Datalink vocabulary

[François Bonnarel]

Hi all,
Le 11/09/2014 10:15, Markus Demleitner a écrit :
> Dear DAL,
>
> [taking the discussion from interop to DAL]
>
> On Thu, Jul 31, 2014 at 08:54:43AM +0200, François Bonnarel wrote:
>>      Thanks to Mireille, Laurent, Norman, Pat and Sarah the datalink
>> vocabulary for semantics of the links in DataLink is now available
>> here:
>> http://www.ivoa.net/rdf/datalink/
> One thing I notice is that the "self" thing I've advocated is missing
> (to mark up the link to the primary dataset); however, I think it is
> important, as for some of the existing datalink prototypes, access to
> the full dataset is *only* possible through datalink, and clients
> need a robust way to discover what link to use to retrieve it.
I prefer "something" to "null". and if possible I would prefer 
"retrieval" to "self". I think the usual DAL word for dowloading a full 
dataset (compared to metadata, cutouts-excerpt, or previews has always 
been retrieval
> I'd also like to make the terms more self-contained, as I expect most
> clients will, at least initially, have to work on these exclusively
> rather than have our little ontology.  And of course, neither
> publishers nor users can be expected, in general, to know the model.
> So, they'd not see #preview/image, they'd just see #image, and they'd
> be confused to see what's coming back is a preview.  This, in
> particular, concerns
>
> image (->  preview-image)
> plot  (->  preview-plot)
So any child will be preceded by its parent ? Otherwise it would be a 
mess, no ?
> proprietary (->  ? -- I'm pretty sure rather than give that publishers
>    should be encouraged to use the other categories -- after all, all
>    data conform to *some* data model; so, they should rather say "this
>    is a characterisation" than "this is in some proprietary data
>    model)
>
> I'm also not sure that "characterization" is a good term for char
> data *in IVOA form*.  obscore and imagedm are less generic and thus
> might be ok (although even there I wouldn't mind an "ivo-" prefix).
> I'd much prefer "ivo-c14n" or something like that, to allow people to
> characterize some other form of c14n data.
"model" is probably a bad term for the parent. Usually what we have here 
is some metadata consistent with an IVOA model (or not IVOA)
so "metadata" or "metadata-model" could be better. Characterization is 
jsut ther for full characterization metadata (char 2 aspects which are 
not in obscore) Maybe ImageDM will eventually encompass this

Cheers
François
PS:
>
> Also, without having read the Turtle specs, I'm pretty sure the #
> shouldn't be part of the names in the TTL file; the way this is
> written would imply, I guess, that the full form of these terms would
> be http://www.ivoa.net/rdf/datalink##preview -- which we certainly don't
> want.  Norman?
according to the discussion between Norman and  Pat, i like "#soemthing"
> Otherwise, I like it.
>
> Cheers,
>
>          Markus