[Proj] MapInfo/Oracle Cylindrical Miller Parameters

[Patrick Mézard]

Hello,
In search of a good-looking world projection I came across Miller
Cylindrical one. Fortunately, both MapInfo and Oracle Spatial know about it,
and even better agree on transformations from WGS84 to Miller for several
points. This is important because it is always easier to debug something
when the input or output can be reproduced on other tools. Now I would like
to use it in proj.4 but as usual I am stuck trying to figure out *the
parameter string*.

I am taking the Oracle WKTEXT description as reference:
"""
PROJCS["Miller Cylindrical", GEOGCS [ "NAD 27 (Continental US)", DATUM ["NAD
27 (Continental US)", SPHEROID ["Clarke 1866", 6378206.400000, 294.978698]],
PRIMEM  [ "Greenwich", 0.000000 ], UNIT ["Decimal Degree",
0.01745329251994330]], PROJECTION ["Miller Cylindrical"], UNIT ["Meter",
1.000000000000]]
"""

It converts (2.35, 48.85) into (261360.667, 5916639.52).

NAD27 datum is defined by:
"""
NAME = NAD 27 (Continental US)
SHIFT_X =  -8
SHIFT_Y =  160
SHIFT_Z = 176
"""

Oracle WGS84 and Clarke 1866 ellipsoids definitions are the same than
proj.4ones.

and finally tried something like:
"""
cs2cs +proj=latlong +datum=WGS84 +to +proj=mill +ellps=clrk66
+towgs84=-8,160,176 +lon_0=0 +units=m
2.35 48.85
261360.67       5889006.49 -12281.453
"""

So, cs2cs output x coordinate perfectly matches Oracle's one, but the y
differs by 27633m.
I tried to replace the Clarke ellipsoid with a sphere defined by the minor
then the major axis but it does not improve the result.

Just to be clear: I do know that configuring Miller projection with a WGS84
datum would give me usable results. What I am looking for is compatibility
with existing applications, to ease debugging and allow release to third
parties without complicated explanations about why this is not the same
thing than the one they see in their GIS dialog boxes while having almost
the same name.

What do you think about this? Are these discrepancies to be expected because
of differences within projection implementations? Did I miss something when
configuring the projection?
Thank you for any idea.

--
Patrick Mézard
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.osgeo.org/pipermail/proj/attachments/20060112/314b333b/attachment.html>