[OSM-talk] Tiles at Home client update

matthew-osm at newtoncomputing.co.uk matthew-osm at newtoncomputing.co.uk
Sun Mar 4 17:13:34 GMT 2007


Hi,

On Sun, Mar 04, 2007 at 05:45:29PM +0100, J.D. Schmidt wrote:
> The proof is in the pudding... See the examples on 
> http://wiki.openstreetmap.org/index.php/Tiles%40home/Problems and then 
> tell me again it is a major issue ?

Sorry, I have to side with Jochen here.

Your examples just don't hit the places that have problems.
In fact, you don't even need a tight curve for the error to
show up, as these examples show (left drawn without bezier,
right drawn with bezier, hoping the tiles haven't been
re-rendered!):

  http://informationfreeway.org/?lat=6943094.31507&lon=-136979.30516&zoom=17&layers=B00

  http://informationfreeway.org/?lat=6939602.69626&lon=-136979.30516&zoom=17&layers=B00

  http://informationfreeway.org/?lat=6941645.59483&lon=-137002.59456&zoom=17&layers=B00

  http://informationfreeway.org/?lat=6938069.17871&lon=-136981.69382&zoom=17&layers=B00

In fact, scrolling up and down that join, I estimate that
about 50% of the roads that cross the join do not meet up
correctly. You could say that the lower zooms matter more,
and it does not show up so much there. I would argue though,
that as zoom 17 is the most detailed level, it is the most
important to be drawn correctly.

_However_, having said that, there is still the question of
whether the error is worth worrying about with the different
clients out there. My personal preference would be to ban
the older clients sooner and have a standard renderer to get
better looking maps. This means that either everyone should
use bezier, or nobody should.

-- 
Matthew




More information about the talk mailing list