Spectral DM 2.0: TCG review feedback

[Michele Sanguillon]

Dear DMs,

I tried to understand if the Pollux database containing metadata about  
synthetic spectra could be Spectral DM 2.0 compliant.
I could read in the wiki that SpectralDM-2.0 has been implemented at  
SVO (TSAP) so I could think that this model concerns both observed and  
theorical spectra.
But when I read the PR-SpectralDM-2.0-20140730, I felt that this model  
only concerns observed spectra (ie §2.1 Illustration 2, the  
multiplicity between Dataset and Target or CoordSys,  
Dataset.CalibLevel, ...).
Could you please give me advices to progress in this task ? Thank you  
in advance.
Cheers,

Michèle


Markus Demleitner <msdemlei at ari.uni-heidelberg.de> a écrit :

> Dear DM,
>
> On Mon, Dec 01, 2014 at 04:37:53PM -0500, CresitelloDittmar, Mark wrote:
>> I think it's OK if any discussion happens on the DM list.
>> I wanted the announcement to be more broad so all concerned were aware of
>> the update.
>>
>> I have, to date, kept the multiplicity out of the heading lines, as they
>> are already in the diagrams and descriptions.
>> If it is to be added, I like how you placed the multiplicity on the
>> attribute "Curation.reference[0..*] : String" rather than on the type
>> "Curation.reference : String [0..*]"
>
> Well, I'm not really concerned with putting this more prominently in
> the headlines, and I believe it'd be typographically, erm,
> unfavourable.  What I'm concerned with is that from the document,
> it's not very clear how to actually write items of multiplicity >1,
> either generically or separately for each format.
>
> One possibility would be to say in the appendix on FITS:
>
>   If multiple values for an attribute need to be represented, append
>   number so each keyword; for instance
>   REFERENCE1 foo
>   REFERENCE2 bar
>
>   Use HIERARCH if the keywords become too long in this way
>
> In VOTable, you might get away with just repeating the params.
>
> Note that I'm not actually proposing this; all effort spent here
> would be better spent ironing out the wrinkles in the general DM
> serialisation within the VO-DML effort, not the least because all
> this totally explodes anyway when there are non-atomic attributes of
> multiplicity >1 (are there in SDM2?).
>
>> Another option, which would be a larger change, could be to reformat the
>> entries to isolate the items.. more in the direction of what VODML
>
> Again, I don't think this is a formatting issue.  As far as I am
> concerned, this is a very fundamental issue in the proposed
> serialisations, and it is a major part of why I'd like to see at
> least examples exercising all the features that are in Spectrum
> DM; writing these up is the only way to figure out spots of
> trouble like the 0..* attribute case.
>
> Cheers,
>
>          Markus


-- 
Michèle Sanguillon
Laboratoire Univers et Particules de Montpellier
Université Montpellier II - Place E. Bataillon - Bât. 13 - CC 072
34095 Montpellier Cedex
Tel : +33 (0)4 67 14 93 59 - Fax : +33 (0)4 67 14 41 90