WFS layer projection problem

[Martin Weinelt]

On Monday 14 March 2005 16:29, Frank Warmerdam wrote:
> On Mon, 14 Mar 2005 16:09:07 +0100, Martin Weinelt
> <mweinelt at planiglobe.com> wrote:
> > From the logs of the WFS-server we can see, that the mapserver
> > CGI sometimes sends strange bounding boxes to the WFS, which
> > do not really cover the area shown by the map.
> >
> > Is there anything special with LAEA and mapserver (proj4). Has anyone
> > encountered s/th similar? Where should we start searching for
> > potential solutions to this problem?
>
> Martin,
>
> There are many challenges in global scale mapping with MapServer or
> many other GIS packages.  In particular, spatial queries are normally
> represented as a bounding box.  If you reproject a bounding box
> from a projection like LAEA, near the pole, to decimal degrees, you can
> encounter many quirky situations where the resulting rectangle does
> not accurately represent the same area of the earth due to the
> different natures of the two coordinate systems (for instance the specialness
> of the north pole in geographic coordinates).
>
> There is a mapserver wiki topic discussing some of the challenges of
> global scale mapping, though unfortunately we can't add to it anymore.
>
> I don't have any useful suggestions on how to resolve the problem for
> your particular case.  The proper solution would *likely* be to make
> the MapServer bounding-rect reprojection function know alot more
> about the nature of coordinate systems.  But that is a hairy area.
>

Frank,

thanks a lot. I visited some of the 'hairy areas' when I wrote a JS routine
for inverse projection from LAEA and mercator which are part of the
interface. For instance, Snyder states that the inverse formula for LAEA
my return undefined values for specific coordinates, but he at the same time
says that you should return a value of so-and-so in this case.

I am still not sure to what extent such semantic 'rules' are implemented
in the projection routines of mapserver. Besides, a LAEA-box - as you say -
is not a box at all 'back' in DD. It's not a box in the first place anyways, it's
four points ...

We may do better with mapscript where we might be able to implements some
of the sematics needed for our purposes, but then - mapserver calls itself
a WFS-client to the specs of the OGC ...

Thanks again, at least for reassuring and your answering at the speed of light ..

Martin