[OSM-talk] Not attaching polygons to roads

Minh Nguyen minh at nguyen.cincinnati.oh.us
Sat Feb 22 23:44:50 UTC 2014

On 16:02 2014-02-21, Bryce Nesbitt wrote:
> On Fri, Feb 21, 2014 at 1:29 PM, moltonel 3x Combo <moltonel at gmail.com
> <mailto:moltonel at gmail.com>> wrote:
>     I agree with the "matter of taste" argument insofar as I dont complain
>     to mappers who initially glue areas to lines. It's just data that can
>     be improved like any other, and if it "tastes easyer" to that mapper,
>     it's fine. You really shouldn't force anybody to be more accurate than
>     they care to be.
> I think it's important to consider that what we put into OSM is a model
> of the real world.
> Roads for example are generally not straight, yet we model them with
> line segments.
> A model.
> To say that "the park occupies the space between these four streets" is
> a very reasonable first approximation model.  You can go pretty far with
> the model: the street has cycle tracks left & right, the utilities are
> on poles overhead, the park edge is fenced.  All the important data is
> present and can be conveniently rendered at any scale or with any emphasis.
> This is very flexible.  Cartography rules often render road width not
> based on physical width, but logical width.  The 25' wide highway at one
> edge of the park is more important than 30' wide residential street on
> the other three sides.  The model handles this just fine: whatever width
> is not used by the road is used by the polygon.  Nobody cares that the
> park just lost a little space, the map looks great and communicates
> clearly to the viewer.
> Flip to a cycling map, and the 30' bicycle-friendly street may be more
> important than the highway: it still renders fine.
> ---
> It's the micro-mapping that brings up hard to process situations.
>   Imagine that same area micromapped:
> road centerline, cycle tracks, power poles, fence, hedge, imported legal
> property boundary of the park.   To render that well for various needs
> you've got to start shoving and pushing element.  To render the highway
> "fat" you need to push out the cycle track and fence lines.  Should you
> also shove over the hedge or just bury it under the roadway?  It's
> unclear what's best.
> Very often the actual legal boundary does not correspond to the
> landscaped boundary.  The park may be landscaped right to the edge of
> the tarmac, but the highway department actually owns a wider swath of
> land.   The fence might be within the legal boundary of the park, or
> outside it. Which one you map depends on your aim:
> the assessors office wants the legal boundary,
> the soccer team will play right to the edge of the fence regardless,
> they only care if the fence is permeable to
> soccer balls.
> ---
> There's no one solution here: in micromapped areas road centerlines are
> not enough.  In many areas (I might hazard 99% of the surface of the
> planet), the /model/ may serve the need better.
> I think that strong tool support for sharing nodes is good, appropriate,
> and a great for first efforts at mapping modelling an area.  The node
> sharing is particularly useful for administrative and other areas that
> do in fact follow a road centerline by fact or convention.
> Just realize that there's disagreement on this point in part because of
> valid differences in scale, scope and aim.  And that we model reality
> because models are often more useful than a direct representation.  Any
> difficulty in editing is a tool issue.

Bryce's entire message resonates with me due to mapping and remapping 
the same suburban areas many times over, each time with increasing 

In most of my mapping area, city limits, subdivision boundaries, and 
individual residential lots all extend to the road centerline by 
default, except perhaps when the road is a divided highway, in which 
case landuse=highway is appropriate anyways.

A residential subdivision here will often place a decorative fence or 
hedge along the road, with a sidewalk in front of it, but the 
subdivision maintains everything up to the curb, where the pavement ends 
(and still owns half the land under the pavement). In other words, the 
subdivision doesn't end at the fence or the sidewalk. Accurately 
"ungluing" the area from the roadway means mapping the curb.

I'm not opposed to curb-mapping -- it's probably quite useful in urban 
areas, for wheelchair accessibility -- but in my area, we're just very, 
very far away from the point where we can ask mappers to worry about 
that level of detail.

Plus, Potlatch makes it very easy to draw an area that follows a windy 
stretch of road: press F repeatedly.

minh at nguyen.cincinnati.oh.us

More information about the talk mailing list