[gdal-dev] EPSG:2065 / South-oriented Krovak support (Czech friends thoughts wanted !)

[Even Rouault]

Hi,

I'm having a look at the following all related issues :
https://trac.osgeo.org/postgis/ticket/1940
https://trac.osgeo.org/proj/ticket/185
https://trac.osgeo.org/gdal/ticket/4762

related to EPSG:2065 support, ie "S-JTSK (Ferro) / Krovak". There's an issue 
about the datum shift used, but that's an easy one compared to a more serious 
one regarding axis order.

The proj.4 krovak method expects the first value of a coordinate pair to be a 
easting, and the second value a northing. Whereas original Krovak (EPSG:2065) 
is supposed to be southing first, westing second.

https://trac.osgeo.org/gdal/ticket/4762 suggests to add a "+to_meter=-1" 
parameter. Which means that the coordinates supplied are first westing, second 
southing (instead of "+to_meter=-1", a cleaner equivalent way is to use 
"+axis=wsu", meaning westing,southing,up )
But EPSG would mandate southing,westing.

But admitedly, for less exotic coordinate systems with official EPSG axis order 
latitude,longitude (EPSG:4326...) or northing,easting (some Finland projected 
CRS for example) we have never introduced in GDAL WKT or proj.4 strings the 
+axis option, and the application/user does the coordinate switch itself 
before passing that to GDAL/proj.

So, does that sound OK that a user providing coordinates advertized to be in 
EPSG:2065 does that in westing, southing order ? Or was this something 
particular to this user ? (I'd note that the dataset mentionned in the ticket 
is a GML resulting from a WFS server, and we all know the axis ordering issues 
related to WFS)

What a mess... 
(Using EPSG:5221 which is the easting, northing variant avoids those 
headaches)

Even

-- 
Spatialys - Geospatial professional services
http://www.spatialys.com