[josm-dev] projection Mercator, set EPSG 3785 instead of EPSG 3857

[Tobias Wendorff]

Am 08.03.2011 16:07, schrieb Josh Doe:
> Okay, first of all the proj.4 parameters you gave aren't entirely
> correct. If you look in the epsg file from proj.4 that I linked to in
> my earlier message, you'll see that the parameters you claim for
> EPSG:3785 (sphere) are in fact what proj.4 uses for both EPSG:3785 and
> EPSG:3857. That's not entirely correct either. If you look at the EPSG
> definition for 3785 that Martin linked to earlier, you'll see that the
> _datum_ is spherical. The proj.4 parameter "+nadgrids=@null" (as I
> linked to earlier) forces proj.4 to NOT convert between the WGS84
> (ellipsoidal) and spherical datums. This is all moot since EPSG:3785
> is invalid and should never be spoken of again. :)

What do nadgrids have to do here? The transition from an ellipsoid
to a sphere can be solved without shift files. Is it a bug of proj4?
I've just dropped +nadgrids=@null: it doesn't change anything on the
results.

EPSG:3785 = uses a sphere with Mercator projection (JOSM is using this)
EPSG:3857 = uses an ellipsoid with Mercator projection (we want this)

> Also, when you talk about using the inverse transform, you're
> incorrectly using a different set of coordinate systems to transform
> between. An application must ensure it is using the correct pair of
> transforms to ensure no errors are introduced. You're trying to
> transform with [1], then back with [3], while you should be going back
> with [4]. Transform [3] should be paired with [2].

But that's what JOSM does! Have a look at JOSM's sourcecode:
http://josm.openstreetmap.de/browser/josm/trunk/src/org/openstreetmap/josm/data/projection/Mercator.java

That's the spherical formular. When you draw a point in JOSM (x,y)
it gets translated back to the *sphere*, not to the ellipsoid.
This spherical information will be stored to the WGS84-database.

JOSM identifies its Mercator projection as 3857, but it isn't...