tl;dr: I'm against a blanket rule when it comes to administrative
boundaries. They're really nuanced, and so should we.
On 2015-03-22 04:32, Serge Wroclawski wrote:
Imagine if Bob and Alice conflict on where a neighborhood boundary is
inside OSM. The issue escalates to an edit war and the DWG is called
in to resolve the conflict. Let's say that Frank is our DWG member.
How is Frank supposed to resolve the conflict between Alice and Bob?
Often neighborhoods don't have administrative recognition, or
administrative recognition is not in alignment with the people. I
imagine this would be especially an issue with neighborhoods where
lots of the under-represented populations live.
This is an important consideration. As I mentioned in a footnote
earlier, even a city with strong neighborhood organization can have
boundary disputes. However, the problem exists for administrative
boundaries in general, all the way up to admin_level=2 boundaries that
cut right across ethnic fault lines.
My point was that we should map neighborhood boundaries in cities where
doing so requires little editorial judgment, thanks to signage,
distinctive lamp posts, etc. And we are quite clear (via the tag value
"administrative") that this isn't the only way by which a community can
be delimited. As numerous threads have pointed out, the USPS has very
different ideas of location (ZIP codes), but that's OK.
When it comes to all our discussions around *administrative* boundaries,
I like this two-point test as a rule of thumb:
1. Are people or property governed differently on one side versus the other?
2. Is this distinction observable on the ground?
Municipalities generally pass both points. Congressional districts pass
#1 but not #2. CDPs generally fail both. School districts can be
observed, but not with the granularity required for mapping a boundary.
City neighborhoods may pass one, both, or neither. Maybe all the locals
you interview can agree on the name of a neighborhood but not its shape
-- in which case it should be nothing more than a POI.
Which brings me to Serge's other point:
First, there are a growing number of people who believe that
administrative data is very useful, but breaks OSM's "ground
observable" rule. That is, someone who is present on the ground should
be able to observe the data in OSM. It's usually not possible to do
that with administrative boundaries.
SteveA has responded more forcefully on this point, and so have I in the
past. [1] Fortunately, Alice and Bob's disagreement sounds pretty
clear-cut. If the city didn't go through the trouble of demarcating any
part of the boundary in some way, perhaps the general public shouldn't
expect OSM to reproduce their two neighborhoods' boundaries at all. But
I see no reason why such a decision would impact boundaries with very
different characteristics.
-*-*-*-
Serge's focus on verifiability relates to a boundary I've spent a lot of
time on lately, so I'm going to go way over my word limit.
Last month, I reminded this list that state borders along the Ohio River
actually follow the river's historical northern bank, not its
present-day thalweg or centerline. [2] Even if you send a diver into the
river, there isn't always going to be a natural feature to verify OSM
data against. We have a few options:
1. Try to be as accurate as possible by tracing USGS topo maps. Treat
these borders as a practical exception to the on-the-ground rule. Use
the source tag rigorously.
2a. Conflate the state borders with the current thalweg. We'd give Ohio
and Indiana various islands and dams that actually belong to Kentucky
and West Virginia, ignoring the Supreme Court ruling. We'd be putting
intentionally inaccurate data into OSM.
2b. Conflate the state borders with the current northern bank, siding
with Kentucky and again ignoring the Supreme Court ruling. We'd give the
entire river to Kentucky and West Virginia, including riverboat casinos
that keep to the Indiana side but are illegal in Kentucky.
3. Omit the river boundaries but leave the rest of the state lines
intact. This approach introduces technical problems like broken
multipolygon relations and just confuses people. Where does West
Virginia end?
4. Omit the entire boundaries of states that border the Ohio River.
It'll look like a mistake, so people will helpfully and sloppily add the
boundaries back in.
5. Omit all state lines, everywhere, throwing away lots and lots of
fixup work done with care by volunteer mappers. And all because Kentucky
wanted the whole river.
Everyone agrees the river is the boundary, just not what "the river"
means. In this case, I say we hold our noses and go with #1 as the most
accurate, least disruptive approach. [3]
[1]
https://lists.openstreetmap.org/pipermail/talk-us/2013-January/010162.html
[2]
https://lists.openstreetmap.org/pipermail/talk-us/2015-February/014307.html
[3] https://en.wikipedia.org/wiki/WP:POINT
--
m...@nguyen.cincinnati.oh.us
_______________________________________________
Talk-us mailing list
Talk-us@openstreetmap.org
https://lists.openstreetmap.org/listinfo/talk-us
