[OSM-talk] to all potlatch and JOSM users - automatic simplification of geometry

[Martin Koppenhoefer]

Hi all,

I want to point attention to the potlatch-funtion tidy-points (similar
to JOSM simplify way). I encourage everybody not to use these
functions (at least not on data someone else entered) as it harms
severly the data. IMHO OSM is a project about drawing a collaborative
map, not about computers drawing it. They don't know the world, e.g.
don't know and can't know, where there is a curve and where the street
is straight. For this reason, they will never be able to do the job of
generalization as good as humans can do. Therefore it is not just
about optimizing the algorithm of performing this action, but it is
generally not possible for computers to do it. Of course an elaborated
algorithm (that knows for example the functions usually used to design
curves in streets, ramps, minimum radius, etc., probably even
considerating the road class and speedlimits) could do a much better
job than the current one, but it will still be "guessing" and not
"knowing" about the optimal geometry. I think that the respect for the
work of the comunity requests us to ban these functions or restrict
the use to ways drawn but not yet uploaded.

An example from the result of the current tidy-points-function here:
http://trac.openstreetmap.org/attachment/ticket/2148/090808_potlatch_tidy-points_.png

I'd like to start a discussion about which automated functionalities
we want to allow and if there is a majority in the community for
setting some limits to what should be possible by automated
interaction. What if I scripted my own tidy-up-function that allows to
specify a location, radius and tag and then deletes all these tags
inside the radius. Might be useful in areas to start certain things
from scratch where data is not reliable (I'm obviously not serious,
that's probably an extreme example, but there might be other functions
that in some cases are useful, in others they aren't).

I know that some people like this tidy-points-function to work on
TIGER-Data, but I tell you: if the TIGER-Data is not good/precise, it
won't get better using this function ;-)

cheers,
Martin