[DataSet Data Model WD review ] comments on version 1.0 20150302

[Louys Mireille]

Dear all,

It seems that this email below did not appear on the dm list last July, 
probably an error on my side . Sorry .
Mireille
_________________________________________________________________
I gave a second read to the Dataset DM specification and have a few 
comments to share, following the pages sequence:

#Acknowledgements : that would be suitable to acknowledge the EURO-VO 
projects :
replace : and from the European commission 6 ...(former program) by:
from the EURO_VO projects (European Commission 7 th program) :
Euro-VO Aida, VO-ICE and CoSADIE.

#Typo : p 10. reads Furture revisions instead of Future.. .

#3.3 Derived Class :
This class gathers the various properties built up from the analysis of 
the dataset.
It was relevant for the Spectral DM to store the redshift, varAmp for 
the source, etc. or estimated signal to noise ratio that should be 
documented precisely.
This object cannot be generalized for any data set, and can remain just 
as a place holder.

#4. Observation-Experiment
This section gathers the primary stage of the Observing process with a 
summary for the Observing Configuration. we could mention here that a 
more detailed IVOA Provenance DM is currently developed and assessed 
(CTA project, GAVO, others?).

#STC2 compatibility
The current version of STC2 in the Dataset DM currently defines a 
Mapping class which encapsulates coordinates transforms.
This is is desirable to identify clearly this Mapping in order to use it 
at the protocol level.

#5.2 Datatypes
I would suggest to add:
Dataset Datamodel gathers and homogeneize data type definitions defined 
in previous specifications like ObsCore(data product), 
VOResource(RightsType) and VODataservice (SpectralBand).
I would prefer SpectralBandType for symetry with other definition.

#5.2.2 Creation type
It seems we are trying to express different concepts here : whether the 
data are static and already archived : 'archival' or dynamicaly 
generated on one hand, and on the other hand, what was the computation 
processed followed to build up this dataset: cutout , filtering , 
binning, mosaicing, etc ... with brings more an idea of the 
'computational Provenance'.
These two ideas should be separated.

#5.2.4 Redshift type with the confidence attribute is not general enough 
to be a type, according to me. STC Redshift class should be used, and 
depending on the application context derived for specific needs.

I think the current document gathers clearly the common concepts worked 
out in various existing models and stitches all in a consistent manner. 
This is a useful and valuable synthesis effort.

Best wishes , Mireille.

-- 
Mireille Louys	, Maître de conférences
Centre de Données ( CDS)		Icube & Télécom Physique University of Strasbourg