[OSM-talk] cadastre, ways, rendering
frederik at remote.org
Mon Jan 22 23:42:18 GMT 2007
> could you elaborate on that? What in particular are non-contiguous
> ways? (Apparently you can put roundabouts in between, and they are
> still contiguous). Why are they bad? ...
In the OSM data model, a way is simply a number of segments thrown
together. You can take one segment from Southern Spain and one from
Norway and make a way from them.
Of course this does not make sense. Some people, me included, maintain
any OSM "way" should only contain contiguous segments (i.e. you must be
able to visit all nodes in the way by only travelling on segments that
are also in the way).
Many people come to OSM with a semantic model in their heads that goes:
"If it is the same road, then it should be the same way". But roads are
not alway contiguous, for example you may have an intersection that
looks like this:
If you want to travel along the horizontal road (assume it is the same
road), you will actually have to travel a bit on the vertical road. If
you put both parts of the vertical road in the same way in OSM then you
have a non-contiguous way. The same is the case with a road when there's
a roundabout in between.
This, however, creates problems with rendering, because wen drawing a
way on the map, many renderers will try to write the name of the way
along its length and have trouble with "gaps". Also, considering
potential later routing applications, it is a nice thing if ways are
contiguous (traversable), therweise you will have to divide each way
into traversable subparts.
So I (and others) strongly advise to make individual ways in such cases.
This means that you may have five ways all having the same name because
they are in reality the same road. No big deal; you have to do that
anyway if attributes change. Since OSM cannot manage properties that
only apply to a part of a way, you will have to split your way if the
road suddenly changes from being a primary road to being a secondary
road (even if it is still the same street with the same name). Or if it
changes from being one-way to being both-ways.
Frederik Ramm ## eMail frederik at remote.org ## N49°00.09' E008°23.33'
More information about the talk