[OSM-talk] Nominatim & US places

john at jfeldredge.com john at jfeldredge.com
Tue Jan 4 03:28:47 GMT 2011

It makes sense to me to label the administrative center, if any, as well as the geometric centroid of the town polygon.  It is not unusual for a town to expand more in one direction than another, particularly if is near a natural barrier such as a mountain range Orr a lake.  Note that towns don't necessarily have their own administrative center; where I live (Nashville, Tennessee, USA), there is a metropolitan government (a merger of the Nashville city government and county government) that administers most of Davidson County), but a few communities have chosen to have their own local government.

-------Original Email-------
Subject :Re: [OSM-talk] Nominatim & US places
From  :mailto:rwelty at averillpark.net
Date  :Mon Jan 03 21:12:16 America/Chicago 2011

On 1/3/11 9:51 PM, Greg Troxel wrote:
> Richard Welty<rwelty at averillpark.net>  writes:
>> what i see a bit right now in the US are places where we have
>> a central node from one import and a boundary with the same
>> name from another, and as a result two names showing up.
>> it's mildly annoying.
> That may be true but Kurt is right.    For most towns in New England
> there is a polygon for the boundary, and then a specific place, often an
> intersection or a village green or a few streets that should properly be
> labeled as a point.
> If having a polygon with the name and a point with the name as a
> "populated place" produces two names on the map, then the rendering is
> arguably broken.  It may be that if the town center point is in the map
> view, the label should be put more or less there.
there is an underlying data problem. the boundaries and center points
are disconnected bits of data, and i would argue that it's not reasonable
to demand that the rendering engines figure that relationship out.

this is why i suggested adding a "centroid" tag to the boundary relations
as a way to convey the place that is by convention considered the
"center" of town. the existing gnis object for West Sand Lake, NY then 
can be
placed in a boundary relation for West Sand Lake with a role of
"centroid". now the rendering engine can just see if there's a centroid and
use that, if not, it can compute its own. if it doesn't check for the 
and computes its own anyway, that's really not particularly a failure, just
not as nice as it could have been.


talk mailing list
talk at openstreetmap.org

John F. Eldredge -- john at jfeldredge.com
"Reserve your right to think, for even to think wrongly
is better than not to think at all." -- Hypatia of Alexandria

More information about the talk mailing list