[OSM-talk] Turn restrictions

Renaud Martinet karouf at gmail.com
Thu Aug 16 13:03:09 BST 2007

On 16 Aug 2007, at 11:58, Steve Coast wrote:

> On 16 Aug 2007, at 10:29, David Earl wrote:
>> On 15/08/2007 22:48, Steve Coast wrote:
>>> So I'm on a train otherwise I'd link to Andys wiki page he pointed
>>> to  today, but I want to try out a turn restrictions scheme,
>>> Given a road direction, and roads 'Baker Street' and 'Marylebone
>>> Road' you tag the way 'Baker Street' with
>>> no_turn_left:'Marylebone Road'
>>> If you can't turn right then it's 'no_turn_right'. If you can't
>>> turn  at all it's 'no_turn'. And thats it. There are three major
>>> problems  with this
>>> First, as far as I know JOSM doesn't let you use keys more than
>>> once  but the API does. So you can't have more than one
>>> no_turn_left. So  fix JOSM.
>>> Second, what if it's no turn for car but ok for bikes, or
>>> whatever.  Well, then shouldn't we look at triple tags?
>>> car:no_turn_left:Marylebone Road ? That is, keyvals now have
>>> third  string along with key and val? We'd add a third keyval
>>> property, call  it what you will, and all the current keyvals will
>>> have that third  proprty set to 'general'. So general:name:Foo
>>> Street and so on. Then,  this third column can be specialised
>>> along the lines of Map Features  as the keys and vals are.
>>> Third, what if someone misspells 'Marylebone Road' or the road
>>> isn't  joined, or it doesn't exist or something. Well, people will
>>> enter bad  data ON PURPOSE and our map data will never be perfect
>>> ANYWAY. And,  it's a wiki so you can fix it. And, it's incredibly
>>> simple compared  with any competing scheme I've seen. Also, JOSM
>>> (or potlatch) can  prompt you and say 'ah-ha Mr. Mapper, that
>>> looks wrong!' And you can  read it, but you can't read
>>> bigscheme:way:324543525234:no_turn:34523452.
>> Frederick's prototype for relationships addresses this properly, by
>> grouping the Ways affected by reference rather than name.
>> See http://openstreetmap.gryph.de/entities/
> Yes but my way is about a billion times simpler, human readable and
> works today. :-)
> have fun,
> SteveC | steve at asklater.com | http://www.asklater.com/steve/

Entities are not that complex for anybody who uses JOSM on a regular  
basis. It keeps the data consistent, it is human readable and it  
works since Frederick made an implementation of it. It provides a  
solution to a lot of things we want and will want to do.
For once that somebody is going through the trouble of proposing  
something and actually implementing it, I think we should give it  
some credit. We'll need to move toward this kind of relationships and  
objects one day or another anyway...


More information about the talk mailing list