Draft note on STC in the Registry

[Markus Demleitner]

Hi TDIG, Hi Registry,

On Mon, Jan 29, 2018 at 04:51:21PM +0100, Ada Nebot wrote:
> I can see that the idea is to combine spatial, wavelength and
> I would like to be sure that discovery for the temporal axes is not
> limited to day but that there will be the possibility of making a
> finer query, say down to the hour or the minute. 

I suppose at least 10 minutes shouldn't be a problem.  Below this
will be, I suspect, a bit of a challenge for many places because most
metadata on images and spectra will be topocentric for the
observatory (or something like that), and transforming that to
barycenter just for coverage calculation sounds like something we had
better not ask for lest too many people do nothing at all.

Keep in mind, though, that this is for *service* discovery.  The
discovery of the actual dataset is then another step.  For this, as
far as I can see, Obscore currently doesn't say anything about
reference position (introducing systematic errors of order 10 minutes
between earth and barycenter) and time scale (jitter of order 10
seconds, say between TT and UTC) in t_min, t_max.

I'd say that this, Obscore, is what we'd need to touch when we want
*dataset* discovery at a precision level below eight minutes.

What I'd like to make clear for the Registry is that reference
positions like Pluto or the galactic center are out.  Whether we can
keep people from using Earth rather than the Barycenter... ah well.
We should politely ask them to use Barycenter, but... full
disclosure: not even my sample records do.

> Section 4
> Further axes? As you mention one could add more axes, but something
> like redshift would already impose questions like: how was it
> calculated? Photometrically or spectroscopically? Personally I
> think this might complicated things. 

I'd say that discovery use cases don't really depend on the
provenance of the redshifts, and it wouldn't be hard to fix the
doppler definition (probably to RELATIVISTIC in STC 1.13 lingo).
But I'd say the discovery cases really are about distances rather
than redshifts as such.  And hence what I'd really like to see are
creative ideas how we could achieve discovery by distance given a
dynamic range starting at the upper edge of the earth's atmosphere
and ending at redshift 1500 (possibly more if we go beyond the
electromagnetic spectrum).

My take: We can always add tables once we better understand what we
want.  Doing away with tables we got wrong is a lot harder.  Let's
postpone this.

> 
> Other ref. systems? Indeed for SSO or moving objects it might not
> be enough. As minimal requirements for time series data we included
> a target name field. For SSOs we should follow the IAU convention,
> I found this https://www.iau.org/public/themes/naming/#inss
> <https://www.iau.org/public/themes/naming/#inss> 

I strongly suspect nothing saves us from having to enumerate what we
want to allow, probably more or less in the style of STC-1.13 (though
perhaps a bit less comprehensive, and PLUTO is now officially out of
the game).  

Be that as it may, I take from the feedback so far that we want
evolvability in this direction, and so I'll just add a column for the
reference system.  I'll try ref_system_name and say it's NULL for
now, and discovery queries should constrain it to NULL to ward
against further development.

> Non electromagnetic coverage? Again, we thought about this in the
> time domain context and it might be useful to add Neutrino and GWs.  

Yes... but how?  They certainly don't really fit into rr.stc_spectral
as it's written now.  If we changed stc_spectral to

ivoid   energy_low  energy_high  species

(with species coming from a controlled vocabulary) I could even see
discovery cases that could leave species unconstrained.

The question is: How much will people hate it if we change the beloved
wavelength to energy?  And is it worth it given that for the
forseeable future, telling apart various bands in neutrinos or, gosh,
gravitonal waves probably won't be common discovery cases.

The alternative are simple keywords (in rr.res_subject).  If resource
providers agreed on using Neutrino and Graviational Waves (or
whatever else) there (and reliably), I guess that'd nicely cover
discovery for the coming decade.

Any further feedback welcome (but I'd say depending on topic to go
either to TDIG or to Registry),

         Markus