FW: Observation: replies to GL and others

[Francois Bonnarel]

Dear Gerard, Dear all,

I try to answer now the mail (actually the note) posted on this list by Gerard Lemson in

 http://ivoa.net/forum/dm/0454.htm

First of all I want to thank him for stating with so much talent and clarity his point of view,
which I think is shared by a large bunch of people in the VO.

Unfortunatly I think I disagree with the background philosophy of this point of view. I mean
I don't think we need VO Data model effort and associated theoretical work to modelize ALL 
astronomy and to DEDUCE from a very abstract and very general data model  (or "ontology") 
specific views for any kind of practical interoperabilty problems in astronomy. And Gerard, this
doesn't mean that I do not think that your attempts for an astronomy domain model is not 
interesting as I hope will become clear in the following sentences. My concern is about the
way you apparently want to use it.

    Interoperabilty in astronomy has a long history and existed before the VO (FITS, bibcode, UCD1, 
ASU, etc ...).
 And as far as the VO time is concerned we have had interoperabilty results such as VOtable , SIA,
 cone search, tommorrow SSA and UCD1+... with tools built on it, demos, allready public services, etc ...
 An this without any major achievment  in the data model field (unachievement which I consider normal,
 because it is so difficult a  task). "The fact preceded the law".
   This said, I must tell that I DO NOT THINK we do not need this data model and abstraction effort, 
and that we can go on just by adding circles around what exists. When an astronomer modelizes a 
cataclysmic binary he is at least doing two things:
        a ) describing what is observed in both a conceptual and quantitative manner.
        b ) trying to do some predictions on other kind of observations which can be done elsewhere.
What are we modelizing in the IVOA datamodel working group? The practical activity of astronomers
an related technical staff in the   "data" field.
     What can we expect from emerging datamodels?
             1 ) describe with as much accuracy as possible the various existing interoperabilty tools, 
       formats, etc...
             2) try to unify, generalize all these, in order to  extend the field where interoperability can
be used,    and btw fix errors or remove unusefull or obsolete features. 2 supposes that we can rely
on 1.

      That's the reason why I think the utype concept is so important, because I think this is the tool
to describe precisly in a conceptual way the various features we are using. There has been a very
interesting attempt made by Jonathan a couple of weeks ago on defining utypes from the IVOA 
Observation data model on the SIA protocol fields. I plan to send on this list  probably next week,
at least before Boston an other attempt with some more utypes on another example given in the
collective note "SIAP Evolution proposal". More than that I think we should try to use utypes
for description of other existing interoperability tools, not only for VOTable fields.

      As far as UCDs are concerned, what you ( Gerard) describe in your note is typically what has been
called in the UCD group UCD3, directly connected to ontologies. I think this does not prevent us to
describe the attributes we define in our emerging datamodels with the current UCD1(+)s which are 
there for comparison of quantities and astronomical concepts across datamodels. 

Cheers
François

=====================================================================
Francois   Bonnarel               Observatoire Astronomique de Strasbourg
CDS (Centre de donnees          11, rue de l'Universite
astronomiques de Strasbourg)    F--67000 Strasbourg (France)

Tel: +33-(0)3 90 24 24 11       WWW: http://cdsweb.u-strasbg.fr/people/fb.html
Fax: +33-(0)3 90 24 24 25       E-mail: bonnarel at astro.u-strasbg.fr
---------------------------------------------------------------------