[OSM-talk] The biggest violation of OpenStreetMap, ever.

Christian Quest cquest at openstreetmap.fr
Mon Jul 14 19:38:20 UTC 2014


2014-07-14 16:40 GMT+02:00 Martin Koppenhoefer <dieterdreist at gmail.com>:

>
> 2014-07-14 12:44 GMT+02:00 Christian Quest <cquest at openstreetmap.fr>:
>
> On the first zoom levels, I'm using the capital=* tag to select the
>> country capitals, then sorting them with decreasing population.
>> It is a very small number of objects, that can easily be maintained.
>>
>
>
> this works not too bad for Europe, but fails e.g. for the US, where just
> Washington appears in Zoom 5, but New York City takes up to zoom level 11
> (!) till it gets spelled out, while there is already NYC (short name) in
> zoom 6 together with a sea of more or less unimportant (at that zoom level)
> cities
> http://tile.openstreetmap.fr/?zoom=6&lat=40.74623&lon=-75.75272&layers=B0000000FFFFFFF
>

Not worse than the current osm.org rendering... but I agree that it is
weird ;)

It is not catched by my query because there is no capital=* tag on it.
Albany is the state capital (something I've just learned thanks to WP). So
more tags may be useful to catch these major places.
There are two tags on the NYC place=* node: importance=international and
rank=0.

importance=* is a proposed tag since 2009, with 700+ occurences.
rank=* is not documented in the wiki and currently have 600 occurences.
For place=* nodes rank=0 has 135 occurences, with a lot of then in Lituania
and several too in Brazil... to avoid too many false positive, it need to
be limited to place=city

Given the emptyness of the area around Clermont-Ferrand, and Brive, it is
quite logical to get them on the map. They are the "major" cities in the
area ;)


San Francisco is hard to find, L.A. doesn't appear before zoom 10 (but is
> hard to spot due to its brevity), and spelled out at zoom 11.
>
> But also in Europe there are some serious problems, e.g. Zurich (typical
> hard case, OK) isn't there at zoom6, unlike "Clermont-Ferrand",
> "Brive-la-Gaillarde", or the famous "Ebingen" on the Swabian Alb ;-)
>
>
Zurich... admin_centre:4=yes, a tag you'll find only in Switzerland... no
capital/is_capital/importance/rank...

   -


IMHO we shouldn't use such a simple approach for the main style. An
> alternative to the "opaque" Natural Earth dataset might be a
> community-generated ranking based on a series of criteria (I named many in
> my previous post), and which is continuously discussed, modified and voted
> upon ;-), or a detail ranking for some subjects with relative ranks (i.e.
> more detailed ranking not mixing up religion and economy in one overall
> ranking, but having detailed ranking to mirror relative importance in
> fields like trade, production, transportation, banking, religion, culture,
> health, public administration, education (e.g. universities), ....
> so everybody creating a map can decide what matters to them.
>
> Or we could use some other external dataset, e.g. "important" cities
> according to the analysis of someone else (usually economy centered), see
> e.g. here: http://en.wikipedia.org/wiki/Global_city
>
>

We need a default uniform tagging scheme, then update OSM. capital=* +
importance=* should be enough, with population to provide a sort order to
help text placements for similar capital/importance values.

importance maybe have different subjects attached to it.
For example, importance:religion=international/national/regional so Mecca
or Lourdes may be promoted on some maps but at least data is there.

By switching the default rendering to a uniform default tagging, this will
quickly push us to improve the data.

The only problem may be a new kind of vandalism based on these tags...
>From time to time thank to my rendering I'm detecting new "countries"
coming from mistakes between "country" and maybe "countryside"...


Maybe we should switch to the tagging list ;)

-- 
Christian Quest - OpenStreetMap France
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.openstreetmap.org/pipermail/talk/attachments/20140714/3a01ae9f/attachment.html>


More information about the talk mailing list