[Talk-us] Talk-us Digest, Vol 116, Issue 20

OSM Volunteer stevea steveaOSM at softworkers.com
Tue Jul 11 20:31:27 UTC 2017

Frederik Ramm <frederik at remote.org> writes:

> I think there might be a misunderstanding here and I would like to chip in...

Thank you for chipping in, Frederik!

> If you have an admin_level 4 entity - like a state - then the boundaries
> with admin_level 4 are the outer demarcation of that, i.e. they separate
> the area where the state is responsible from the area where the state is
> not responsible.

Right; I'm with you so far.

> The only reason to have an admin level 4 boundary inside a state, would
> be if there was somehow a piece of *federal* territory inside the state.
> Only then would the state have a "hole" in it that would be tagged with
> admin level 4! An area inside the state that is state-governed because
> of a lack of admin_level 5+ entity does not need its own boundary. It is
> defined by the boundaries of the admin_level 5+ entities that surround it.

OK, I'll take your word for it.  But I ask you to please further clarify that in that first case (where you say *federal*):  is it more correct to say "anything above state?"  That is, not just admin_level=2 as *federal,* but 3 or 1 would be included, too, if there were such things.  Right?  BECAUSE they are "above."  (A lower admin_level value).

>> without using a multipolygon relation, 
> You will be using boundary relations which are practically identical to
> multipolygon relations. Any attempt to create a "lower 48 states"
> polygon without relations would hit the 2000 node limit.

Yes, I know this too, as I have built several multipolygons which needed more than one segment because of our 2000-node limit (e.g. various large state parks and national forests here).  I was offering this example not to demonstrate my knowledge of multipolygons (or lack thereof, though I don't think so as I'm facile with them) but rather I try to gain clarification about the "relative levels" of admin_level when using (boundary, as you say, not multipolygon) relations.  It's both the numbers (values) on the admin_level tag as well as the geography of "holes" in these blending of geometry that was unclear.  You are clearing it up for me, so thank you.

>> is it correct within OSM to tag, say a very large "lower 48 states" polygon with admin_level=2 AND ALSO tag admin_level=2 on, say, a national_park inside of it
> That would only be correct if the national park was *not* part of the
> lower 48 states but somehow part of another nation.

Again, I have to take some time to fully parse/understand this, but I think I get it.

> I'm not 100% sure what you want to achieve but think of it like coloured
> polygons. If you have an admin_level 2 area for the USA, think of that
> as one colour, and then you have a lot of states, each with a different
> colour. In those areas where the "USA colour" shines through, because
> they're not covered by any state, that's automatically federal territory
> and you do *not* want an admin_level 2 boundary surrounding that
> (because then not even the "USA colour" would shine through, there would
> be nothing there).

Ah, yes!  Your wording of "shine through" is very helpful!

>> Guidance by knowledgable people with real answers might guide us on a number of these situations, not just "Gores" (et al) but other kinds of "hole" tagging without multipolygons.  
> If you mean not only "without multipolygons" but "without boundary
> relations" too then I think you should stop right here and leave it to
> people who can work with relations.

Well, I have worked with hundreds, maybe thousands of relations, so as I say, I'm facile with them.  I think I misspoke/mistyped regarding multipolygon relations (with specific role tags to "punch holes" with inner/outer) and type=boundary relations, which ARE the precise data structure we are talking about here.  Good of you to point that out; I'll be less confused in the future both in my mind and as I might ask questions here.

My major thrust with US_admin_level has been to document what the USA believes should be the correct value of the tag on specific jurisdictional entities, not so much the geometry.  So, thank you for clarifying my understanding of the geometrical aspect of type=boundary relations in that context, it does help very much, especially your "shines through" wording.  (Because of the way that each of our 50 United States can be and are different than each other, I believe it is important to document these differences so we tag correctly.  It has gotten better and we are closer to a finish line).

I am absorbing the importance and implication of what Kevin is saying, too:  it might seem hierarchical here, and while we are doing yeoman work to shoehorn as much of it as we can into a hierarchy (and we're doing OK, maybe even pretty good, but certainly not perfect), there ARE exceptions, and there will be places where it simply breaks down and we can't fit parts of the USA into a hierarchy.  I do accept that, but I continue in my quest to do what we can where we can to utter (tag) our very best where we can do so.

Again, thank you for great dialog and the additional clarification,


More information about the Talk-us mailing list