[Tagging] Chinese counties (was: Wiki: Chinese road names)

Minh Nguyen minh at nguyen.cincinnati.oh.us
Sat May 15 08:34:32 UTC 2021


Vào lúc 01:31 2021-05-14, Sarah Hoffmann đã viết:
> On Fri, May 14, 2021 at 08:28:59AM +0200, Jeroen Hoek wrote:
>> On 13-05-2021 23:45, Minh Nguyen wrote:
>>> Is this the same user who suggested deprecating population=* in favor of
>>> china_population=*?
>>
>> Looks like the same user. I hope someone with Chinese language skills
>> can reach out to them, because this has all the makings of a
>> well-meaning individual establishing dubious de facto mapping practices
>> on their own for a single (huge) country.
>>
>> china_population=* is a way to fix a rendering problem without having to
>> get in touch with Carto or working with others at all. Convenient, but
>> ultimately harmful to the global database. Filling population=* with
>> values designed to trigger render behaviour is not in line with OSM's
>> guidelines.
> 
> China has a huge problem with tagging-for-the-render classification
> of place tags. They basically have a 1:1 mapping between place type
> and admin_level, instead of describing the kind of place. There is
> even a huge warning on the China tagging page not to use population:
> https://wiki.openstreetmap.org/wiki/Chinese_tagging_guidelines
> 
> I was made aware that the tagging is explicitly chosen, so that the
> county names render:
> https://github.com/osm-search/Nominatim/issues/2251#issuecomment-813424984
> 
> I'm a bit at a loss what to do about it. It's square against OSM tagging
> practises and it does mess up geocoding. But it's been the convention for
> many years now and is widley accepted in the Chinese community, it seems.

Some countries are principally organized according to space-filling [1] 
administrative regions, which correspond to administrative boundary 
relations, rather than population centers (human settlements). A map 
design that emphasizes administrative regions would typically label as 
many as space allows, regardless of population, whereas a design that 
emphasizes population centers is more likely to distinguish them by 
population.

OSM-based renderers typically label place=* POIs prominently; if they 
label the centroids of administrative boundaries at all, those labels 
tend to be more subtle. Until 2019, openstreetmap-carto only labeled 
administrative boundary centroids at z16+. Nowadays it would label 
China's admin_level=6 county centroids at z13+ in much the way that 
provinces are labeled at lower zoom levels. I wonder if the Chinese 
mapping community would be open to revisiting their stance on place=* 
and population=* in light of these changes. As far as I know, renderers 
don't consider population=* at all in the treatment of boundary relations.

I'm not especially familiar with Chinese geography, but Vietnam follows 
a similar model of space-filling administrative regions with complex 
relationships among them. [2] The Vietnamese community has adopted a 
more conventional approach of using administrative boundary relations 
for the space-filling administrative units (such as communes and 
districts) and place=* POIs for population centers (which may or may not 
correspond to administrative units), even though it's common for 
Vietnamese maps to label communes and districts prominently. [3]

In the U.S., New England states are also organized according to 
space-filling administrative regions (cities and towns). Some of the 
states are blanketed in POIs tagged place=city or place=town according 
to their legal status as cities or towns, regardless of population. I 
don't know whether mappers there were unaware of that place=* is 
supposed to be more holistic, or whether they took this approach because 
it results in a pleasing distribution of labels at every zoom level. But 
you can see a clear difference between Connecticut and Massachusetts on 
the one hand and New York and Vermont on the other hand, which can't 
quite be explained by a difference in settlement patterns. [4] There's 
something to be said for the evenly spaced labels that evoke print maps 
with hand-placed labels.

These are regions with much lower population density than China in 
general. Mappers in New England would rather promote town labels as 
opposed to how mappers in China have demoted county labels. But I think 
the principle is the same: only population centers should be mapped as 
place=* POIs that would tempt a renderer to consider the population in 
the first place.

[1] By which I mean that these regions cover the whole territory without 
any gaps and don't necessarily conform to settlement patterns.
[2] 
https://commons.wikimedia.org/wiki/File:Administrative_subdivisions_of_Vietnam_as_of_2016.svg
[3] http://maps.vietbando.com/maps/?t=0&st=0&l=9&kv=11.6031399,108.066329
[4] https://www.openstreetmap.org/#map=9/42.4093/-73.2349

-- 
minh at nguyen.cincinnati.oh.us




More information about the Tagging mailing list