[Benchmarking] Updated on Mapnik status

[Dane Springmeyer]

Hi all,

Just a heads up that my first pass at generating Mapnik results last
week -  I have now removed from the google doc. Thomas, very kindly,
helped me spot check the correspondence of sql queries used in the
Mapnik style and the Mapserver style, and we (Mapnik team) made
several mistakes when authoring the style (We use TileMill a GUI
around Mapnik) such that we were unintentionally pulling much less
data than MapServer.

Originally this had not been a problem because I started by using
python/regex to try to automatically assemble the Mapnik style from
osm-google.map. This effort failed ultimately because the rendering
model of MapServer is different enough my brain was melting trying to
auto-convert things like casings and labels so they rendered
correctly. It just was not looking like I could get things to 100%, so
I gave up. Good learning experience though!

So, after I ditched that effort AJ Ashton, who works with me, started
fresh by branching an internal stylesheet we (Development Seed) use
called OSM Bright, which is highly optimized to imposm generalized
tables (and previously osm2pgsql schema). Long story short, I'm still
working on adapting this style, basically properly unoptimizing it so
that I can try to match the osm-google.map sql queries exactly (to
ensure to the best of ability the same exact features are being
pulled). I'm still working on this, so this is why I've removed the
previous invalid results and hope to have new ones soon.

I'm at SOTM now, and looking forward to seeing/meeting many of the
people involved in the benchmark very soon as you all arrive for
foss4g.

And as a side note: Mapnik does not support raster reprojection, so I
have not done a run of the raster/vector combined. But, I am working
to add raster reprojection this week so I might be able to do a run
for this test. 50/50 chance I may just have to sit out this test.

Cheers,

Dane