[OSM-talk] Crossing ways

David Earl david at frankieandshadow.com
Sat Aug 11 19:18:24 BST 2007

On 11/08/2007 17:49, Jon Bright wrote:
> Hi David,
> David Earl wrote:
>> Here's the typical way to do it.
>>                   |
>>     --------------x------------
>>                   |
>>     --------------x------------
>>                   |
>>                   |
> I considered this, but wasn't sure.  There are two problems that I see. 
>  The first is, as you mention, that the turning restrictions at the 
> junction can't be represented.  

... yet. It will come in one form or another soon, so making really 
complicated junction topology is probably overkill - you'remaking a lot 
of work for yourself I suspect, which will probably need to be converted 
into structure designed for this case in the not too distant future.

 > The second is that, at least for some
> junctions, the resulting distance between the minor road and the 
> opposite carriageway is incorrect.

The carriageways have to be somewhat notional anyway when they are clsoe 
together: put them too close and you can't see it is a dual carriageway.

> In my case, the junction actually ends up more like
>         a
> ----<----------------<---
>           \  /
>            \/
> ---->------+--------->---
>          \ |
>           \|
>            |c

I think that would be fine if there were a physical island shaped like 
that, otherwise it will be misleading on an ordinary map, and a routing 
algorithm probably will have a hard time not saying "proceed for 2m and 
turn left".

We need a technique for modelling junctions beyond their physical 
layout, and I'm sure we'll get it soon.


More information about the talk mailing list