Doubts with the facility name term

[Markus Demleitner]

Dear Semantics folks,

On Thu, Feb 03, 2022 at 02:20:46PM +0100, Tamara Civera wrote:
> I am writing to this group, because trying to unify the facility name in the
> different VO services offered by my observatory (SIAP, Obscore,...) the
> following doubt has arisen: *Should the facility name describe only the
> facility or observatory, the telescope or both?*
> 
> According to the different VO protocols documentation:
> 
> Obscore: "B7.1 The Facility class codes information about the observatory or
> facility used to collect the data."
> 
> Obscore: "C.3 <FIELDname="facility_name"datatype="char"ucd="meta.id;instr.tel"utype="obscore:Provenance.ObsConfig.Facility.name"xtype="adql:VARCHAR"arraysize="128*"><DESCRIPTION>telescope
> name</DESCRIPTION></FIELD>"
> 
> SIAP 2.0: "2.1.12 The FACILITY parameter is a string-valued parameter that
> specifies the name of the facility (usually telescope) where the data was
> acquired. The value is compared with the facility_name from the ObsCore [7]
> data model"
> 
> VoDataService: "3.1.1 The observatory or facility used to collect the data
> contained or managed by this resource."

Tamara raises a very valid point here, one that I've often briefly
puzzled over only to just move on and ignore the problem.  

I think that if we want these fields to be reliably usable, we ought
to make up our minds (and then perhaps clarify the various standards
using errata).

As usual, I'd suggest we start by working out what functionality
should be enabled by the feature; historically: what motivated the
addition of facility_name in Obscore and VODataService (and,
conversely, what made the SSAP authors to only have instrument).  If
anyone reading this remembers, please chime in.

If you ask me, I'd say the main conceivable use case would be queries
-- whether for data sets as in obscore or for data collections in the
Registry -- of the type "Give me data from the HST" (which then would
include any sort of detector; my impression is that there's
relatively wide consensus that that would be in the instrument
field).  This, however, gets more complicated when there are multiple
telescopes at one site (facility?).  Let's take ESO's La Silla as an
example, where (I think) there are about a dozen individual
telescopes (or telescope-like... things, to avoid the term
"instrument" that we've just used for detectors).  

Which reminds me: The FITS standard
<https://heasarc.gsfc.nasa.gov/docs/fcg/standard_dict.html> has the
TELESCOP keyword for pretty much this -- does anyone remember why we
didn't just take that over?  Was it because we wanted to cover
devices that don't feel like telescopes at all like, perhaps the GBM
aboard Fermi, the Icecube-like neutrino detectors, or LIGO/Virgo?)

Resuming the La Silla example, I'd venture giving the concrete
telescope ("ESO 3.6 m", "NTT", ...) will more likely be useful for
actual searches than the generic "La Silla", as the individual
telescopes have so different properties that it seems unlikely to me
that a scientifically relevant question could use them in grouped
form.

Against that could be a situation in which someone remembers that
some data set they've seen, say, at a conference, was "taken
somewhere at La Silla".  That, for one, seems a less likely scenario
to me; also, if and when we have usable
dictionary/vocabulary/taxonomy of "instruments" (now in the general
meaning), it would be fairly straightforward to translate such
a query into a query based on the individual telescope names.

After this, I'm frankly tempted to try and clearly state that
"facility" should be, to first order, the telescope, and we ought to
explictly link it to FITS' TELESCOP.  Which, by the way, is defined
as

  The value field shall contain a character string identifying the
  telescope used to acquire the data associated with the header.

I can't say I'm too happy with using "telescope" in this definition
as our "collectors" are becoming more diverse and exotic.  In a way,
I'd be tempted to try and separate facility_name and instrument into
"device that collects/concentrates messenger particles" and "device
that detects messenger particles".  For Icecube-like devices, that
would happen to coincide, but that's perhaps not problematic.

Well, having said all this: I suspect there's a bit of prior art on
this.  If you have pointers, by all means share them.  And would
anyone volunteer to run a discussion on this problem at the Interop?

            -- Markus