[OSM-talk] colour fills in osmarender

Richard Fairhurst richard at systemeD.net
Fri Jun 9 15:06:07 BST 2006


[big chunks of text quoted because Etienne's original went, I think  
erroneously, to me rather than to the list]

Quoting Etienne <80n80n at gmail.com>:

> On 6/9/06, Richard Fairhurst <richard at systemed.net> wrote:
>> Yes, agreed. So roundabouts could be:
>>
>> - defined as individual ways
>> - multiple 'ref' tags, one for each road that meets at the roundabout
>> (e.g. ref=A361,ref=A44,ref=A3400 - this avoids the need to say "turn
>> onto the A361" then "turn back onto the A44", because you simply
>> prefer the ref you're already on)
>> - junction=roundabout
>> - oneway=clockwise|anticlockwise
>
> I think I agree that roundabouts should be a way on its own that is
> distinct from all the ways leading to it.
>
> The oneway attribute only needs to be true/false.  True meaning the
> traffic is in the direction in which the nodes are ordered in the way,
> false meaning the opposite.    I think it would be difficult to deduce
> what clockwise means.  For example:
>
> <way oneway=clockwise>
>  <node id=1 lat=50.926579090483266 lon=-1.7636343126730878 />
>  <node id=2 lat=50.92650013591016 lon=-1.7633346667141547 />
>  <node id=3 lat=50.92660735124737 lon=-1.7633605302878226 />
>  <node id=1 lat=50.926579090483266 lon=-1.7636343126730878 />
> </way>
>
> What is the clockwise order of these nodes?  And what is the algorithm
> to derive this?

Clockwise is "you turn left when coming onto the roundabout",  
anticlockwise is "you turn right". But I guess you're right,  
oneway=true|false would do.

>> (In general, FWIW, my personal preference would be to simplify the
>> data model by eliminating segments... but I'm not really that bothered.)
>
> I haven't heard anyone yet defend the need for segments as they are
> currently defined.  I can't think of a good reason for them at the
> moment - but then maybe someone has more insight than me and they
> exist for a reason...

Anyone?

Hey, let's see if we can reinvent the whole OSM data model while  
Steve's in SF and see if he notices. ;)

cheers
Richard





More information about the talk mailing list