[Proj] ED50 from C API

[Frank Warmerdam]

Roger Oberholtzer wrote:
> On Thu, 2007-03-29 at 09:05 -0700, Eric Miller wrote:
>> I don't know about the "esri" file, but you have two different ellipsoids and no information to convert between the different models.  Don't you need a datum or towwg84 parameter for the ED50 definition?  I think the ED50 towgs84 parameters are missing because there is more than one candidate.  See, http://crs.bkg.bund.de/crs-eu/
>>
>> There isn't really much difference between using cs2cs and pj_transform.  The init strings are the same.  Same for proj and pj_fwd/pj_inv.
> 
> I have since tried a few ED50 definitions to see which is closest to the
> target locations:
> 
> "+towgs84=-87,-98,-121 "     // "EUROPEAN 1950, Mean"
> 
> "+towgs84=-87,-96,-120 "     // "EUROPEAN 1950, Western Europe"
> 
> "+towgs84=-84,-107,-120 "       // "EUROPEAN 1950, Portugal &
> Spain" (best)
> 
> "+towgs84=-86,-98,-119 "     // "EUROPEAN 1979"
> 
> The best is "+towgs84=-84,-107,-120 ", which gets me to within 5 meters.
> Still too far, but better.
> 
> It still remains a mystery to me how anyone actually uses the esri file
> when it is quite often incomplete. To the point it does not work at all.
> Especially for a conversion as 'big' as one used for all roads in Spain.

Roger,

I'm afraid I haven't been following this thread closely.  Is this incomplete
in the sense of missing the datum shift, or missing the actual projection
definition?

The "esri" file was a bulk translation I did some time ago and is not,
to the best of my knowledge, widely used.  Nor have I done much review
of it.  There may have been errors in translation beyond the fact that
the bulk translator doesn't know how to handle some projections.

> Perhaps proj needs its own file that contains complete working
> definitions? proj is a great library. But without proper definitions is
> is less useful than it could be. I understand the complexity of the
> thing since so many things can be converted into so many things,
> resulting in a zillion combinations. But there must be something
> possible. Or is the solution staring me in the face and I just don't see
> it? Do all users of proj make all the definitions locally (read
> "reinvent the wheel")?

It has several files with definitions, but it is important to be able
to relate coordinate systems in PROJ with coordinate systems in other
software, or according to well known standards like EPSG.

And there is generally no single agreed on best datum shift solution
for a given datum so it is hard to handle this in an authoritative way.

I share your frustration, but PROJ trying to produce an authoritative
and complete set of coordinate systems on it's own is not practical!

I do think anyone wanting precision in transformations from an unusual
datum to WGS84 needs to be very aware of what datum shift parameters
are being used.

Best regards,
-- 
---------------------------------------+--------------------------------------
I set the clouds in motion - turn up   | Frank Warmerdam, warmerdam at pobox.com
light and sound - activate the windows | http://pobox.com/~warmerdam
and watch the world go round - Rush    | President OSGeo, http://osgeo.org