Comment by Freak_NL | Hacker News Reader

Comment by Freak_NL | Hacker News Reader

As an OpenStreetMap contributor this looks weird. Some names are off from what is recorded in OpenStreetMap. I wonder if…

…and sure enough. This app uses one of those moronic map styles which doesn't just use the names for places in the OpenStreetMap database, but also draws them for Wikidata. Great. Those often suck for use on a map, because they tend to include stuff like 'Province X' for a province named 'X', and because the translations stored in Wikidata are often archaic or plain wrong.

The argument the developers who use the wikidata IDs stored in many place name objects to draw in 'missing' translations is that this is surely better than just showing the local name, but this completely misses that in OpenStreetMap name data is already really quite accurate, whereas Wikidata…

As an OpenStreetMap contributor I am not interested in having to keep two databases up-to-date to get names (exonyms and endonyms) to show up correctly. I know a number of developers are pushing this dumb solution, but it is hurting the project more than it is helping.

Screwups happen all the time. For example in Mapbox basemaps Skorki (some village in Poland) was both:

- transalated as Dermaptera (which incidentally are called skorki in Polish)

- shown at totally inappropriate low zoom level because of the previous point (presumably that Wikipedia article was more linked to or had much more languages, which could be used to decide importance rank)

Maptiler tiles prefer Wikidata over OSM. More info is here: https://github.com/zbycz/osmapp/issues/221

Map labels are complicated and Wikidata is (in general) a better place than OSM to sort that out on the data level.

Consider all the various *name* tags in OSM and how they are (mis-)used to please specific renderers and geocoders. E.g. do you apply (semi-)standard abbreviations of street names or do you write names in full so that they fill the map? It's fuzzy in OSM: https://wiki.openstreetmap.org/wiki/Abbreviations

(I'm all for everyone using the map style that works the best for them - after all, it's open data and open source.)

Using the full names isn't to 'fill the map', it's because the process of going from full name to abbreviation has a lot less ambiguity than going from abbreviation to full name.

Seems the search box is tripping on that (not considering places on North 14th Street for a search like N 14th St).

The full names will be too much in some cases, and there is no abbreviation algorithm that will work everywhere (OSM is a global project). Similarly, "N 14th St" has no meaning in most of the world.

Hence, it's complicated and not "moronic" e.g. to make use of Wikidata for labels and to use external datasets to improve geocoding.

An algorithm that uses location and other hints to detect language will probably come pretty close to working everywhere.

I'm not really arguing with you, just questioned the 'fill the map' characterization and pointed out the more fundamental motivation.

Completely unrelated, but Google Maps seems to have (or had) a rule in place to expand "Int" to "Interchange".

Meaning here in Singapore there are a lot of local bus interchanges that get labels like "Bedok International".

OSMAnd seems to do a good job, how does that handle it?

OSMAnd uses the data from OpenStreetMap itself, and doesn't add any labels from Wikidata when a translation is missing. This is good, because a place name without an explicit name in some other language (i.e., most villages worldwide) won't get some made up name entered on Wikipedia and picked up in Wikidata automatically at some point as they are sister projects.

Names don't always have exonyms.

If a name tag is broken in OpenStreetMap, I can fix it in OpenStreetMap. If a name label is broken in Wikidata, I now have to fix it there as well, and get into edit wars with folk from Wikipedia who use different rules for naming things. So not only are there now two places to fix bugs, there are also two sets of sometimes conflicting policies for naming things. This is because for Wikipedia (which is Wikidata's sister project and has a large influence on it) it may make sense to call the Belgian district of Luxembourg 'District Luxemburg' in English (which is what OSMApp does via Wikidata), but on a map this region should just be labelled 'Luxemburg' (if using English), or preferably just 'Luxembourg' (the endonym). That it is a Belgian district, and not the neighbouring country can be made sufficiently clear by styling and geography.

After all, you don't expect the literal label 'New York City' for the city and 'New York State' for the US state on a map; just 'New York' for both, with styling and size indicating which is which. Any map which does want to render these names with their classification can do so by adding '(city)' or '(state)' to the label if they want to. It just shouldn't be part of the name in the data source.

Map labels are complicated, so leave that complexity to a project which deals with map data; i.e., OpenStreetMap.

This is a discussion more about data vs. maps, there are data purists in OSM as well. So the correct way to do it according to a large portion of OSMers would be to write out the full name.

I agree with you.

Can you provide some examples ?

Also note that you can chose your map style.

https://osmapp.org/#16.72/46.3903/14.0007

Municipality of Bled... in the middle of the forrest... who the hell searches for "Municipality of bled" and wants to be pointed to a forrest? But somehow, it's probably the middle of the polygon of the actual municipality, and the label is put there and visible even when very zoomed out (along with other municipalities,... but not city names, like actual Bled - https://osmapp.org/#8.33/46.3350/14.1244 )

Not sure if this is directly related but all province names in Belgium are displayed in Dutch, even in the French speaking regions. Waals-Brabant or Provincie Namen, for example.

Another criticism I would make is that different levels are displayed the same, for example in France I can see both "Pas-de-Calais" and "Hauts-de-France" in the same style, at the same time, although one is part of the other.

I noticed Namen/Namur too. That is one where the label seems to be drawn from Wikidata instead of using OpenStreetMap data which has the names correct for each language.

For the city Namur though, it's the French name that is displayed.

And for some reason, the province of Luxembourg becomes "District Luxemburg", which seems to suggest it's Dutch, but the Dutch name would be "Provincie Luxemburg".

Luxembourg itself (the country) does have districts, but this is the Belgian province, not a district of the country of Luxembourg.

I think Switzerland has the same kind of problem but with German being used for most names.

Thanks ! Looks like a problem with MapTiler.

Reply on news.ycombinator.com