UTYPEs for SIAP

[Francois Bonnarel]

Hi there,

About Jonathan proposal and some of the answers:
   A utype tells you what attribute is your VOTable field in a given data
model (at least UML data models). I think I differ from Tom's and Brian's 
point of view: data models should not only deal with interfaces but also
with class contents: attributes, and links between them which build a
"logical" structure of the data. (The "class diagram" in UML)

    In this respect I like the kind of mapping that Jonathan does    
between the datamodel attributes and the SIAP fields. The apparent      
biunivocity with UCD1+ is just due to the small number of fields 
we are dealing with in SIAP. If we add new fields in a next version of SIAP
it may happen that several fields will share the same UCD, while they
will be described by different utypes.
   Jonathan, a few remarks however:
    The "new" (for Observation datamodel) "dataset" is what other people
(Alberto Micol, and eventually CDS) call "Packaging"?
    About the two kind of Mapping that you are proposing, I definitly
prefers the first one, because it's clearer to the user. Independantly from 
the SIA context all data model attributeis  should have a UCD. 
For these attributes allready covered in SIAP it is exactly the UCD of the 
field! Jonathan's main argument for the second kind of mapping  is  the 
necessity of generalization and abstraction for families of datasets. I think 
this could be achieved by deriving the actual datamodel classes from generic 
or abstract classes that several models can have in common, rather than 
defining "abstract" attributes which get flesh only by assigning a value to 
their ucd.


   One last word about Data Models , XML and VOTable in answer to Roy, Jonathan,
and Brian.
   I desagree with this opposition between VOtable fans and DataModel/pure
XML fans.
   The model behind VOtable is the relational one, not the "flat" one.
It does'nt prevent to code some kind of structure.

   So the relationship between VOTable and datamodels is the same kind of   
relationship an object oriented programmer encounters while dealing with
a relational DBMS. Utype is only one of the available tools for this
kind of mapping. Mireille Louys and I have written a note about that to appear
next week.

François Bonnarel
(after discusssing Jonathan's proposal with Mireille Louys)
   

SAUVONS LA RECHERCHE :  <http://recherche-en-danger.apinc.org/>

=====================================================================
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
---------------------------------------------------------------------