[Imports] Oahu, Hawaii CDP boundaries

[stevea]

On Sep 15, 2019, at 10:40 AM, Brian M. Sperlongano <zelonewolf at gmail.com> wrote:
> Thanks Steve, this is good background.  I agree with most of this.

Hi Brian:

Great, you are welcome!  (Fair warning, lengthy reply ahead).

I'll reiterate salient (OSM-based, not Oahu-based) points I hope we agree on, to make sure:

1)  CDPs are statistical boundaries.  While they do "snapshot something in time," they aren't the best data to enter into OSM, especially as they are frequently confused with truly administratively governed areas (CDPs are no such thing, though they often lend a name / identity to an area, perpetuating this confusion).  They can enter (and sometimes have entered) OSM where they make sense to do so (e.g. Alaska's enormous Unorganized Borough, "sense" here largely being thanks to boundary collaboration work between the Census Bureau and Alaska state government).  CDPs are properly tagged boundary=census and never given an admin_level=* tag of any value.  Most CDPs are likely better entered (at least initially) as a node tagged place=[town, village, hamlet, isolated_dwelling].  Similar things are true about ZIP codes, though more strongly:  ZIPs have no boundaries (if they did, they would be tagged boundary=postal_code, but they aren't in the USA) and they certainly are not administrative (or tagged admin_level) in any way (though again, post offices have been known to lend a sense of identity to especially rural areas with a name).  All that said, and because I believe that you have fresh CDP data, I'd prefer (and I believe OSM prefers) that a better OSM representation of these would be as nodes as I describe them above, not as polygons tagged boundary=census:  while the latter are technically correct, they instantaneously drift out of correctness, a major reason why these aren't the best data to enter OSM.

2)  In (incorporated) cities (in the USA, in OSM), we tag the values place=[borough, suburb, quarter, neighborhood] with rather specific rules, though these are not always followed.  See, e.g. our suburb wiki, but roughly stated:

a borough is a distinctly administrative unit (sub-city-level government, like a borough council) of a city,
a suburb is a part of a city (or rarely, town) "with a distinct and recognized local name and identity,"
(often best mapped with a node, unless "the (suburb) boundary is clearly defined")
a quarter is "for areas smaller than a suburb and larger than a neighbourhood" and
a neighbourhood is "a specific area within a place=suburb."

These are often "less appropriate" tags in USA, but they do exist here.  I don't know if this sense of "hierarchy within a city / large conurbation" logically "fits" Honolulu, but if these tags DO fit there, please use them.  Some rather careful thinking about the actual method by which Honolulu locals do this (as well as the government structure, I understand it is rather complex as mentioned in my previous posts) is absolutely required.  If these tags fit, they should be used because there ARE distinctly administrative units or there ARE areas smaller than a suburb and larger than a neighborhood — in Honolulu.  You and locals, not me, have better knowledge to make such determinations.  Though, government (structure) does assert its existence (with council meetings, funding, services, sometimes signage and so on).  These are real and I believe should be well-mapped in OSM.

3)  There are also place tag values in OSM different than those in 2), like place=[town, village, hamlet, isolated_dwelling].  The major distinction is that these "stand alone," that is, they do not subordinate to a larger conurbation of a city (or rarely, town).  Again, I am not personally aware of Honolulu's subtitles in this regard (you and locals, more so) but if the "glom" (conurbation, technically the consolidated city-county  / CCC of Honolulu) INCLUDES the various villages/hamlets/suburbs/neighborhoods — whatever they are — within Honolulu, then paragraph 2) should guide tagging (subordinating), paragraph 3 less so.

There can be a sticky problem here:  the government (which ostensibly both represents and expresses the sense of organization of administrative structure) might hew to one "flavor," (I'm guessing a complex hierarchy of city-county, hence paragraph 2-style subordinating suburbs and neighborhoods) while "locals" (the People) might hew to another (maybe the paragraph 3 flavor of seeing things, where "we aren't part of HONOLULU around here!" even though, technically, administratively, they are).  I believe I read that in your reply.

Yes, this can be (in this case, likely is) complex and often takes many words to untangle — I see no way around that.  Yet, loquacious as I am often accused of being, it remains true that I wish to see accurate tagging in Honolulu as we discuss these issues.

I don't wish to ruffle feathers by saying so, as I appear to come into direct contradiction with you as I do so, yet exactly this sort of "sticky problem" seems to have emerged here:  there really IS a CCC which is both "the county of Oahu" and "the city of Honolulu" which share a coterminous boundary.  As there are 39 other CCCs in the USA, and for years OSM in the USA has wiki-described (after much talk-us and backchannel consensus) and agreed that two relations tagged boundary=administrative, the former with an admin_level=6 tag and the latter with an admin_level=8 tag are "the correct OSM method" to map this, I continue to believe this is a "technically" correct solution (speaking, say, as a political scientist, which I am not).

ADDITIONALLY, I believe it is in the best interests of OSM (especially Hawaiians!) to also capture the semantics you describe as (roughly) "Honolulu is the urban area on the SE coast" (of Oahu).  The existing relation you mention (119231), especially as it seems to have derived from CDP data (e.g. border_type=census, source:population and all the tiger: tags) simply seems wrong in a number of ways.  For one, "border_type" is a deprecating tag, sometimes accurate on maritime boundaries when used with specific values, but most of the time on land areas, it is simply a confused mess (as it appears to be here).  Secondly, while the admin_level=8 tag does its best to say "city" (when what we might better mean to express is "urban area as understood by locals"), without a boundary=administrative value to combine with the admin_level=8 tag, it seems either meaningless or only meaning something in a very specific context (Honolulu's), though what that context is isn't well-specified.  And I hesitate to re-invent on a case-by-case basis what OSM means to say when we tag a city, starting with Honolulu.  I do remain in listening mode to better solve this, though.

It may be that some "early work" ("throwing darts at the board rather than nailing down exact details") can further the good data we all wish to see enter our map.  As I've said, this might be that we sharpen focus on relation/119231's tags, removing border_type and admin_level, maybe leaving boundary=census, maybe deleting it, I don't know.  (I'll remind that the actual polygon is almost certainly out-of-date, though it can serve as a rough guide as to "what locals mean as a dated approximation of the urbanized area, though likely not exactly correct today.")  "Throwing darts" might also include adding nodes to the non-urbanized areas of Honolulu with place=village (or even place=town?) tags, giving them a sense of a "more independent, less part of Honolulu" identity.  Technically, a Honolulu boundary=administrative, admin_level=8 tagged coterminous polygon with Oahu's admin_level=6 is something I'd like to see eventually, even as these additional place=village (for example) nodes continue to co-exist within it.  (This makes them suburbs, but I don't with to digress).  But I hear you loud and clear that place=neighborhood "feels inappropriate" on Kaneohe, even as it might be exactly accurate tagging for OSM!

> I can't buy into the proposed handling of the current Honolulu level 8 relation (119231).  Replacing this relation with one coterminous with the current level 6 county relation would totally break the local understanding of the difference between "Honolulu" and the political entity "City and County of Honolulu".  Anyone that lives in Kailua, or Hale'iwa, or Kapolei would tell you that they definitely live in a different place from Honolulu, which a local understands to be the urban area on the southeast coast stretching from the ocean to the Ko'olau crestline, and from Halawa Stream to Makapu'u Point.  That's the boundary that a local would expect to see when searching for Honolulu.  Other places on the island simply aren't part of Honolulu despite being serviced by the C&C government.  Waianae or Kaneohe or Mililani are NOT neighborhoods of Honolulu!

We seem to be largely in agreement (I would find it helpful if you replied with whether you agree or disagree with my paragraphs 1, 2 and 3 above).  However, we do seem to disagree (slightly) on one point:  you say that (for Honolulu) "admin_level 8 isn't right because it's not an administrative boundary" (as you say "you agree," but I don't know who you are agreeing with or in what context).  I'd like to politely say that "the City of Honolulu" (as a distinct entity of the CCC — the "city" portion) IS exactly this, even as it is coterminous with the county of Oahu (island, plus a lot more — all the Northwestern Islands, too!)  YES, I can certainly see how locals say "don't call Kapaeloa or Dillingham Airfield part of Honolulu — WE don't!" yet technically, politically they are (as is Green Island Airstrip, absolutely wacky as that seems, being ~2300 km from Honolulu!).

> IMO, there needs to be some way to mark the real-world-use boundary of Honolulu.  I do agree that an admin_level 8 isn't right because it's not an administrative boundary.  Perhaps there is some other combination of tags that can fulfill this purpose.  Defining "Honolulu" by the C&C boundaries may be pedantically correct by the definition used on the mainland but it's wrong on the ground.  A place marker alone is unsatisfying because there are clear-cut boundaries that exist and are understood.  The county is not a CCC in the way it's understood in other places because there's NO municipalities in Hawaii.  Therefore, there should be NO level 8 boundaries in Hawaii, period.  Coterminous boundaries make sense in other places because they have both CCCs as well as standalone municipalities that are sub regions of counties.  "C&C of Honolulu" is a naming convention rather than a merger of levels of government that are otherwise separate in other places in the state.  If they had chosen the name "Oahu County" instead and changed nothing else, we'd all agree that this is a level 6 boundary only.

What to do?  For now, I listen.  Loudest in my ears now is your statement "there needs to be SOME way to mark the real-world-use boundary of Honolulu" (as I understand you and other locals mean this, "the urbanized SE coast of Oahu" (stated roughly and quickly).  I agree with you there.  I do not (right now) know a way to capture this in OSM.  I believe I do know (and have stated) how I believe the CCC should be entered into OSM, and I continue to believe it should be.  I also believe that OSM might have to evolve some tagging, or a tagging approach, to better solve this.

SteveA
California