[geotk] Spherical-earth CRS

[Martin Desruisseaux]

Hello Jon

Le 01/04/11 10:41, Jon Blower a écrit :
> I don't think I have lenient=true set anywhere - it was a simple Java test program with a main() method, so unless it's set elsewhere (e.g. in a config file) I'm not sure how this could have been set.

There is no configuration file for this aspect, so I'm not sure neither why no 
exception were thrown... Maybe a matter of Geotk version (but I don't remember 
any change in this aspect).

Just as a side note, I'm exploring OSGi and JSR-330 (which provide Inversion of 
Control) for the next Geotk release (3.18). Hopefully it will make the 
configuration easier, or at least more standard. The fact that a JSR exists for 
Inversion of Control, that containers like Glassfish are JSR-330 ready, and that 
JDK 8 will be compatible with OSGi make all those solutions more attractive.

> I think the CF conventions have only recently added support for specifying the ellipsoid/sphere, and there aren't many files out there in the real world that include this metadata.  So far in the absence of an explicitly-specified datum I've been using WGS84, which is unlikely to be correct for a numerical model (but more likely for observations).  I was wondering if the SPHERE CRS might be a better (or at least less bad!) approximation.  I guess there isn't an easy answer.

I would bet that the SPHERE CRS has good chances to be closer than WGS84... I 
guess that peoples use a sphere instead than an ellipsoid for simplicity. The 
SPHERE constant uses a simple radius (6371000 metres) which is only 7 metres 
away from the GRS 1980 Authalic sphere, and GRS 1980 is itself pretty close to 
WGS 84.

     Martin