Data from a scan project initiated by the late Kurt Birkle.

The raw scans come down from Landessternwarte (talk to Otmar Stahl,
o.stahl@lsw.uni-heidelberg.de).  They are calibrated using
bin/calibrate.py (which uses astrometry.net).

Every now and then a plate catalogue comes in from Landessternwarte.
Then do:

 * copy the file to res/, following naming conventions.
 * link to it from plattenkatalog.csv
 * run bin/calibrate.py
 * run bin/calibrate.py --verbose --report
 * if plates are not solved, try to find out why (see processors docs)
 * gavoimp res/positions
 * send bin/calibrate.py to whoever sent you the plate catalog

TODO: Add header information as per

https://www.plate-archive.org/wiki/index.php/FITS_header_format

Mainly as a proving ground for https://www.ivoa.net/documents/BibVO,
we've collected a few publication -> plates links in
res/bibliography.txt. Perhaps this can be expanded.  If you'd like to
tackle this, the format is <bibcode> <plateid>+.

The calibrations of several plates are so severely off that the polygons
end up non-convex.  I for now removed the coverages from the database
using:

update lsw.plates set coverage=NULL
where accref in (
'lswscans/data/part2/Wolf/FITS/G1970.fits',
'lswscans/data/part2/Wolf/FITS/G3315.fits',
'lswscans/data/part2/Wolf/FITS/G3318.fits',
'lswscans/data/part2/Wolf/FITS/G3183.fits',
'lswscans/data/part2/Wolf/FITS/G3354.fits',
'lswscans/data/part2/Wolf/FITS/G5079.fits')

Also, the calibrations of all Bruce Plates are wrong if the
circumference/area of the resulting coverage isn't somewhere between
0.018 and 0.037

update lsw.plates set coverage=NULL where accref in (
select accref from (
select accref, ar/cf as rat from (
select @-@coverage as cf, area(coverage) as ar, accref
from lsw.plates
where  plateid like 'B%') a) b
where not rat between 0.018 and 0.037)


Instrument identifiers
----------------------

For a test with b2find and b2inst, I've created "P"IDs for the
telescopes here.  I did that on https://b2inst-test.gwdg.de/
The login there is Markus' uni-id with some custom password (don't ask).

These people want a lot of metadata, none of which really apply to our
case.  I liberally used DataCite NULL values (:unas, :unap) for them.
Well, and then I still don't know how to pull the generated identifiers
from them.