Re: [Talk-us] National Forest boundaries (Mike Thompson)
On Jun 21, 2020, at 5:58 PM, Mike Thompson wrote: > 1) Not all "inholdings" are completely surrounded by the National Forest, > they are "bites" off the edge in some cases. I don't think one can have an > inner ring and an outer ring which are at all coincident (they can't share an > edge) and still have a valid multipolygon. I don't wish to sound dismissive as it seems we largely agree, but this is merely quibbling over constructing an outer edge. If truly a "bite off the edge," then it seems the outer polygon should shrink to accommodate, no need for edge mumbo-jumbo (though sometimes these edge memberships take on a tagging life of their own and it gets to be a high-wire act as to how "loaded with tags" each one might become). There are a lot of methods to capture semantics using syntax that is crisp and unambiguous, I believe some methods are smarter (less or even no ambiguity about the semantics that are "meant") and cleaner (fewer data) than others. There is what might be characterized as "a wide zoo of tagging in these realms" (nationally at the enormous polygon scope). Thank you (again) to Kevin for the word "menagerie" here. This also enters what some have dubbed "higher math" (multipolygon edge tagging in a relational database and how deep these semantics can be relied upon are a "topology of deep genus"). > 2) Holes (inner rings) are not part of the polygon. Thus if one did an > analysis of (for example) a series of points, any points that fall in one of > the holes would not register as being inside the multipolygon, even though > they are inside the outer ring. That sounds right. And a snappy-efficient way to achieve what is "truly excluded as PRIVATE" as an inner member of an "outer polygon that describes geographic extents of this PUBLIC forest" is by simply tagging what IS the inner member with "what it is." That might seem fancy word salad, so I want to break down what we say as understandable to both of us. We're using the double-duty that in a public forest (with a large, enclosing geography, but no larger than necessary or truly) which is tagged as outer role, anything we tag inner role is EXCLUDED from the public forest. That "inholding," in every sense I've ever seen it, is private, hence, it's an inner to the outer, as private isn't public and vice versa. The logic of opposites, the power of roles (inner and outer) in a relational database and the geography (in 2-space) of what we "mean" (quite intentionally) by inner and outer become powerful. Let's simply tag the "thing inside, different from its enclosing polygon and so excluded from that polygon" for what it is, then include it as an inner member of the relation. We do this, the logic of "nodes registering or not" is already built into "the space." Think of a grassland in otherwise-land-filled-with-trees, it's a (mathematical) "hole" (of the 2-space, lat-long of nodes OSM lives in). It simply works, like math, geography, software. (Um, "well written" software!) Meet you off list? Steve ___ Talk-us mailing list Talk-us@openstreetmap.org https://lists.openstreetmap.org/listinfo/talk-us
Re: [Talk-us] National Forest boundaries (Mike Thompson)
Steve, Perhaps I am not understanding what you are saying, but: 1) Not all "inholdings" are completely surrounded by the National Forest, they are "bites" off the edge in some cases. I don't think one can have an inner ring and an outer ring which are at all coincident (they can't share an edge) and still have a valid multipolygon. 2) Holes (inner rings) are not part of the polygon. Thus if one did an analysis of (for example) a series of points, any points that fall in one of the holes would not register as being inside the multipolygon, even though they are inside the outer ring. Mike On Sun, Jun 21, 2020 at 6:39 PM stevea wrote: > Continuing from my previous post, we even have an especially data-compact > (efficient) way of representing that: the member of the forest relation > which is an inholding (tagged with role inner) IS the polygon of, say, a > private residence "inside of" the forest. > > For example (I'm making this up), say we have a national forest with a > small shopping area (for food, supplies...) near its center for > convenience. I could see one polygon (tagged landuse=retail, name=ABC > Forest Shopping Center) both BEING exactly that, AND being included in the > (enclosing) forest multipolygon as a member tagged "inner." Voilà, > double-duty and done. > > SteveA ___ Talk-us mailing list Talk-us@openstreetmap.org https://lists.openstreetmap.org/listinfo/talk-us
Re: [Talk-us] National Forest boundaries (Mike Thompson)
Continuing from my previous post, we even have an especially data-compact (efficient) way of representing that: the member of the forest relation which is an inholding (tagged with role inner) IS the polygon of, say, a private residence "inside of" the forest. For example (I'm making this up), say we have a national forest with a small shopping area (for food, supplies...) near its center for convenience. I could see one polygon (tagged landuse=retail, name=ABC Forest Shopping Center) both BEING exactly that, AND being included in the (enclosing) forest multipolygon as a member tagged "inner." Voilà, double-duty and done. SteveA ___ Talk-us mailing list Talk-us@openstreetmap.org https://lists.openstreetmap.org/listinfo/talk-us
Re: [Talk-us] National Forest boundaries (Mike Thompson)
Mike Thompson wrote: > One polygon for the administrative boundary of the NF which was established > by Congress. > Zero or more polygons describing limitations on access (no need for polygons > to for access=yes, we can assume that in a NF generally), whether they be due > to private ownership, or other reasons. > The above are two separate concepts, so it is ok to have two separate OSM > elements, in my opinion. > A relation for all would be ok too, as long as the private inholdings are not > removed from the NF (which I think has been done in some cases). If I'm not mistaken, we already have the machinery to do that with how we build multipolygons. To wit, a single multipolygon (well tagged as to name of national forest, protect_class=6, ownership=national...representing the forest) has one or more polygons with role "outer" where all those tags apply and one or more polygons with role "inner" where there are inholdings and "something else, not national forest" are, and the tags on this multipolygon do NOT apply. (These appear as "holes" in the usual way inner members do in a multipolygon). There is nothing stopping us (and sometimes we do) from adding additional polygons that are "coincident with the holes" which represent "what that particular inholding is." It seems to me THOSE are the places where any access tagging (if necessary) might apply, should your fancy run to tagging those specificities. We've been tagging "large public areas with inholdings" like this (using multipolygons with inner members) for as long as OSM has had multipolygons. Why might we (re-)establish "two separate concepts" in two separate data structures when we already achieve this with one data structure (and possibly others, by that I mean "one multipolygon representing the forest, which might have inner members," while noting that ADDITIONAL polygons can describe what the inholdings ARE and superimpose on top of the holes represented by the inner members? Am I missing something? SteveA ___ Talk-us mailing list Talk-us@openstreetmap.org https://lists.openstreetmap.org/listinfo/talk-us
