[OSM-dev] Boundary data

Roland Olbricht roland.olbricht at gmx.de
Thu Oct 30 22:11:44 GMT 2008

Hello everybody,

the boundary in OSM is maybe suboptimal structured. What is the best option?

1) Leave the data as it is and use precautions in the software using it
2) Silently correct the data by running an appropriate script
3) Discuss somewhere (where?) what the script should correct and what not
4) Encourage somehow (how?) the mapper community to correct it manually

In detail, there are
a) the "left:country" and "right:country" attributes in the boundary ways.
b) the nodes with tag k="place" v="country" which contain more information on 
a country.

The information in a) is only partially present, it is sensitive to typos or 
name variants of countries (e.g. "Belgium", "Belgique" or "Congo", "DR 
Congo"), it is redundant, sometimes it is simply inaccurate (e.g. if a way 
passes a point where three countries meet, there is no way to note the change 
of the country on one side) and it does provide no chance for additional 
information, e.g. the names of a country in different languages.

But nonetheless, it is currently used by tiles at home for giving the borders 
names. So silently stripping of the tag might affect the rendering.

Storing the information in b) would leave more space for information about the 
country (currently used for example for the name of the country in different 
languages) but it has almost no connection to the boundary itself an it 
doesn't provide any information about exclaves of the country.

A possible solution would be to have instead of b) a relation which contains 
all the tags of b) and as members all ways that constitute the border of that 
country. A good example would be relation 47796 for the Netherlands.


More information about the dev mailing list