[Proj] Roussilhe Projection

[Oscar van Vlijmen]

Mr. Macke wrote: 
> is there a chance that the rouss projection from Gerald Evenden's libproj
> version will be included in proj.4?

By coincidence I was about to comment on that projection too.
I've translated the libproj4 rouss projection to JavaScript and found an
anomaly.
[Sorry mr. Warmerdam, there I go again questioning the code of
proj-libproj-etc or the way they are used!]

Mr. Mugnier gives in his article "Grids and Datums: The Syrian Arab
Republic"
<http://www.asprs.org/resources/grids/09-2001-syria.pdf>
a test point for the Roussilhe projection.

I was puzzled by the fact that libproj rouss apparently doesn't do anything
with the scale factor k0 (m0).
(See also Mr. Evenden's posting dated Wed Apr  7 20:48:19 2004)

Working with the Syria data:
lat=37d; lon=42d9m; lat0=34d12m; lon0=39d9m; x0=0; y0=0;
Clarke 1880 ellipsoid
I obtained with rouss:
x = 267202.985811455, y = 314789.193884167 m
But the article gives:
X = 267078.434 m, Y = 314642.621 m
If I multiply the calculated x and y by the mentioned scale factor of
m0=0.9995341 (note that x0,y0 are 0), then I get:
x = 267078.495940365, y = 314642.533598736
(not all decimals are significant!)
That's much closer. Still 6 and 9 cm off, but closer than the current
libproj rouss function gets [I dare to think].


Has anybody proven testpoints for Poland, which seems (seemed?) to use the
Roussilhe projection as well?

Try for instance:
lat=50d52m; lon=20d37m; // Kielce, in Strefa I
lat0=50d37m30s; lon0=21d5m; x0=5467000; y0=4637000; m0=0.9998;
Krasowski ellipsoid
I got with my "corrected" rouss:
x = 5434153.30654471, y = 4663982.52582146 (don't use all decimals)
Does this make sense?