[mapserver-users] Projection problems

[Jan Hartmann]

On 01/07/11 20:57, Frank Warmerdam wrote:
> On 11-01-07 02:40 PM, Jeff McKenna wrote:
>> On 11-01-07 10:49 AM, Jan Hartmann wrote:
>>>
>>> Yes, the transformation to WGS84 for the Dutch coordinate system
>>> ("Amersfoort", "Rijksdriehoeksnet) is still *not* part of the latest
>>> PROJ4 release (4.7.0, downloaded yesterday).
>>
>> I have requested many changes to the EPSG 'geodetic dataset' 
>> throughout the
>> years, so maybe I can give you some advice:
>>
>> - enter your change request through a form on the EPSG site:
>> http://www.epsg.org/Comms/Comment.asp
>> - it may take upto one year for the change to be accepted by the EPSG 
>> group and
>> included in the next EPSG geodetic dataset
>> - you will get an email from the EPSG group once that happens
>> - however it will take much longer for software that uses EPSG codes 
>> to update
>> their own releases with those new EPSG codes (this includes PROJ, 
>> QGIS, MS4W,
>> etc.)
>>
>> Hope that sheds some light on the issue for you.
>
> Jeff,
>
> I think Jan's concern is that the transformation does exist in EPSG but
> is not reflected in the last PROJ.4 release.  I believe this is fixed
> in CVS and I really just need to push out a new PROJ.4 release reasonably
> soon.
>
> Jan - I'd appreciate it if you could confirm that:
>
>   http://svn.osgeo.org/metacrs/proj/trunk/proj/nad/epsg
>
> has the right value.
>
> Best regards,

Hi Jeff, Frank,

Yes that was indeed the issue. The problem is that the EPSG database 
*does* have the WGS84 transform for EPSG 28992, but gives a whopping six 
possibilities, without offering a good explanation what they amount to. 
That's why the automatic retrieval by PROJ of those parameters from the 
EPSG database doesn't work too well for this EPSG number.

The parameters in SVN are not the most recent ones: see 
http://trac.osgeo.org/geotiff/ticket/22 for the right numbers, and the 
attached spreadsheet for the way they were computed. Both series result 
in almost the same ground positions (differences of a few centimeters), 
so it doesn't matter much in practice whether you adapt SVN by hand or 
not. Of course it would be better to have just one set of parameters 
available everywhere. I have always wondered why there is nowhere an 
official Dutch site were this matter is kept up to date in an 
easy-to-understand way. To translate Milo's Dutch proverb, everyone 
seems to fish in his own pond over here.

By the way, I have the impression that the method by which the WGS84 
shifts are retrieved from the EPSG database and put into PROJ is not 
100% robust, so perhaps this could be looked into too.

I hadn't realised that PROJ is so stable by now that new releases take 
so long to appear. Sorry if I sounded a bit grumpy; to quote Milo again, 
this problem has plagued Open Source GIS people in Holland for years.

Thanks,

Jan