[Tagging] R: Request for new tag "natural=upland" (as way) orenabling "way" for "place" tags

Christoph Hormann chris_hormann at gmx.de
Fri Jun 10 11:28:39 UTC 2016


On Thursday 09 June 2016, Amacri wrote:
> > On a general note - when things are mapped as nodes this is
> > frequently done with the implicit notion that this is a location
> > with a certain tolerance margin.  You might think of mapping
> > something with a linear way as a method to specify an anisotropic
> > tolerance, well localized in one direction but poorly in another. 
> > However that is not what you actually do when you map it as a way -
> > on the contrary you much more specifically localize it.
>
> I would ask you to elaborate your concept that I might not understand
> well.

If you in the simplest case map something either as a node or as a two 
node way extending a distance on either side of the node in the other 
case the way does not indicate a higher tolerance in localization in 
direction of the way.  Instead it indicates a larger extent of the 
feature in that direction and generally a lower tolerance in 
localization since you specify two coordinate pairs instead of just 
one.

In other words: when you map something as a way this always says 
something about the shape of the thing you map, it does not represent 
the shape of the probability distribution for the location of an 
otherwise shapeless feature.

> As you mention, documenting the feature by highlighting anisotropic
> tolerance of its extension, as per analysis of historical maps (e.g.,
> official maps from one side but being the only source for data from
> the other side,

A historic map that is old enough so you may legally use it without 
permission of the creator is usually not an appropriate source for 
geometry data and labels in maps are never a suitable source of 
geometry data - after all label placement is subject to a lot of 
influences completely unrelated to the geographic reality.  So this 
line of argumentation does not really convince.

As i said before - you best think about what you know about the thing 
you want to map in terms of verifiable, observable facts.  Based on 
this you can best decide how to represent such a feature in the 
database.

-- 
Christoph Hormann
http://www.imagico.de/



More information about the Tagging mailing list