[OSM-talk] Komuna e Malishevës, Serbia ?

Paul Norman penorman at mac.com
Sun Apr 1 10:07:15 BST 2012


> From: Lester Caine [mailto:lester at lsces.co.uk]
> Subject: Re: [OSM-talk] Komuna e Malishevës, Serbia ?
> 
> Mike Dupont wrote:
> > 2012/4/1 Lester Caine <lester at lsces.co.uk <mailto:lester at lsces.co.uk>>
> >
> >     Altin Ukshini wrote:
> >
> >         Does anyone care about these reports ?
> >         I said it before and I'm saying it again, we can't
> monitor/administer
> >         the map
> >         everyday and you know this, we don't even have enough people
> >         contributing in OSM
> >         here in Kosovo and Serbians are using this opportunity to
> change
> >         whatever they
> >         want in the map.
> >
> >         I don't think that you might want to loose contributors from
> Kosovo !?
> >         You have to report these violations... how much will this last
> until someone
> >         makes a decision ?
> >         We have already reported so many violations till now.

If you have reports of vandalism that the local community can't deal with,
send them to data at osmfoundation.org. Include lists of objects and
changesets. Without those you can't tell who changed what.

> >     'Location' is somewhat hit or miss everywhere. My own diary
> postings give
> >     some strange textual descriptions of where they were supposed to
> be located.
> >     The problem here is simply that there is no clear mechanism for
> identifying
> >     'where' a location is, so until all of the relevant boundaries are
> mapped
> >     and tested in the same way as the intensive effort on getting the
> coastline
> >     complete it is going to be difficult to fix some of these
> irritations.
> >
> >     I've given up asking for a proper check on hierarchy place and
> is_in which
> >     would at least get towns in the right country ... rather than
> relying on the
> >     mapped boundaries ...
> >
> >
> > this has changed, it used to be kosovo.
> > it needs to be fixed, please.
> 
> My point is ... does anybody actually know where it is broken?
> It is not the 'is_in' tag ... that isn't used to determine location ...
> last time I tried looking at this problem in the UK nobody seemed to
> know how the location was ACTUALLY generated :( It looks to the
> 'administrative boundaries'
> areas and so presumably someone has been playing with them? It's a pain
> trying to find what has been 'modified' ...
> 
> ( No bloody politics here RB !!! )

The best way I've found to debug the location information is to look up the
location on http://nominatim.openstreetmap.org

For the example diary entry linked, this gives you
http://nominatim.openstreetmap.org/details.php?place_id=4568477

This tells you that the place is a place=hamlet node in the
boundary=administrative relation "Komuna e Malishevës"
(http://nominatim.openstreetmap.org/details.php?place_id=128946600)

Now it gets complicated. The node is within three admin_level=2 or
admin_level=3 boundaries. 
These are:
http://www.openstreetmap.org/browse/relation/2088990 with name:en="Kosovo
and Metohija" admin_level=2 created March 18th

http://www.openstreetmap.org/browse/relation/53295 with name:en="Kosovo and
Metohija" admin_level=2 created 2008. The admin_level started off as 3, was
changed to 2 in July 2010 and back to admin_level=3 Feb 2012. The name was
also changed from Kosovo to Kosovo and Metohija at the same time (in all
languages).

http://www.openstreetmap.org/browse/relation/1741311 with name:en="Serbia"
admin_level=2 created in 2011. 

Because regenerating all of the address data inside a country would be too
much of a performance drain, they are cached for large boundary objects.

My guess is what happened is at some point in the past the 53295 relation
was broken and Nominatim cached it. 




More information about the talk mailing list