Talk:Tag:place=city

When should one tag a city as a town?

I have updated the lead to clarify the difference between cities as defined by official designation and the use of the city tag in OSM. In particular to highlight when it is appropriate to tag a settlement as town in OSM even though it is official designated at a city. Lets talk about it if we need to work on this more. PeterIto (talk) 11:13, 14 February 2013 (UTC)

This lack of understanding is becoming a serious problem. I find numerous regions (edited by many different people) where tiny populated places of a few hundred people are being upgraded to place=city, "because that place is a city". iD says in its pop-up help,

• city: The largest urban settlement or settlements within the territory.
• town: An important urban centre, between a village and a city in size.
• village: A settlement with between 1,000 and 10,000 inhabitants.
• hamlet: A settlement with less than 100-200 inhabitants.
• locality: A named place that has no population.

This is all fine and good -- if you know to look for those tags, and if you understand that these tags are used to rank importance of a place based on population and distance from other populated places of importance. A new iD user who looks at a populated place and thinks 'But that's a city', and types in city may very well tag it that way.

The problem is that OSM's place tag names often overlap with the formal administrative titles of populated places, and it's a confusion I can't see going away.

If we really want to end this kind of confusion, it would help a lot if we simply used one tag, like this:

• place=settlement (Used regardless of the size or title of a human settlement.)
• name=Howard (Obvious. Used the same as it is today.)
• layer=0 (Optional. Used exactly how layer is today with other data sets. One number higher than the other means that settlement is 'rendered on top' when a render conflict exists. Assumed zero by default.)
• population=9999 (When layer numbers match, population can be used to determine importance.)
• admin_level=0 (Used how it's used today. Doesn't determine render importance, but is used by nominatim for administrative boundary relation purposes.)
• admin_title=town/village/city etc., an optional User Defined field. A default title is assumed when an admin_level exists, and this field does not. (Not used to determine render importance at all, but could be used to supplement the name of a place, like 'Village of Howard' or 'City of Appleton', or 'Acadia Parish' in Louisiana where counties don't exist at all.) These can be used by a map renderer, or by Nominatim to title a given administrative area with any name desired. Deprecates place=county/state/etc.)

I'm just tossing this idea out there as 'an idea', and I grant that a change this sweeping would be a stretch to see approved. That said, unless something like the above, or edit locks are implemented to prevent this kind of place tag upgrading, I imagine it happening for the indefinite future. Skybunny (talk) 22:32, 9 June 2014 (UTC)

I generated a formal proposal for the classification criteria here: Proposed_features/Populated_settlements_classification --Iagocasabiell (talk) 01:56, 7 September 2019 (UTC)

"Charter cities"

I have a problem with the 2018-added (fairly recently) comment by Dcapillae regarding "charter cities" in the United States: huh? First of all, some states in the USA have these, some don't (California, where he gives the example of Alameda, does). Second, this is a legal definition that classifies governance and has little to do with population or whether another city (larger or smaller) is proximate to the charter city. Third, the suggestion that we do this smacks mightily of "tagging for the renderer." Let's tag cities according to their existing definitions, guided by their population and let renderers deal with it. I propose we remove Dcapillae's suggestion. After all, OSM is about the data, not the rendering. Stevea (talk) 19:34, 22 April 2019 (UTC)

I agree with that. Although maybe not tagging for the rendering, but tagging for the "gazetteer search results." Generally, the criteria for when to use a tag shouldn't be based on it's geographical proximity to something else. Population is way more scientific and probably makes better sense to the average person. --Adamant1 (talk) 09:12, 25 April 2019 (UTC)

Population tag values of place=city nodes

I've downloaded all the place=city nodes in the database, and all the place=town and place=village in a number of countries for comparison. See Talk:Tag:place=town for details about towns and Talk:Tag:place=village for villages. I didn't download ways or relations, since most of those are duplicated by nodes (and only 14% of place=town are also mapped as an area).

It turns out that the previous suggestion on the wiki that villages have 1000 to 10k population wasn't based on the data: the median (average) village size is 500 to 1000 in most countries, and the 90% or 95% range is about 100 to 10,000k (though many villages do not have a population tag, which makes this analysis limited). Towns mostly range in population from 1k to 100k, and the median in each country ranges from as low as 3k in Canada and 4k in Greece, to as high as 30k in Japan and 50k in South Africa (I believe many smaller towns lack a population tag in South Africa, but many of the towns in Japan have a tag, so it's numbers are relatively reliable).

Looking at cities, the median city population tag value globally is 134k. I suspect that many larger cities actually have a population value for the municipality rather than for the entire urban area which is referred to by that name, but this is more of a problem for cities with population >200k. There are 9812 nodes with place=city tags globally, and 6185 (63%) have a population tag. Of these, 3838 (62%) are >100k, so 38% are less than 100k, with median of 134k worldwide (again, among those tagged with population=*). 1101 (17%) are less than 50k, 295 (5%) are less than 20k, and 194 are less than 10k (3%). Only 7 place=city have population recorded as less than 1000k.

I would suggest adding some of these statistics to the page. I believe it would be useful to add a summary something like this:

"The minimum population of a place tagged as a place=city varies depending on the country and for different regions within large countries. For this reason, it is helpful to add a population=* when possible. Currently 63% of place=city have a population tag. Of these, 95% of place=city have population=20000 or greater, and only 3% of population=* values are less than 10,000. However, a third (33%) of place=city have a population= value between 20,000 and 100,000.

This would help remind mappers to consider the local context and how the tag is used in their country or region rather than focusing on particular population cut-offs, but also give some information about values that are lower than average; e.g. places with <10k are rarely tagged as cities. I've also seen a number of places that "a city should have more than 100k population", and these numbers show that this opinion is not widely supported by current usage, at least in many countries. --Jeisenbe (talk) 01:11, 20 August 2019 (UTC)

I appreciate the fruits of research and analysis Jeisenbe presents here. With the exception of one minor typo I've corrected in his proposed Summary statement, I not only agree with it fully, I believe inclusion would benefit the Page. There are opinions that "city must be ≥100k," but where those prevail, the data bear that out (i.e., THERE, town and village are used for what other places often call city, a "small city," to be sure, but tagged place=city nonetheless due to more-local convention). I notice as this has evolved, Jeisenbe has omitted (I paraphrase and embellish somewhat) "Including population data (when known) can help determine whether 'edge cases' might be better tagged with the next place=* value up or down, as populations do change over time." I believe that (or something like it) is a helpful addition to the proposed Summary statement. Putting smart "hand-holds" like these into our wiki makes data better over time, strengthening our tagging with clear intent as more consistent. Stevea (talk) 01:56, 20 August 2019 (UTC)

I included the sentence "For this reason, it is helpful to add a population=* when possible" in the example text, which is meant to be similar to what you suggest above. I'm not sure if we should imply that an increase in population automatically warrants changing the classification of a settlement: if all of the nearby cities and towns have also gained (or lost) a similar percentage of their population, then it wouldn't be necessary. It's more that having the population=* tag lets database users make their own assumptions, especially about how to render or interpret 3 place=city with 1) population=30k 2) populatino=130k and 3) poplution=900k (most will treat the later specially, by displaying at earlier zoom levels, or placing search results higher on the list, etc, and some may distinguish between the first and second examples, but this depends on the database user). --Jeisenbe (talk) 02:03, 20 August 2019 (UTC)

Sound logic, well explained and so, "fair enough." What we fully agree upon is that "this depends on the database user," and we only slightly differ on offering the person entering the data good guidance. Again, I believe the Summary text is well-researched, well-written and deserves a place (heh) in the Page. "As is," or if others wish to offer additional suggestions. One clarification (usually gets added to "currently") would be a date to that context, like "as of August, 2019"). Stevea (talk) 02:08, 20 August 2019 (UTC)

I've added the new section now. --Jeisenbe (talk) 06:15, 24 August 2019 (UTC)

Recently, the Portuguese community decided to apply the city tag to all cities in the country, regardless their size and regional importance, a definition that must comply to certain parameters so that the Parliament approves the place as an official city. That option was made due to the [[1]] parameter, which makes easy for any database consumer to fetch all cities in Portugal, all towns, etc. Using population is 1st) not verifiable (a statistic that changes with time); 2nd) debatable regarding where mappers draw the line on what is a city (more than 50 thousand? More than 70? 100?); 3rd) not easily updated due to population changes. Wouldn't it be more verifiable and easier to understand (by mappers and data consumers) to define places' rendering and hierarchy based on other tags, like capital=*, admin_centre roles, etc? --AntMadeira (talk) 23:26, 2 October 2021 (UTC)

Rendering example?

This file was just added as a rendering example: File:Place-6.svg which looks like:

But this is only the symbol used by the OpenStreetMap Carto style at low zoom levels. At most zoom levels, a place=city is only represented by the text of it's name=* tag. I'm not sure if there is a better way to show this, but I worry that the current rendering sample is misleading. --Jeisenbe (talk) 05:13, 7 November 2020 (UTC)

National folklore ruins the usefulness of place=city

Road sign indicating the following cities in descending order of distance

What is a city in OSM ? Short definition from the wiki: the largest settlement or settlements within a territory. This definition is compatible with the first good practice: Map what's on the ground. But how? Interestingly, population isn't on the ground in the sense that we need it: it moves, and it is difficult to verify on the ground.
So, what is on the ground? Buildings. Airports for wide-body aircraft. Railway stations where all trains stop. Motorways. Universities. Major institutional buildings... And that's not all: Cities are important far and wide, often within the area of influence of another cities (same could be written for towns, of course on a smaller scale, and dwarfed by the surrounding cities). And this distant influence can also be verified on the ground: a city is indicated by distant road signs; it is a rail destination on long-distance lines...

Cities are one of the most important geographical landmarks used by people. When you meet someone abroad who asks where you're from, answering "Fiães" or "St Davids" straight away would be impolite unless you're the local idiot who imagines that the whole world knows your village. We start with the name of the country, and if the person knows a little about the country, we may mention the name of the nearest city. Highlighting cities is therefore a necessary quality of maps.

I would argue that OSM's general definition of a city, combined with a little thought and pragmatism, results in a useful and reasonably verifiable mapping that can be applied in most countries.

So why do some communities allow charming but insignificant little towns and villages to be mapped as cities in OSM? Why is OSM so much worse than other maps at consistently mapping the world's cities? Look at Portugal at low zoom level [2]. It would be easier to find the cities of Portugal on a paper map put through the washing machine. In the title, I wrote "National folklore" to be nice. But the reasons seem to me to be far less glamorous.

Laziness

To cut through the long debates on the classification of unimportant towns, which some local contributor always wants to see bigger (not smaller) on the map, we choose an official list, and the more cities it has, the less the community will hear from local promoters who want to be more visible.

Dogmatism

OSM tags values, like city, are just computer codes matching an OSM definition. But it is spelled C I T Y, and in the Middle Ages some villages were granted a "city" (see, same letters C I T Y) status because they built a cathedral. Yes, it's stupid as it sounds.

Short-sightedness

The usefulness of place=city becomes apparent at low zoom levels, which often display several countries at the same time. Applying a national rule with marked national idiosyncrasies shows a lack of consideration for the global map that smacks of short-sightedness (if not chauvinism).

Marc Mongenet (talk) 23:02, 30 June 2023 (UTC)

Very rational and valid points, but you still didn't define what's a city. Defining what is a city in a paper map is easy. It's a decision of the producer based on a clear hierarchy from the biggest to the smallest (size and importance, and more or less based on the common sense rather than data) and based on the space available on paper. In OSM you don't go that way. You need something which defines what is a city that is objective and without dubious/unverifiable criteria. Portugal did that. It tried, not with chauvinism, laziness or dogmatism, but rather on official data specific from the country and approved in the Parliament, to define what is a city. It's not perfect? No, it isn't, but it's an honest try to frame this tag in verifiable parameters. Trying to heterogeneize what a city is (or town, or village) in the entire OSM map is condemned to failure, unless you decide all the cities of the world from a list of hierarchical importance for each country, based on a bunch of criteria that will always be a matter of opinion. In my view, it should be the render to decide, based not only in the place=city tag, but in a group of tags and relations. Regards. --AntMadeira (talk) 01:09, 1 July 2023 (UTC)

I am thinking about what data is missing to allow the renderers to do a good job. In Openstreetmap default renderer (which is my favorite renderer), cities appear at zoom level 4. In Europe, country names take a lot of space, so I guess the renderer had to make choices that could explain why several major cities like Amsterdam, Milano, and Madrid are missing. But why are cities like Chicago, Boston, and San Francisco missing? I have no idea.

It gets more interesting at zoom level 5. I see that San Jose is displayed instead of San Francisco. But most travelers looking a map at zoom level 5 are probably plane travelers, and San Francisco is the common destination of international flights, not San Jose. So I am thinking about new tags to tell the importance of a city in the air traffic network, in the road network (like how far from the city do you see signs to the city), in the railway network, academic network, and so on. These tags would have numeric values instead of an arbitrary list like the place tag. And it would allow renderer to select the most relevant cities for each use case: cities relevant for https://www.openstreetmap.org/#map=7/45.082/4.373&layers=T (railway-oriented) may not be exactly the same than cities relevant for https://www.openstreetmap.org/#map=7/45.082/4.373 (road-oriented).

You are right, I was thinking about Portugal when I wrote laziness. But avoiding the work of producing a good map by reusing a list that wasn't designed for that purpose, that doesn't match the OSM description of the tag value, that only sounds like the computer code city, and producing an extremely bad map, could also be called by other, possibly harsher names.

I'd like to emphasize a point that seems to escape contributors who speak a language that has a word directly matching city (like city in English, or cidade in Portuguese): the list of values for the place tag are just computer codes, and they could be renamed overnight to settlement_level_1, settlement_level_2, settlement_level_3... What is important is that the used code matches the OSM definition (largest settlement or settlements within a territory). Marc Mongenet (talk) 01:56, 2 July 2023 (UTC)

You're still going in circles and don't define what is a city. Saying "largest settlement or settlements within a territory" could be anything to one's liking, it's not a verifiable way of tagging something. Also, a tag is not just a computer code, it as semantics, it as a meaning. FYI, before the change in Portugal, this tag was used in settlements that were district capitals, but that has the problem of the size and importance in more urban areas, because there are district capitals which are way smaller than others that are not. So, as you see, unless you come up with a concrete and verifiable definition, saying that "the map is ugly" is not a valid argument. I didn't and won't check if all the countries in the world have the same description for what a "city" means, but I can say with a good amount of certainty that the vast majority of them have a well structured way to define what is a city, what is a town, what is a village, etc. Portugal has it and the Portuguese community adopted it, but this is not a dogmatic decision. We'll be more than happy to change it if you or other members of the OSM community presents better arguments than "the map is ugly". --AntMadeira (talk) 13:20, 3 July 2023 (UTC)

The "city" of Ezmoriz...

...dwarfing Porto.

First I would like to clear some misunderstanding. I fully agree that "the map is ugly" is not a valid argument. It would be valid in a discussion about renderers, but not here in a discussion about tagging. When I write that the map of Portugal is terrible at zoom level 7, I don't write that because I find the map ugly; I write that because the map fails at its primary role: to display well-chosen landmarks to guide the viewer. The viewer (me) is lost in a deluge of cities. In [3] I don't see Porto, but I see Esmoriz. And it's not the renderer's fault; the OSM community of Portugal decided that Porto is not really more important than Esmoriz.

"Also, a tag is not just a computer code, it as semantics, it as a meaning." I am not sure what you mean. What is the semantics, meaning, of tag tracktype=grade4? It is fully given by this wiki (in https://wiki.openstreetmap.org/wiki/Key:tracktype). For place=city it is the same: the meaning is given by this wiki (in https://wiki.openstreetmap.org/wiki/Tag:place%3Dcity). Ignoring the wiki definition, and trying to find a word in one's own language that sounds like city (cidade maybe) to reuse a list of cidade made with criteria unrelated to this wiki would be a headfirst jump into a kind of etymological fallacy (and I write this thinking of a country where it rains more than Portugal).

I don't agree that largest settlement or settlements within a territory" could be anything to one's liking. If I ask someone (in Esmoriz maybe?), where is the nearest big town, what will be the answer? People know where to go when they can't find something (a product, a service) in the nearest town. For instance when they need higher education (university); when they need to travel far, using a fast train or plane; when they need specialized medical care (say heart surgery); when they want a large choice of entertainment (opera)...

I read in https://en.wikipedia.org/wiki/List_of_cities_in_Portugal that a locality can only be called a city if more than 8,000 inhabitants live in the city's urban area. In addition, at least half of the following infrastructure must be present: pharmacy, hotel, pre-school and kindergarten... That's in fact a pretty reasonable criteria for OSM place=town! Change 8000 to 80000, kindergarten to university, and you quickly get a pretty reasonable criteria for OSM place=city.

This criteria based on population and infrastructure is a good start, but it's done for municipal authorities. The best map in the world needs better criteria. It is difficult to describe, but here are some initial clues to what I have in mind:

https://fr.wikipedia.org/wiki/Aire_d%27attraction_d%27une_ville landmarks useful at zoom levels displaying motorways, national highways

https://www.flightconnections.com/flights-from-geneva-gva to identify landmarks useful at zoom levels used by long distance (i.e. airplane) travelers

"And it's not the renderer's fault; the OSM community of Portugal decided that Porto is not really more important than Esmoriz." I'll stop this discussion here, not only because it won't go anywhere, but because you're not interested in really discussing it seriously. Using a render problem in the Humanitarian layer (try using the main layer, instead) and saying what you said it's not an honest and serious approach to try and solve a tagging problem. If you want to continue, open a topic in the tagging mailing list or on the community forum, which are better places to discuss this. BTW, by that criteria of flights, Portugal would have only three cities, and one of them is no bigger than the top 10 in the country. Regards. --AntMadeira (talk) 17:45, 8 July 2023 (UTC)