[OSM-talk] Crossing ways

Mike Collinson mike at ayeltd.biz
Sat Aug 11 17:17:32 BST 2007

Looking at http://discord.ensued.net/crossingways1.png, I'd expect to see the minor road physically divide into three divide into three link roads if I drove along it, two of which (implicitly) cross the dual carriageway by bridge or tunnel.  I suspect, forgive me if I am wrong, that instead you are trying to model the logical arrangement dictated by traffic lights and signs to show that U-turns are not allowed?  I don't think you can do that with the current OSM model; as indeed you suggest, some sort of extra tagging would be required.

Assuming the real physical layout of the junction is simple (?), I personally just show the minor road connecting to a node on the nearest carriageway and carrying on to finish as a node on the second carriageway.  That would then render correctly but loses information about turn restrictions for future navigation/routing software.  There has been debate on this issue, but no consensus reached on a solution that I am aware of.


At 03:21 PM 11/08/2007, Jon Bright wrote:
>Now that I've found the validation panel in JOSM, I have a question 
>about crossing ways.
>I have a dual carriageway, which has a junction with a minor road. 
>Traffic on the minor road can join the dual carriageway in either 
>direction.  Traffic on either part of the dual carriageway can join the 
>minor road.  Traffic on the dual carriageway cannot perform a U-turn.
>I'd tried to represent this with the ways in the screenshot at
>The validator complains that the selected ways cross one another.  There 
>are no bridges or tunnels involved here - the whole thing's controlled 
>with traffic lights.  My question is, what to do in this situation:
>- Should I have organised the junction differently?
>- Should I just ignore the validator's warning?
>- Should I put set one of the ways as layer=1, even though they're not 
>really above one another?
>- Something else I didn't think of?
>My current strategy is to ignore the warnings, but this gradually tends 
>towards making the warnings about crossing ways no longer be useful. 
>One other strategy I thought of was something like a "highway=crossing" 
>tag - something along the lines of "this way isn't really a way per se, 
>but does provide information to routing systems about the allowable 
>routes between these other ways".  Renderers might then choose to ignore 
>such ways.

More information about the talk mailing list