[OSM-talk] Projections

[Christopher Schmidt]

On Fri, Jan 19, 2007 at 04:12:37PM +0000, Robert (Jamie) Munro wrote:
> Christopher Schmidt wrote:
> > On Fri, Jan 19, 2007 at 03:37:19PM +0000, SteveC wrote:
> >> * @ 19/01/07 03:30:15 PM andrew.rowbottom at gmail.com wrote:
> >>>> Are you trying to say that there is a standard for tiling in Mercator
> >>>> projection written down somewhere, and we are breaking it?
> >>> The big issue here is that the program used to display the tiles must
> >>> know the projection used if it wants to convert pixel coordinates (or
> >>> tile names) to lat/long. The current implementation assumes a
> >>> particular conversion that is not the same for all the different sets
> >>> of tiles.
> >> The current (top two layers) tiles can be dropped in to google maps -
> >> it's a pretty compelling reason to stick with it.
> > 
> > Unless you're using a different definition than I am, I don't think
> > that's true... maybe I'm misunderstanding something though. Google, in
> > my experience, uses a simple mercator projection, and OSM isn't creating
> > tiles that match that...
> 
> It is true. Compare:
>  http://mt3.google.com/mt?n=404&v=w2.37&x=4066&y=2713&zoom=4
>  http://tile.openstreetmap.org/test/13/4066/2713.png

Okay. I was wrong.

If someone can set up a Mapnik WMS server that can serve the images in
the same projection as OSM does, I can learn from that.

There is no projection code in TileCache/WMS-C. There is nothing at all
except a call out to Mapnik with the bounding box sent in the request. 

All the logic is in mapnik, which means that my mapnik config is
probably broken. 

http://labs.metacarta.com/osm/osm-merc.xml 

That's my Mapnik config. Tell me how to change it so it serves the right
data, and I will. Any code that you think I might be hiding, I'm willing
to share: all I'm using is TileCache, which is Open Source and released
at http://labs.metacarta.com/wms-c/ . 

I can't figure it out. If someone else can, I'll commit the fix and get
it into the WMS-C instance.

I'm sorry I persisted in being wrong for so long. I don't know how to
change it now though, and that's where I need help.

Regards,
-- 
Christopher Schmidt
MetaCarta