[Talk-us] Admin boundaries tied to roads

[Frederik Ramm]

Hi,

Alan Mintz wrote:
> At 2010-04-19 10:45, Mike N. wrote:
>>   I see that the separate VS tangled argument has been settled in the US by
>> the "Duplicate Node attack bots", who have blindly merged all duplicate
>> nodes.
>>
>> http://www.openstreetmap.org/browse/way/38855677
> 
> Is this really happening? Can someone describe exactly what criteria are 
> being used, and just how it was decided that this was a good idea? 

It seems that someone is, more or less blindly, using the JOSM validator 
de-duplication. Doesn't look like a bot but, as Richard said, has 
similar results.

For the record, the proper way of aligning a boundary with a road (at 
least if the alignment runs for more than a few nodes) is NOT to have 
the boundary use the same nodes as the road, but instead to create a 
multipolygon which, on one side, uses the road itself as part of the 
"outer" ring. (This would also be a very clear expression, data model 
wise, of the fact that the boundary is defined by the road, instead of 
accidentally happening to coincide with the road.)

This is especially important if you create a mesh of boundaries: There 
should only ever be *one* way connecting any two nodes, and neighbouring 
boundary multipolygons would then both use the boundary way as part of 
their "outer" ring.

Of course if you have a very small mesh-like structure like a patch of 
adjoining buildings then you would probably not make the effort of 
creating a multipolygon relation for each. But as soon as you are 
dealing with larger structure, this makes a lot of sense.

Bye
Frederik