Le 03/02/11 18:30, John Caron a écrit :
i guess whats missing is a method to convert lat/lon on one ellipsoid to
lat/lon on another?
Close. A method to transform lat/long on one *datum* to lat/lon on another.
Many different datum may exist for the same ellipsoid. The ellipsoid axis length
alone are not suffisient.
(minor note: in ISO 19111 terminology, this is a "transformation", not a
"conversion". A Map projection is a "conversion" but a datum shift is a
"transformation". The difference is that conversions are determined by
mathematical formulas with infinite (in theory) precision while transformations
are determined by empirical parameters with statistical errors).
However there is no generic method for doing the above transformation. Datum
shift require empirical parameters (typically the "Bursa-Wolf parameters", but
other methods are possible) which need to be stored in some kind of database (a
real database like the http://www.epsg.org one, or a text file, etc.).
On a related note, the GeoAPI 3.0 interfaces are now going through the final
voting process (after much delay for reasons that we don't control) in the Open
Geospatial Consortium. ESRI, CNES (a French space agency) and other are among
the submitters. I would suggest to set the NetCDF referencing classes as
implementation of GeoAPI interfaces (I can provide help for that if you wish),
so NetCDF users can start with a relatively "lightweight" referencing
implementation, while other implementations are possible. For example the
Geotoolkit.org referencing module alone (not including dependencies like ISO
19115 metadata) is a 900 kb JAR file, which suggest that referencing stuff may
complex enough to allow the possibility to use exernal libraries.