Crab et al., a warning Re: wcs.ps

[Eric Greisen]

Lucio Chiappetti writes:
 > On Thu, 1 Apr 1999, Eric Greisen wrote:
 > 
 > > Arnold Rots writes:
 > >  > I noticed that your list was missing the unit that is dear to many HEA
 > >  > types' hearts: the "Crab".  
 > 
 > >      if the Crab is as well defined as the Solar units, we could add
 > > it.  What is the definition?
 > 
 > About the definition of "1 Crab" I guess it could be good to agree a standard
 > definition. I believe everybody has one's own variant ...

      Arnold's earlier reply that "we don't want to go there" about
Crabs is well explained by your description of them.  I have no
intention of putting them in the paper, until the community gets a
much better definition.

 > 
 >                            BUT .. BUT ... BUT ...
 > 
 > One moment. Are we talking here of standardization of unit indication in
 > generic FITS files (tables) or just of WCS ?
 > And does not WCS apply "primarily" (or "originally") to IMAGES and
 > "secondarily" (or "more recently") to SPECTRA ?

     WCS applies to any observation of a physical quantity.  The issue
is most pressing when we attempt to associate coordinates with the
voxels in an image, but images as well as single pixels can and do
occur in tables and elsewhere.

 > 
 > I would expect it would be very difficult that the "spatial" axes of an image,
 > or the "dispersion" axis of a spectrum, would be measured in Crab units (or
 > in Ohm, if that matters).
 > So aren't we discussing about "the sex of angels" ?

     Derived quantities may also be sampled into images or tables and
expressed against whatever physical coordinates seem appropriate to
the Physics of the observation.  I have removed trigonometric
functions from wcs.ps but am concerned about matters that are indeed
functions of e.g. cos(latitude).  Maybe it is wrong to remove them.
I agree that Ohm is unlikely in astronomical data and so we are free
to change its spelling, but Crabs and Angstroms are not unlikely.

 > 
 >                             AND MORE BUT'S ...
 > 
 > Even if we would be discussing standardization of units in generic tables, I
 > reiterate my provocation, is it really worth while to go in extreme detail ?
 > For what are essentially "plot labels" ?

     WCS are not plot labels.  They have that function, but they have
a great deal more as well.  If units are expressed properly, then they
may be parsed in software and converted to other comparable units,
after which vast tables may be searched for matches, sub-set
selection, etc etc etc  This is now done with tabular data a great
deal with for example the NVSS survey here at NRAO and, judging by
Francois Ochsenbein's remarks, many of the tables at the Strasbourg
data center.  WCS for imagery is often used to regrid one image to the
coordinates of another so that the physics of the two may be compared
visually and numerically.  WCS is not a plot tool, it is a Physics
tool.

 > I'd expect humans (astronomers) be able to read a column heading and know what
 > it is (and convert to their favorite customary units) anyhow, while software
 > will just copy and use labels as they are, as "semantic-less" strings.
 > Do we really expect to have an all-encompassing universal routine library
 > which will convert any unit to any unit ?

      Yes - I do not have such a thing.  But I am told that numerous
Object Oriented tool sets (e.g. aips++) do have exactly that.  Current
data sets, tables, and images are already awfully large to depend on
astronomers to read them visually.  And as you point out below, the
working astronomer may not be the best equipped to understand
technical details of what you would force him to read.

 > 
 > I heard a saying "build a system than any fool can use, and only a fool will
 > want to use it". But it could also be reversed "build a standard that is too
 > clever, and nobody will be clever enough to want to use it".

      FITS as a whole was claimed when we started to be too hard
because we allowed and required BINARY data.  Where would we be today
if we had done only character-form data?  Where are ASCII tables
today?  I agree that we can get too fancy and I think that some
hierarchal relational data-base proposals that I have heard are just
that.  However, we do need also to get things right.  To leave things
inexact and prone to error and misinterpretation simply because "it is
too hard" is wrong.  The whole subject of WCS is hard and that is why
we have had to leave it until later in the FITS process.  It was
needed at the beginning (hence CDELTA, CTYPE etc in the first FITS
paper) and has been used in detail in some software from the early
1980s (e.g. aips).

 > 
 > I'm thinking of a recent discussion about a largish database private to a
 > smallish consortium (less than 10 institutes), of course to me it was obvious
 > to say FITS had to be used for interchange and not ASCII ... but it was
 > pointed to me that the programming competences of the "mean quadratic
 > astronomer" were quite below that mark ...
 > 
 > And if he/she will have difficulty in using age-old consolidated good old
 > plain FITS, is it likely it will adhere, or even check, a "marginal"
 > convention ?

     The initial software may not be the best that it could be.  But
with a good standard and the 3-sigma astronomer it can get better with
time if there is a standard to which it may aspire.  FITS tapes are
usually written even today with little or no WCS info.  They are
useful, but would be very much more useful if they had better WCS,
especially WCS that software would recognize.

 > 
 > Even when major agencies (funded also by my taxes) involved in major missions
 > produce and distribute tools which measure fluxes in eV/mm2/s (!) instead of
 > the more usual erg/cm2/s ?  :-) [sorry, I could not resist]

     But there is no problem with either of these units if they are
described properly.  The point of wcs.ps is to allow even misguided
folks to express their misguided opinions in such a fashion that other
misguided folks may correctly convert the data to their particular
religion.


     I knew that the units section would get people to talk about the
paper.  I just hope you are all also reading the rest of it and Paper
III (and Paper II when it becomes available).  The spectroscopy paper
is new and important.  FITS itself has guided how astronomers think of
their data and how software systems think of data for their users.
These WCS papers will, if adopted, affect how you will see data and
how you will receive them from your colleagues.


Eric Greisen