[MetaCRS] "FOSS Maintained" Source of CRS Definitions

[Martin Desruisseaux]

Frank Warmerdam a écrit :
> Can you point us to a public document on this model?  I'm frankly somewhat
> intimidated by it, and concerned that a lot of stuff it describes is not
> practically useful.  But I really need to do a current review before I
> spout off much more.

The ISO 19111 specification is not free. However their UML schemas are available 
for free (Rational Rose format I believe):

     http://www.isotc211.org/HMMG/

You can also look at the GeoAPI javadoc, which is derived from ISO 19111 (and 
copyrighted to OGC, so they sort out the copyright issue between OGC and ISO):

     http://geoapi.sourceforge.net/snapshot/javadoc/index.html

Just looks at the 5 packages starting by "org.opengis.referencing" on the top of 
the page and ignore the remainding. Everything derived from ISO 19111 is clearly 
flagged with a @UML annotation. Example: if you look at CoordinateReferenceSystem:

http://geoapi.sourceforge.net/snapshot/javadoc/org/opengis/referencing/crs/CoordinateReferenceSystem.html

You can see at the top of this class the following annotation:

     @UML(identifier="SC_CRS", specification=ISO_19111)

This means that this interface is derived directly from ISO 19111 and is called 
"SC_CRS" in ISO (we renamed it in GeoAPI in order to comply with Java usage, but 
never do that without telling what is the official ISO identifier in an 
annotation that developpers can parse if they wish).

If you scroll down in this class, you will see this @UML annotation repeated for 
methods. If a particular methods do not derive from ISO 19111 but is an addition 
that we did ourself in GeoAPI, it is clearly flagged by an the following annotation:

     @Extension

This model is the one that we implemented in GeoTools. I have experimented both 
the legacy OGC 01-009 model and the ISO 19111 one (I initially implemented OGC 
01-009 before I refactored everything as ISO 19111). I believe that this model 
is very good. Maybe incomplete (this is why GeoAPI has @Extension), but the part 
covered by ISO 19111 is very good in my opinion.


> I'd be interested to hear from the CS-MAP folks if they model datum
> shifts between arbitrary datums, as opposed to mapping everything through
> WGS84 (as PROJ4 is inclined to do).

I can said that GeoTools models datum shift between arbitrary datum. WGS84 is 
not treated especially except during WKT parsing and formatting.



> A few limitations to WKT that I am concerned about:
>  * There is no way to express datum shift methods other than 3/7 parameter
>    transformations to wgs84 (via TOWGS84).  In particular, how am I
>    supposed to document use of grid shift files in WKT?
>  * There is no way to express that there are more than one possible
>    datum shift approximation available for use.
>  * No way to mark coordinate systems as deprecated, categorize them for
>    display to the user and such.  If we produce a "super dictionary" it
>    would be nice if user interfaces could provide a useful navigation
>    mechanism to user based on it.


Can resit to the temptation to point out that EPSG database support of that :).


> I can imagine picking the ISO model as our central CRS data model, and
> distributing definitions in some format that maps well to that model, like
> GML, but I'm concerned we make our dictionaries so complicated that they
> become very difficult to exploit.

If we peekup the ISO model, we peekup a database schema quite close to EPSG. The 
EPSG database is not exactly the same than ISO, but is pretty close (same 
concepts, just a few different names).

	Martin