[Proj] Gridshifts give slightly wrong results

[Paul Kelly]

Hello Uwe

On Wed, 7 Mar 2007, Uwe Schmitz wrote:

> Hello,
>
> I've upgraded from proj-4.4.7 to 4.5.0 and
> notice a slight difference of my transformation
> results when grid shifts are involved.
>
> Executing the command:
> cs2cs -f "%.12f" \
>      +proj=latlong +ellps=bessel +nadgrids=./BETA2007.gsb \
>   +to \
>      +proj=latlong +ellps=GRS80
> 7.483333333333E 53.500000000000N
>
> results in
> 7.482506019176  53.498461143331 0.000000000000   # 4.4.7 correct!
> 7.482506019176  53.498461144236 0.000067942776   # 4.5.0
>
> Through debugging I found out, that now (V4.5.0) there
> is an additional step converting the geodetic coordinates
> to geocentric an back again. This is caused by the different
> ellipsoids.

What happens if you add +towgs84=0,0,0 to the second projection 
definition? That might stop it trying to do an ellipsoid conversion. Have 
to admit though that I am quite hazy on how this works in PROJ.

> IMO, this is a bug, because the datum shift is completely
> done by adding the shift values.

Is this relevant? http://bugzilla.remotesensing.org/show_bug.cgi?id=368
Particularly Frank's comments right at the end?

Paul