the HST
   images were off in HFF by as much as 500_mas. Indeed, the final
   viewgraph, the conclusions of the talk, had as its final bullet a
   statement to the effect that further study of the relationship of the
   radio imagery to the optical must await improved coordinates for the
   latter. I felt embarrassed that this purely technical issue loomed so
   large in this scientific presentation. As I thought about it, I got
   'ionized', and decided to wait until today to express my feelings to
   all of you. Today I still feel strongly about this matter:

            THIS SITUATION IS INTOLERABLE! SCIENCE IS *HURTING*!

   In the question period after the talk I alluded to the new
   Hipparcos/radio reference frame and asked whether there was any reason
   why this would not solve the problem. The answer I got from the users
   in the room was that WFPC distortions were not properly accounted-for
   in the matching of the radio and optical imagery. This is a dramatic
   illustration of the *scientific* need for us to agree on a FITS
   convention for the interchange of pixel correction functions.

I agree with the sentiments expressed in your email.  Astronomy has been
the loser.

The WCS proposal was first put for public comment in Feb/1993 - more than
eight years ago.  In fact, its history predates the Feb/1988 NASA-sponsored
meeting in Charlottesville.

Surely this timespan alone, 13 years out of FITS' 20 year official lifetime,
should be enough to tell you that the process is dysfunctional.  Thirteen
years.  That's longer than it took NASA to put a man on the moon!

The cause is no mystery; for extended periods of time there has been no
activity, absolutely none.  The tardy development and adoption of the WCS
standard is due to its lackadaisical prosecution.  The fiction of secret
backroom discussions is simply a disguise for this inactivity.  As far as I
can tell, in the 5 years between Feb/1988 and Feb/1993, nothing happened;
the Hanish & Wells (1988) document still has DRAFT written across the title
page.

During the 8 years of Eric's and my tenure, we have had periods of activity
lasting a month or maybe two, followed by eerie silences lasting for six to
twelve months.

This WCS proposal is not a fine bottle of wine, to be laid down in a quiet
place undisturbed for years.  It was never going to get better simply by
getting older!  Quite the reverse:

   1) The reason I became involved in FITS WCS in the first place was to
      ensure that the general celestial coordinate types I envisaged for
      aips++ would be transportable and archivable.

      This effort has been completely frustrated, thus...

   2) ...the demonstrated lack of an efficient process must discourage
      further standards proposals from us or others.

      How would you answer the question "Will I spend 10 years to promote
      this usage as a FITS standard?"?

      Given the agony of this process we have declined to consider some
      natural extensions to it, in particular, time measurement.

   3) Because of the proposal's extended gestation and the ongoing
      uncertainty over its status, FITS writers must choose either to

      a) start using it and risk incompatible changes: the PROJPn and PCij
         came and went, leaving some doubt about whether any data was
         written with these keywords.

      b) defer using it until it's ratified - but, what to use in the
         meantime?  Hence the many informal "temporary" WCS conventions
         (AZ/ALT, -LIN, COi_j, -ZPX, -TNX), with the risk that these
         superfluous constructs will force their way back into the proposal
         and/or its software implementation.

      It's a no-win situation.

   4) WCSLIB languishes until the proposal is accepted lest its interfaces
      change once again.

   5) The process is free floating, people feel no sense of urgency, they
      can always raise an objection later and turn the proposal back through
      yet another of its seemingly interminable loops.  Anyone can do it!
      Your authors and those waiting for the standard are at their mercy.

      Every time we appeared to have convergence we were frustrated once
      again.  And so, having been stuffed around like this for so long, you
      have now alienated your two authors.

I certainly do not deny that the proposal has also benefitted from its long
gestation, but it's way out of balance.  The same result could have been
achieved very much faster if you had driven the process with more vigour.

Mark Calabretta, ATNF
Eric Greisen, NRAO



>>> Included message >>>



Date: Wed, 15 Jul 1998 14:14:34 -0400 (EDT) (Thu 04:14 EST)
From: Don Wells <dwells at nrao.edu>
To: fitswcs at fits.cv.nrao.edu
Subject: [fitswcs] Exhortation

Dear FITS WCS friends,

In spite of feelings of frustration and dismay, I haven't given up on
a WCS agreement, and I trust that none of you have either.

	   -=- Scientific requirement for WCS agreement -=-

There was a lunch talk here at NRAO-Charlottesville yesterday on VLA
and MERLIN imaging of the Hubble Deep Field [HDF] and Hubble Flanking
Field [HFF] areas. There are many fewer radio sources than optical
sources in the HDF and HFF, but the questions raised by the detections
and non-detections concern several of the hottest topics of current
astrophysics: (1) the radio and infrared observations provide
invaluable handles on the history of star formation rates in the
Universe and (2) a few radio detections which are blank in the HST
optical imagery raise questions about completeness, about whether the
census has reached the 'edge of the Universe'.

Throughout the talk the speaker (a radio astronomer) noted failure of
radio contours superposed on the optical imagery to align with optical
features, and stated that these misalignments were errors in the
optical imagery.  The radio coordinates were good to 40_mas
(0.04_arcsec, MERLIN resolution is comparable to HST), but the HST
images were off in HFF by as much as 500_mas. Indeed, the final
viewgraph, the conclusions of the talk, had as its final bullet a
statement to the effect that further study of the relationship of the
radio imagery to the optical must await improved coordinates for the
latter. I felt embarrassed that this purely technical issue loomed so
large in this scientific presentation. As I thought about it, I got
'ionized', and decided to wait until today to express my feelings to
all of you. Today I still feel strongly about this matter:

	 THIS SITUATION IS INTOLERABLE! SCIENCE IS *HURTING*!

In the question period after the talk I alluded to the new
Hipparcos/radio reference frame and asked whether there was any reason
why this would not solve the problem. The answer I got from the users
in the room was that WFPC distortions were not properly accounted-for
in the matching of the radio and optical imagery. This is a dramatic
illustration of the *scientific* need for us to agree on a FITS
convention for the interchange of pixel correction functions.

I don't want to give the impression that WCS interchange problems are
peculiar to HST data. Not so! The problems are already pervasive and
are destined to become worse unless we implement consensus WCS
solutions. New groundbased optical telescopes are commonly producing
imagery with PSFs of 400_mas
(http://www.eso.org/outreach/press-rel/pr-1998/phot-20-98.html). Under
good conditions the new optical telescopes have even higher resolution
(270_mas in
http://www.eso.org/outreach/press-rel/pr-1998/phot-15-98.html).  A
considerable number of these telescopes (Gemini, Subaru, etc) are
about to achieve first light.  AXAF
(http://asc.harvard.edu/AXAF-description.html), to be launched this
year, expects to get 500_mas resolution; when X-ray cameras had
4000_mas resolution we could afford to be sloppy, but not with AXAF!
Intercomparisons of radio, HST, ground-based optical and X-ray imagery
with 500_mas resolution are going to be severe tests of World
Coordinate System technology.

I also do not want to give the impression that HDF/HFF studies are the
only science problems that need WCS technology. For example, the study
of gamma ray bursters is inherently a multi-wavelength ground- and
space-based subject and is critically dependent on WCS techniques.
Studies of protostars will depend on aligning NICMOS imagery with
radio imagery. Active galactic nuclei are now studied in all frequency
ranges and important questions now depend on comparisons of
radio/optical/X-ray imagery.

Our community must solve the WCS interchange problem. We can no longer
put it off.  The Hipparcos reference frame can be aligned with the
radio reference frame to 0.5_mas RMS by means of a single small
rotation. Therefore, there is now one universal reference frame (the
ICRS, http://riemann.usno.navy.mil/AA/data/docs/ICRS_links.html) to
use for all of our imagery, and we should all be using it. No
colloquium speaker should spend his/her time apologizing for bad
coordinates in radio/optical/Xray/infrared image comparisons.

		    -=- Other responsibilities -=-

I confess that I have not worked on this business myself for the past
three or four months, because I was concentrating all of my abilities
on the problem of understanding the focus tracking algorithm for the
Green Bank Telescope. I recently completed that, and so I am starting
to think about WCS again. My bet is that many of you are also
distracted by your formal responsibilities for your major projects,
so I am sympathetic to your situations, but I hope that you will all
agree with me that achieving agreement on WCS interchange conventions
is still a critical goal of the FITS community. 

				 -=-

I would like to be able to report to ADASS'98 that we have resolved
the major WCS issues which were still in contention at the Sonthofen
ADASS last year. In particular, I want us to report that we have
agreed to split the G&C paper and that revision of the text is
work-in-progress. I want us to report consensus on the form of the
linear transformation (CDij versus CDELTi/PCij). I want us to report
consensus on the solution to the multiple-WCS requirement. Finally, I
want us to report real progress on the pixel correction function
problem. It would really help if we could report that our operational
software is making progress toward a universal interchange solution.

Please join with me to achieve these goals!

Regards,
Don
-- 
  Donald C. Wells         Associate Scientist         dwells at nrao.edu
		    http://www.cv.nrao.edu/~dwells
  National Radio Astronomy Observatory                +1-804-296-0277
  520 Edgemont Road,   Charlottesville, Virginia       22903-2475 USA