It is a common misconception that place=city (Butte, MT) and place=town (Anaconda, MT) nodes represent the same thing as a municipal administrative boundary. In actuality they are different, but related concepts. A municipal administrative boundary represents the exact area controlled by the municipal government. A place node represents the general center of a named settlement (populated area of some density) in the popular consciousness. The two often coexist, but not always. An unincorporated community is mapped with a place node, but no administrative boundary. A rural municipality can have an administrative boundary with no place node sharing its name.
1 Like
Right, sorry for insinuating that. It looks like the Anaconda place node, like all the others, was imported originally from GNIS. We only imported GNIS coverage of populated places, which doesn’t distinguish between incorporated and unincorporated places, let alone consolidated governments. GNIS also has a Civil feature type that corresponds to incorporated places, but we didn’t import it.
Most local boundaries were imported from a completely different source, TIGER. That import also omitted lots of relevant boundaries, such as all the townships in several states. We did import Anaconda as a boundary, but it was deleted in 2015, probably because “Anaconda-Deer Lodge County” and ”Deer Lodge County” sounded too similar.
As TIGER comes from the Census Bureau, it may not come as a surprise that it also hews to the bureau’s practice of representing Anaconda as a census county division coextensive with Deer Lodge County. We didn’t import CCDs because they’re very obscure in general.
This is (sort of) off topic, but I’m just taking a moment to say I appreciate all the responses here.
I wish I had realized there was a robust community for me to discuss these things with prior to making changes to the OSM data.
I viewed OSM similar to Wikipedia / Wikidata / Geonames where there’s not a lot of community discussion (that I know of anyway).
I’m also learning much more about the coding through the discussion and the research.
Anyway, thanks for taking the time to tackle this and educate me!
6 Likes
It’s come up in New York City. I had duplicate entries for each of the boroughs and their corresponding counties due to changes in admin_level over time that caused me to accumulate both copies.
And oh, let me tell you about NYC community boards (trigger warning)
I would love to have a way to not have these “single city in a county” entries like Philadelphia, but that’s really a wikidata handling/data modeling issue on my side. That said, I don’t think we should replicate that into OSM unless the city and county are strongly separate concepts.
See, “Denver County” exists. That you find that it does “pointless” might be the root of at least some frustration of yours. But that doesn’t mean that “Denver County” doesn’t exist (or shouldn’t be represented properly in our map data).
As Minh says, San Francisco being tagged border_type=city;county might be workable.
Minh is on the right track here: EVERYTHING is a case-by-case basis. We (in OSM) do work hard to characterize (and somewhat categorize, even as relations are not categories) as best we can so that reality is “best represented” (maybe we can only get to “better”) in our map (data).
There is a lot that is correct in reasoning that in Orleans Parish, it is the level-6 entity which publishes birth certificates and in San Francisco (the City of), it is the level-8 entity which payrolled Minh’s job. I totally nod my head at that, it is the reality that I am told (first-hand) and I can make perfect OSM-sense out of it like that.
Minh’s description of how the county was abolished and the city renamed “City and County of…” is one more example of how CCCs (a loose term, one Minh acknowledges is a Census Bureau-invented category) differ from one another in slight, subtle, variable ways. OSM would also like to “capture the sense of” (as a semantic) with crisp syntax. “Double-entry” (with essentially contiguous polygons tagged 6 and 8) has been our initial foray into doing this. What I think we are now discussing is how to tune the fine points so we have wide agreement about them, especially for downstream use cases.
And, OSM is most certainly not the first to struggle with this (concept of CCCs). But I’m glad to see that our struggle remains reasonable, civil and can even be said to be making progress. We have good people, good dialog, good syntax (at least a rough-cut, initial set of tags that get us most of the way there…) and are using good decorum / language / manners to discuss this. Good for us!
Denver County has never existed, except as a substring in the official name of the City of Denver. This is in addition to my finding it pointless.
OK, does San Francisco (County) exist, other than in the name of the CCC?
I assert it does. I have seen both its (city) police officers and its (county) sheriff deputies.
(Just to be clear, I didn’t work for the City and County, just in the city-and-county, as in geographically.)
They’re all employees of the city. State law doesn’t distinguish one entity or geography from another, not since the county was abolished. The term “County of San Francisco” is still used by the court system, which is organized by county. Some other states have counties that primarily exist in the justice system, like Connecticut.
According to:
It sounds like no. But I would defer to people that know more about the situation there.
“City and County of Denver” on the other hand, was created from whole cloth at its creation. The County of Denver never existed.
Either we or somebody is splitting hairs, notwithstanding that Brian seems to continue to find problems parsing our geo data.
If the “sheriff deputies work for the City” that’s OK, but they are doing so in a capacity where they are blending a whole bunch of things not city-related (the charter that created San Francisco, including its modifications / amendments, state statutes in guise of the California Government Code, the California Constitution…) under an umbrella of “now City.” That is REALLY splitting hairs, and something like that could be done for every single CCC in the U.S., meaning we’re back to the assertion that “they’re all unique, there is no such categorization as a CCC.” That’d be a shame, because (rather roughly) we do assert there are such things.
It’s definitely a gray (grey) zone, but asserting that isn’t especially helpful answering “so, how do we tag them?” (so all of us are content).
And just like that (as in Connecticut), here is evidence SF County does exist (and should in OSM).
My current position, not strongly held, is that San Francisco is one territory. Hardly any effort is made by either the local or state government to distinguish the city from the county in general. You don’t see a lot of intentional references to the county as a geography. Some governmental departments may refer to San Francisco County its own special purposes, but we don’t map CAL FIRE districts as administrative boundaries just because CAL FIRE is part of the government. The court system is similarly an exception in its use of “San Francisco County”.
But this is not a one-size-fits-all solution. My other current position, strongly held, is that Orleans Parish is a territory distinct from New Orleans, if not spatially then semantically. If we must combine the Orleans Parish and New Orleans boundaries into one, then we would do the same to Washington and the District of Columbia, and we would we explain that Orleans Parish gets treated differently than East Baton Rouge Parish because the former doesn’t matter so much, like the District of Columbia doesn’t matter that much.
These two differing approaches can be justified legally if need be, but I would much rather just point to the boundary sign. After all, OSM has always had a difficult relationship with boundary mapping to begin with. The existence of boundary signs has been the most effective tool in quieting criticism of the boundary relations.
From where I sit, I see a bunch of cities that are one feature with the word “city” and “county” in their name. And, a name with many words in it does not magically make it two things. I see it as simple nomenclature, syntactic sugar around a pattern that mostly looks like:
admin_level=6border_type=city;countyname=(X)official_name=City and County of (X)alt_name=(X) County
If we are examining centuries-old history to contort ourselves into locating a second entity, maybe we can just step back and let the light bulb come on that these are really just single entities. If there are truly cases we can point to with a clear litmus test for two coextensive territories, then let’s trot them out.
While it is valuable to call this out, my current position (rather strongly held) is that OSM tags “what is” (to the best of our ability to do so, even amidst spirited debate!) We shouldn’t (necessarily) use “how much effort is made…to distinguish.” We should “simply” distinguish. To the best of our ability. And that apparently isn’t particularly easy.
It’s somewhat where the carving knife slices through the semantics: is it a two-headed (single) beast? Or is it two beasts?
I guess I really don’t understand how San Francisco City/County can be one entity but New Orleans/Orleans Parish and Washington/District of Columbia are not. Is it really just because of how the highway department chooses to sign them? Or California’s habit of naming counties after their largest city?
The way I see it, representing San Francisco as just admin_level=6 implies to me that its territory is not part of any admin_level=8 entity in the administrative structure of California. But that’s not true, if you stand on Market Street you’re standing in an incorporated city, the City and County of San Francisco, a city just like nearby cities like Oakland or Daly City or wherever. It’s not equivalent to being in, say, unincorporated San Mateo County or something. San Francisco is, as its official name plainly states, a City and County, so I think it should be tagged as both. I’d be amenable to admin_level=6;8 on the relation, and maybe that’s “the most literally correct”, but I don’t really see an issue with using two separate relations to represent the duality if it’s easier for the vast majority of data consumers to process.
Some see and agree to the existence of the duality (that makes more sense to them, and they have evidence to back it up), some do not see the duality and insist that only the City exists, not the County (and they have evidence to back it up).
I remind that CCCs are unique entities. Some days I feel like I have to both squint and stand on my head to see things properly.
What to do?!
1 Like
I hear you loud and clear: this is a geography project; it is neither a history project nor a government org chart. Never mind that a New Orleanian simply cannot discuss geography without talking your ear off about history.
Hypothetically, let’s say the critics of boundary mapping have a point – let’s ignore the classification system coming out of Washington, D.C., and rely solely on the on-the-ground rule and other bits of present-day ephemera that contribute to source=local_knowledge. In the following post, replace “township” with “parish”:
Replace the photo you found from Facebook with these eight Orleans Parish limit signs. (There are more Orleans Parish limit signs than New Orleans corporation limit signs.)
Replace the news clippings with ones such as:
Are people just using an alternative name for New Orleans? Kind of. I’m sure you can convince people to refer to the two geographies interchangeably. People say they live in “D.C.” and fly out of “Washington”. This doesn’t mean a geographic database can’t maintain coextensive boundaries for both. Heck, sometimes I say I live in San Francisco when I really mean San José. But even so, would people choose to refer to San Francisco County instead of San Francisco, except in statistics and government reports? Hecka no.
As I said, I’m not wedded to the idea of representing this hybrid as only one feature for the sake of database predictability. There’s precedent in how we sometimes map a hybrid POI as multiple POIs because we’re scared to break data consumers with semicolons in primary feature tags.
So data consumers can easily query for all the county equivalents in California via admin_level=6 instead of border_type=county, and all the cities and towns in California via admin_level=8 instead of looking for both border_type=city and border_type=town. These queries will even port over to Louisiana or New York State without any additional effort. I won’t deny that this is an advantage, but is it accurate? Should we put an asterisk somewhere to remind data consumers not to say something wacky like “the City and County of San Francisco is a city in the City and County of San Francisco”?
Well then, it’s settled. If it goes by a consolidated name, e.g. “City and County of X”, it’s one territory. If they have distinct names, it’s two.
1 Like
This feels like:
-
A hack that is a solution only to make a narrow interpretation appear to be true, or to ease downstream parsing (using a particular method, not necessarily a correct / accurate method, if there is such a thing), and
-
Something we’ll regret in the future and be forced to revisit to make adjustments the next time somebody realizes this was too short-sighted.
But I could be wrong, it’s just a feeling I have.
I’ve learned: beware of being “certain” about such things.
Our map saying that there is only one boundary for two distinct political entities (which we represent by tagging with admin_level) feels distinctly wrong to me. And I, too, have worked (and lived) in San Francisco.