[OSM-talk] Not attaching polygons to roads

Peter Wendorff wendorff at uni-paderborn.de
Fri Feb 21 10:44:51 UTC 2014


Am 21.02.2014 10:44, schrieb Pieren:
> On Fri, Feb 21, 2014 at 9:19 AM, Peter Wendorff
> <wendorff at uni-paderborn.de> wrote:
> 
>> If the other mapper shares nodes between the road and the field, and the
>> field is surrounded (and tagged as such) with a fence, so the field is
>> e.g. landuse=farmland, barrier=fence, then this is an error in the map
>> as it states that the fence is in the middle of the street or it's not
>> possible to decide where relative to the street the fence is.
> 
> That's maybe an original issue with landuse polygons. Once you go to
> this level of details (fence), the border line between landuse's is
> not a clear sharp line. As suggested by Shaun, once you add fences and
> detach the farmland, you should also fill the gap created and make the
> landuse=highway. 
> People detaching landuse from road lines are most of
> the time doing half of the job.
I may agree here, but in OSM I think doing half the job is better than
mapping wrong stuff.
OSM is a database, not (only) a map, and there isn't something like
"once you go to this level of details".
Let's extend the example slightly:
Let there be from left to right a field, a fence, a street, a hedge and
a park.
Let the fence in addition be not around the field, but only at the
borderline to the street (so it's not a tag on the field polygon any
more, while the hedge surrounds the park (where entrances are mapped as
such on nodes).

Now Mapper A starts mapping with low detail from aerial imagery: he
draws a polygon for the field, another polygon for the park, and a way
for the street, and tags it as landuse, leisure and highway
respectively. He omits the fence and the wall.

As you said, this is perfectly valid (although it's a little bit ugly to
detect that it's not a park directly beside a field, because you would
have to create the corresponding buffer for the highway for that; it's
not possible to calculate the exact area of the field, as we're wrong by
6 meters for half the street width.)

Mapper B is on the ground a while later and recognizes that there's a
fence and a hedge.
Adding the hedge seems to be easy: she adds the barrier=hedge-tag to the
park.
Adding the fence is easy, too, but how to do this? She definitively has
to draw a new way as there's no geometry matching the fence. But where?
By the "rules" applied in this scenario up to now it would be fine to
add the fence as a way sharing the nodes that are already shared by
park, highway, hedge and field.
But what happens when doing this?
There is a set of nodes that is hedge and fence. Might be possible: I
would interpret this as a fence inside a hedge, which is possible and
well known in the wild. But what's the matter with the highway? well...
then the highway must be in between the hedge with fence and the field
with the wall...
Well - wait... it could be a fence on top of a wall instead - now the
fence is on the other side of the way...
Or it could be a fence in the middle of the street - strange...

In fact the map says, there's a fence, a wall and a street's center line
at the same position.
Independent of the level of detail I would assume for the application
this is simply wrong, and keep in mind: the coordinates are in a level
of detail of 10cm or better, with no way to see what level of detail is
ment by any particular mapper with any particular object.

Let's invite Mapper C. He - as you suggests would like to clean up the
mess produced by A and B, and it's going to be hard work.
Without being on the ground it seems to be possible to detach the park
with the hedge from the way on the first glance, but damn it - what to
do with the wall? Isn't it wrong to detatch the fence if it might be
possible that in fact the fence is on top of the wall? If so it would be
necessary to detatch the fence, too and let fence and wall share nodes.
If not, this would create a different but completely wrong situation.
So nothing can be made better without being on the ground.

C decides to take his bike and and ride a whole day to that place as
unfortunately there's no active mapper any more (A and B stopped mapping
months ago). Now he knows what's the situation on the ground and has to
detatch lines from each other, stumbling over several ugly issues in the
osm editor software available:
- How to select one of many ways sharing the same nodes?
- how to minimize breaking object history - but remapping everything
would be more simple to do.

If A and B would have drawn lines in parallel, leaving a gap for the
highway in between it would be much easier.
>> Changing the way of mapping without adding value/improvement (!) is
>> not okay.
> 
> At least, new contributions shouldn't decrease the quality.
+10,
but filling the map without empty gaps isn't a good measurement of quality.

regards
Peter



More information about the talk mailing list