On 10/06/16 at 10:46am, Greg Troxel wrote:
>
> Martin Koppenhoefer <dieterdre...@gmail.com> writes:
>
> (This is just a longer note about non-admin-boundary settlements and why
> they are particularly tricky in a lot of New England, sort of separate
> From the node/line/way discussion.)
>
> >> Il giorno 09 giu 2016, alle ore 18:50, Christoph Hormann
> >> <chris_horm...@gmx.de> ha scritto:
> >>
> >> If you can verifiably map a settlement as a linear way you can also map
> >> it as an area. Usually neither is the case so most populated places
> >> are mapped as nodes.
> >
> > I think that most settlements could be mapped as areas, there's often
> > a border where many people would agree that inside is the settlement
> > and outside is not
>
> In rural areas, I think that's true. But around me (mostly built up
> because of proximity to Boston), it isn't true and the borders are very
> hard to know. Even the people that live there will find it hard and
> disagree and tell you that the question doesn't quite make sense.
>
> As a concrete example, consider the "hamlet" of South Acton:
>
> https://www.openstreetmap.org/node/158813473
>
> This is not an administative boundary (which would be "Acton",
> admin_level=8), but a name for a populated place.
>
> This place name is longstanding, probably dating to before the American
> Revolution. It appears on USGS topo maps. It has a railway station of
> the same name. It is known to to the locals. I believe that the
> railway went there (~1846) and has a stop because the village was
> important.
>
> To those who think of South Acton as the historical village, it's clear
> that the buildings on either side of the railroad bridge are in the
> village. But as you go away from the center, it's very hard to draw a
> line. (There's a further complication that the intersection of 27/111,
> traditionally "Kelly's Corner", is often called South Acton, as it is
> more significant commerce-wise today.)
>
> There are many other examples, where what used to be a village with very
> little on the road to the next village is now a place where there are a
> clump of older houses among a sea of houses covering the whole town.
>
> Still, it's entirely reasonable to try to draw a polygon, as long as its
> done by the locals over beer. And also to have a node, which is far
> easier to place uncontroversially, as there is usually an obvious
> cluster of houses much older than the rest, and useful even if there is
> a polygon.
Agree - for me the big problem with placenames as just nodes is that it
then assumes that all areas are circular when doing reverse geocoding to
get the name of an area. But where I live this simply isn't the case -
For example Edinburgh Old Town looks like this:
http://nominatim.openstreetmap.org/search.php?q=edinburgh+old+town=1=
So for the various services that do the reverse lookups to work properly
we ended up dividing the city up into suburbs, all as areas. As has been
said, I agree this is really hard, and has involved discussions over
beer. But it does mean that at that level the classifications are much
better than with just nodes.
At that point, I discovered that some renderers like the cycle map layer
with render names for nodes and ways so we end up with duplicates :(
I did submit a pull request:
https://github.com/gravitystorm/openstreetmap-carto/pull/427
But it all got very complicated, but these changes never went very far,
see
https://github.com/gravitystorm/openstreetmap-carto/pull/546
Cheers
Chris
___
Tagging mailing list
Tagging@openstreetmap.org
https://lists.openstreetmap.org/listinfo/tagging
