avij wrote:The Y axis is more difficult, and as can be seen from the European map the vertical resolution varies between 2 pixels (in the south) to 6 pixels (in the north). What kind of rounding would fix that problem? Or what kind of projection would be immune to this kind of problems?
Pretty much no projection because, well, the earth is a ball (more or less) and the screen is flat (more or less).
Mercator projection, however, is not quite optimal for maps which raise to higher geographic latitudes. Wikipedia
has (section "controversy", at the moment) a good illustration of the area distortion which is the larger the closer you get to the poles. This applies for the standard Mercator projection whose reference line is the equator (basically, everything is projected onto a cylinder which encompasses the earth/ball, and the cylinder touches the ball at the equator). Now the Transverse Mercator projection
could use a different meridian as reference line, probably some meridian that runs through central Europe. Then Portugal and Turkey would appear larger than they are now, Belgium smaller and Finland a bit smaller as well. In summary, it wouldn't help too much.
The main structural problem with Mercator is that it is not area-preserving. This is typical for cylindrical projections.
There is an extensive overview of various projection methods at the website of the USGS
and a more instructive overview here
Generally very good results are obtained with either of the two Lambert projections. The Lambert Conic
is, like Mercator, angle-, not area-preserving but still much better than the former. The Lambert Azimuthal
is area-preserving which comes at the expense that shapes get distorted, but still in a very acceptable manner.
A significant problem with the Lambert methods is to work out the mathematics they use: They are much more messy than Standard Mercator. Plus: Once we had that, and implemented that, we shall need a new map, and already the creation of the last map
was quite an effort. And yes, we almost surely can't just simply download some map because of
2) quite specific needs (just coastlines and borders)
3) the parameters used during the creation of that other map have to be the same as in our projection which draws dots, but these parameters are normally not indicated on a downloadable map (parameters: reference meridian for Mercator, tangential point(s) for Lambert)
I feel that we are doing quite well with the current projection despite its known problems and, for the moment, focus on other issues of site development. In any case (as to my knowledge), we are the only tracking site which provides maps at all. Maybe the text above gives some hints, why.