Talk:Proposed features/drop recommendation for place name

From OpenStreetMap Wiki
Jump to: navigation, search

Double rendering in some cases

You forgot one major point in your wiki. The "place_name" has been created for a rendering issue when you have a "landuse=residential" AND a "place" node for a village. If you change the "place_name" to "name" on the landuse area, then the village name will be rendered twice. -- Pieren 15:16, 22 November 2012 (UTC)

I get that it was created to sort a rendering issue. And it might not exactly fall in the recommended Tagging for the renderer, but IMHO, if a rendering issue is sortable some other way, it should be used instead of creating tags made for that and create doubts for mappers because they might fail to understand why we sometimes use name, and some time place_name.
The question is, is it avoidable ? IMHO : yes. And one way to avoid double rendering is to follow One_feature,_one_OSM_element. This makes it simple for mappers to understand as it is allready the recommended practice for shops (and others) : If you can't record the shape, use a node, if you can record the shape, remove the node and copy tags to the area. A town could then be recorded either by a node (as usual) or by an area (but avoiding to tag both)
In your example, this means to either add the name to the "landuse=residential" + add a place=whatever to the same area, or to not add a name to the area at all and only tag informations on the node.
One case however I haven't though of, is that if we consider the landuse and the area place as two seperate on the ground feature, we then break One_feature,_one_OSM_element and we might encounter rendering difficulties again. But this either is nitpicking and irrelevant (render could choose if they render both name for the same object since in overwhelming cases it is the same name) or we could just use two seperate supperposed ways (I'm part of people thinking a landuse=residential is rarely the whole place area) or if we accept more complexity, create a type=multipolygon+place=x relation with the landuse as member. sletuffe 18:07, 22 November 2012 (UTC)

A settlement usually consists of a mosaic of landuses (residential, commercial, industrial...). If you set the name of the settlement on each landuse, it's just wrong for each of them. It's not the name of any of the fragments. It's the name of the whole thing. Moreover, each fragment may have an actual name on its own, e.g. Industrial Park X or Y Corporation. In summary, you shouldn't set name=* of an area to the settlement name.
But I don't see any point in using an other key like place_name=*, either. A settlement is usually spread around a center, with high density in the center and less density in the outskirts. The density just decreases with radius. You cannot define a line where a town ends (except for administrative boundaries). Therefore, you cannot decide which areas belong to the town and which don't (except for the administrative boundaries, again). Just set a place node with name=* in the center and forget about the the place_name bogus. --Fkv 19:14, 26 November 2012 (UTC)
There are plenty of area features that are somewhat fuzzy, but we still map them and try to make a reasonable assumption. A rough evaluation of a place area still provides more information than a single node. Otherwise, it would be useful to additionaly map the recognized center of a place, and this deserves a specific key. --Kaitu 14:09, 27 November 2012 (UTC)
+1 to Kaitu, not only we can still map this way in fuzzy boundaries some times, but there are juridiction, in france for instance, where this has been defined as not fuzzy at all but extremly precisly, and therefore, is mappable. sletuffe 15:57, 27 November 2012 (UTC)

The standard way of mapping a town with a distinct downtown is a relation for the boundary and a node downtown. Deleting the node loses the information about where downtown is (which is where any decent map would actually label the town). But there are ways to solve this that do not involve tags like place_name, such as adding the node to the boundary relation. --NE2 01:02, 11 December 2012 (UTC)