[OSM-talk] calculating lon/lat bounds from tilename

Martijn van Oosterhout kleptog at gmail.com
Sat Jun 23 17:30:27 BST 2007


On 6/22/07, Jon Burgess <jburgess777 at googlemail.com> wrote:
> Whether this distortion is more or less acceptable than tiles where the
> lat/lon are non-linear across the tile depends on what you are trying to
> achieve. I think a 3D representation is the only thing that people agree
> is 'fair'. All 2D projections suffer from at least one weakness or
> another.

Well, Mercator is pretty popular because:
- lat/long lines are all straight
- preservation of angles (north-east is always halfway between north and east)
- people are used to it

Especially the angle-preservation makes it ideal for streetmaps,
because it most closely corresponds to how people perceive the
real-world. Mercator looks wierd on a large scale, but on a small
scale it preserves shape also, squares that are not too big remain
square.

> By the sound of things people have applications which want to work in a
> nice simple square lat/lon grid. For these applications, tiles which can
> map to a lat/lon grid are probably much easier to work with than the
> current slippy mercator projected tiles.

Hmm, if you mean 1 degree is the same anywhere then I agree with other
people, that looks wierd. JOSM has two choices for the projection, and
the other one looks very very strange.

For the application given though, all he want is the coords of the
corners, which is a SMOP (simple matter of programming). No need to go
reproject the entire online map for that...

Have a nice day,
-- 
Martijn van Oosterhout <kleptog at gmail.com> http://svana.org/kleptog/




More information about the talk mailing list