[fitsbits] FITS vis a vis XML

[Francois Ochsenbein]

In the debate of XML and FITS, just a few words about our motivations
which lead to propose the "astrores" XML format for providing tabular
data as results of queries to catalogued data, rather then using FITS
tables (definitions at http://vizier.u-strasbg.fr/doc/astrores.htx):
which was presented at the last ADASS meeting.

First, the context: an application has to query astronomical catalogues
typically to find out which are the stars, galaxies, radiosources, etc
which exist in some region of the sky, for further processing (eg overplot
on digital images). The application requires obviously minimal metadata
describing the tables returned as the result of the query; why did we
choose an XML formatting rather then FITS which in an obvious candidate
for returning tables with their associate descriptions ?

The difficulties with FITS in this context:
=> FITS has 2 formats for tabular data; the binary tables (which seem to
   be preferred ?) have either to be encoded in some way, or to be loaded
   by a second connection; iven with the "ascii tables" format, the data
   can't just be displayed with the standard tools that are available 
   on a PC or a Unix station.
=> The interpretation of FITS tables in Java currently requires large pieces 
   of code -- too large to be useable efficiently within applets
=> there is no facility (yet?) in FITS for managing links -- definitions
   which specify how to get the related data or documentation, which is
   a requirement in a distributed environment. Just the limitation to 70
   characters in the headers prevents from writing (not so long) URLs !

XML was preferred because of these limitations, but also because going
this way allows immediately to _display_ the (eventually nicely formatted)
document on XML-compliant browsers, _and_ find in the same document the
structured metadata required by the application for the intepretation of
the data.

However, knowing the large usage of FITS in astronomy, we tried to keep
in our XML definitions the elements which allow to re-create the FITS (ascii)
table from the XML tags and attributes.

Just some final remarks: 
=> astrores is only dealing with tables --- FITS is of course dealing with 
   many more entities 
=> mixing XML and FITS which both provide metada describing the data 
   contents and structures in the same document would inevitably generate 
   conflicts --- any modification in the XML dscription would have to be 
   reflected in the FITS headers and vice-versa; the convertibility
   paradigm looks more appropriate, and safer too.
=> astrores _is_ currently working and being used, over 2,000 catalogues
   represented by 6,000 tables can be queried with results supplied in 
   this "astrores" XML output, try e.g.
   http://vizier.u-strasbg.fr/cgi-bin/asu-xml?-source=GSC1.2&-c=3C+273

Any comment about the subject welcome -- things are not yet frozen,
hopefully !
======================================================
Francois Ochsenbein
Centre de Données astronomiques de Strasbourg
francois at astro.u-strasbg.fr / ochsenbein at evc.net
======================================================