Summary

Some objects, such as cities, are tagged with both place=* and boundary=administrative+admin_level=*. Should it be one or the other, and if so, which one, or is there a conceivable benefit for having both sets of tags?

Detailed problem description

This is a Nominatim related question. I don't think I will ever run out of those.

Here's the outline polygon for my favorite city: http://www.openstreetmap.org/browse/way/33261354

It's tagged with every possible key and it may be a case of TMI (too much information):

boundary = administrative
admin_level = 8
border_type = city
name = Irvine
place = city
place_name = Irvine

plus a full complement of is_in tags (which may or may not be needed depending on who you ask) and the usual tiger: schema tags.

I don't know if place_name and border_type tags are of any consequence. However, I believe that because it is tagged as both a place=city and boundary=administrative + admin_level=8, Nominatim ends up importing it twice as two separate places, and it bothers me. Check it out:

Irvine as admin boundary polygon
Irvine as place polygon

Both point to the same OSM polygon.

I assume Nominatim has its reason for importing OSM places and admin boundaries separately. I am guessing it does not expect the same object to be tagged as both. What ends up happening, if you look at the children, is that, as far as Nominatim is concerned, one "Irvine" is a parent of the zip codes, most neighborhoods, streets, parks, etc., while the "other Irvine" is a parent of small, but not insignificant number of neighborhoods, shopping plazas, streets, etc.

In practical terms, when Nominatim returns a search result set or does reverse geocoding, it is going to list any street or park as being in "Irvine". Whether its one "Irvine" or the "other Irvine" is not that important to the end user, but you have to agree that the situation is pretty clumsy and is rife with potential for data integrity problems (in Nominatim, that is).

The fact that there is also a "third Irvine"

Irvine as node

represented by a node tagged with place=city, with its own subset of children that are missing from either of the other two Irvines, is a whole other story.

asked 14 Apr '11, 00:46

ponzu's gravatar image

ponzu
2.0k496382
accept rate: 0%

edited 20 Apr '11, 01:20

Hey, this sounds quite similar to my recent question: http://help.openstreetmap.org/questions/4068/when-does-it-make-sense-to-use-place-on-an-area Good that there's some more discussion on where the place tag is appropriate.

(14 Apr '11, 11:07) sleske

Similar, yet different. You ask when to use one set of tags vs. the other. I ask if it;s ever a good idea to use both. I am slowly arriving at a conclusion that there isn't, albeit without any input from the gurus.

(14 Apr '11, 19:22) ponzu

Why didn't you guys tell me that my links don't work (because of the escaped underscore?) Let me update them and see if the question gets an actual answer.

(14 Apr '11, 23:28) ponzu

Check out the "As Areas" section in the wiki page for place. One suggestion there is to mark a node as the "centre" of the city (where the name will go) with place=city. Create a boundary=administrative relation for the perimeter (with the appropriate admin level), and make the node a member of the relation with role=admin_centre. See also the Relation:boundary page.

permanent link

answered 14 Apr '11, 02:43

Ebenezer's gravatar image

Ebenezer
94821121
accept rate: 9%

Yeah, I have seen those. The "As areas" best practice is actually a pretty poor piece of advice written without any thought of what having a place node and an admin boundary for the same city, county, etc. object will do to Nominatim searches. I have researched and talked about this issue extensively over the past two days. The consensus today is to delete the place node when a polygon is available, not add one. Adding a node to a relation with a "label" role is a swell idea, but it is not currently supported by renderers.

(14 Apr '11, 03:39) ponzu

Adding a node to a relation with a "label" role is a swell idea, but it is not currently supported by renderers.

True, but doesn't that mean the correct course would be 'add nodes with no tags but with role=label', and lean on render developers to support that?

Particularly since "delete the place node when a polygon is available, not add one" seems to be the best/accepted practice, and in that case, most renderers show nothing at all on a map for a border relation without a place node -- right?

(30 May '14, 22:03) Skybunny

Note that this is a pretty old discussion. One of the Nominatim maintainers recently wrote (here? talk@?) that it is not a good idea to remove the place node if there is a boundary.

(31 May '14, 04:42) cartinus
Your answer
toggle preview

Follow this question

By Email:

Once you sign in you will be able to subscribe for any updates here

By RSS:

Answers

Answers and Comments

Markdown Basics

  • *italic* or _italic_
  • **bold** or __bold__
  • link:[text](http://url.com/ "title")
  • image?![alt text](/path/img.jpg "title")
  • numbered list: 1. Foo 2. Bar
  • to add a line break simply add two spaces to where you would like the new line to be.
  • basic HTML tags are also supported

Question tags:

×824
×549
×31
×12

question asked: 14 Apr '11, 00:46

question was seen: 4,640 times

last updated: 31 May '14, 04:42

powered by OSQA