DAL Running Meeting #19 - Datalink and multiple locations

[BONNAREL FRANCOIS gmail]

Dear all,
    I will not attend the meeting, but here are my 2 cents on the 
questions raised

    a ) DAL at IRSA :
         - in Obscore there is a distinction between "observation" 
identified by obs_id and the individual unit "dataset" identified by 
obs_publisher_did
This allows to group by a common obs_id several datasets considered to 
belong to the same "observation". The definition of what is an 
observation is actually provider dependent. SO in the mapping with CAOM 
I guess that planeid is more aligned with obs_publisher_did and the 
dataset concept.
         - for "fast" DataLink as far as I understand the question, the 
ID parameter  in the DataLink service is not required to be 
obs_publisher_did, so any more direct index to the CAOM artifact table 
can be used there.

     b ) How should ObsTAP+DataLink provide options to retrieve the same 
product from multiple locations (including cloud)?

              - prefered provider : the DataLink access is not required 
to be given by the access_url FIELD (or equivalent) in a data discovery 
response. It can be provided through a DataLink service descriptor as a 
secondary RESOURCE in the respeonse. In that case the access_url can 
provide direct link to the preferred provider.
              - multi provider issue : with the current standard, I 
think providing the provider tag in the local_semantics is the "less 
bad" solution if not the better. local_semantics is a free text FIELD, 
but should repeat the same strings for different datasets in the main 
service in order to identify related or similar raws in the DataLink 
response. So something like
                           ID = id1
                           semantics : #this
                           content_type : application/fits
content_qualifier : image
local_semantics : aws
                           ............
                           ID = id1
                           semantics : #this
content_type : application/fits
content_qualifier : image
local_semantics : google
.............
                           ID = id1
                           semantics : #progenitor
content_type : application/fits
content_qualifier : cube
local_semantics : aws
.............
                           ID = id1
                           semantics : #progenitor
content_type : application/fits
content_qualifier : cube
local_semantics : google
.............
                           ID = id1
                           semantics : #progenitor
                           content_type : hdf5
                           content_qualifier : cube
                           local_semantics : aws
                           .............
                           ID = id1
                           semantics : #progenitor
                           content_type : hdf5
                           content_qualifier : cube
                           local_semantics : google

and then the same kind of answer for ID = id2

will allow the client : to find out all accesses to the full (image ) 
fits format dataset served by aws (or google or ...)
                                 to find out all accesses to the (cube) 
progenitor in fits (or hdf5) served by google (or aws ...)
etc....

Cheers
François




Le 09/12/2024 à 11:10, Grégory Mantelet via dal a écrit :
> Dear DAL members,
>
> Here is an announcement for the next DAL Running Meeting.
>
> Anyone who recently attended the DAL session in Malta may remember the 
> first
> talks given by Anastasia Laity (pdf 
> <https://wiki.ivoa.net/internal/IVOA/InterOpNov2024DAL/DAL_at_IRSA_Laity.pdf>) 
> and Tess Jaffe (pdf 
> <https://wiki.ivoa.net/internal/IVOA/InterOpNov2024DAL/NAVO_cloud_DataLink_202411_interop.pdf>). 
> They raise an interesting
> question about ways to convey multiple locations for a single product 
> provided
> through Datalink. This question was raised and Tess suggested some 
> solutions ;
> see session notes 
> <https://wiki.ivoa.net/internal/IVOA/InterOpNov2024DAL/IVOA_Nov24_DAL.pdf>. 
> Unfortunately we did not have a lot of time for discussions.
> Hence this DAL Running Meeting.
>
> If you are interested by this question, here is the information about this
> Running Meeting:
>
>     *Date    :*Dec 11, 2024
>     *Time    :* 04:00 PM UTC
>     *Duration:* 1 hour
>     *Location:*
>     https://cnrs.zoom.us/j/92095660540?pwd=3dAHll6Na9X71S8HAgUnNtxzjo3r9v.1
>     /*Meeting ID:*//920 9566 0540/
>     */Passcode/**/  :/*/sjM0h4/
>
>
> Cheers,
> Grégory Mantelet & Joshua Fraustro

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mail.ivoa.net/pipermail/dal/attachments/20241211/19aff838/attachment.htm>