[OSM-talk] Area-type objects and ways along its boundaries

[Nathan Edgars II]

Ed Avis wrote:
>Nathan Edgars II <neroute2 <at> gmail.com> writes:
>>Depends what the way is. If it's a street, the area most likely stops
>>at the right-of-way line, and does not extend to the middle of the
>>street, so it would be incorrect to extend the area into the street.
>>On the other hand, if it's something like a fence or wall that forms
>>the de facto edge of the area, it probably should be connected (if not
>>using the same way with a relation for the area).
>
>I agree with this, and would add that for 'pedestrian areas' such as squares
>or plazas, which are often mapped as an area tagged with highway=pedestrian, it
>makes sense for the area to extend up to the surrounding road, since you can walk
>all the way to the middle of the road or even cross it.

Or, even better, draw the sidewalks and connect it to those.
>
>>If the area is an administrative boundary (or a useless
>>census-designated place), it shouldn't be joined to the way, even if
>>it's drawn along it. Boundaries (at least in the US) are usually
>>defined such that neither minor nor major realignments of highways
>>will affect them.
>
>That may be so, but if the boundary is defined to be in the same place as the
>current line of the road, and the road is slightly inaccurate in OSM, then you
>would expect that adjusting the road position by a few metres to reflect more
>accurate survey data would also adjust the boundary.  If there were roadworks
>and the road were changed in the real world, then of course it would need to be
>moved separately from the administrative boundary.  But that is a fairly rare
>event, whereas minor tweaks to object positions in OSM are quite common.

It's extremely common in the US that imported TIGER data will have a
road in the old position, which happens to match a boundary (or
parallel it, with TIGER having it erroneously on the road), or that
TIGER will make the same mistaken assumption you are that the boundary
is on the road, rather than where the dusty predecessor was 300 years
ago. It is significantly easier to fix these errors when the
boundaries are not improperly joined to roads. It's also easier then
to download the boundaries in JOSM with a XAPI query and be able to
fix them up without worrying about messing up highways.
>
>Similarly, if a boundary is defined to follow the course of a river, I would
>expect it to share the same nodes as the river's centre line, so that tweaks
>to the river's surveyed position would be reflected in the boundary. Otherwise
>you end up with a river and a boundary line that criss-crosses it, which can
>be confusing.
>
No - you'd have a boundary between the two river banks. Now if the
boundary is defined to lie on one of the banks (mean high water, I
believe) and we use the same definition for drawing the bank, it would
make sense to use the same way.