Obscore 1.1 errata page

[Laurent MICHEL]

Markus,

First of all: Happy New year DMers!

I'm sorry to insist, but the purpose of the errata is to make UCDs 
consistent with the field definitions, not to change the definition 
themselves.

obs_publisher_did and publisher_id are both defined as IVOA identifiers:

*Section 4.5* Publisher Dataset Identifier (obs_publisher_did)
The obs_publisher_did column contains the IVOA dataset identifier 
(Plante and al. 2007)

*B.4.2.* Publisher Identifier (publisher_id)
The IVOA ID for the data provider as defined in the Spectrum DM.

The publisher_id definition references spectrumDM furthermore.

My personal comment on the substance is that multi-purpose columns is 
not a good thing for a table used for data discovery. This would require 
clients to run inferences on values to guess what they really are 
(IVOAID, ORCID, ...)
This would turn back to a situation similar to access_url for which 
clients have no way to guess whether the URL refers to a data product or 
to a datalink.

Cheers
LM

Le 20/12/2018 à 10:55, Markus Demleitner a écrit :
> Dear Laurent,
> 
> On Wed, Dec 19, 2018 at 02:22:26PM +0100, Laurent Michel wrote:
>>> On Wed, Dec 19, 2018 at 10:48:42AM +0100, Laurent Michel wrote:
>>>> 1) OK, I'll split the page in 2 errata
>>>> 2) Is your footnote [1] suggesting to go back from meta.ref.ivoid
>>>> to meta.curation;meta.ref.uri?
>>
>> To me, it's an important change in the model.
>> Those 2 fields are clearly defined (in the text) as being IVOA identifiers not URIs.
>> Modifying this is a bit more than fixing an UCD inconstancy, this changes
>> the scope of these 2 identifiers (obs_publisher_did and publisher_id)
>> I would say it's another topic.
> 
> 
> Well, I'd say the move from meta.ref.uri;meta.curation to
> meta.curation;meta.ref.uri is smaller that going all the way to
> meta.ref.ivorn.
> 
> 
> On the other hand, I'm positively surprised that obs_publisher_did in
> practice actually contains IVOIDs in the services that have them
> (look for yourself:
> 
> import pyvo
> 
> for svc in pyvo.regsearch(datamodel="obscore"):
> 	try:
> 		res = svc.service.run_sync("SELECT TOP 1 obs_publisher_did"
> 			" FROM ivoa.obscore WHERE obs_publisher_did IS NOT NULL")
> 		if res:
> 			print(res[0]["obs_publisher_did"])
> 	except:
> 		pass
> )
> 
> Incidentally, for publisher_id, this script hasn't found an non-NULL
> ones at all in the current VO -- which makes me suspect that no
> client will have a look at the column in the first place, which
> probably means we could operate with impunity here.
> 
>           -- Markus
> 

-- 
---- Laurent MICHEL              Tel  (33 0) 3 68 85 24 37
      Observatoire de Strasbourg  Fax  (33 0) 3 68 85 24 32
      11 Rue de l'Universite      Mail laurent.michel at astro.unistra.fr
      67000 Strasbourg (France)   Web  http://astro.u-strasbg.fr/~michel
---