[OSM-talk] The segments vs ways vs superways question again...

David Earl david at frankieandshadow.com
Wed Jan 3 09:57:52 GMT 2007


...
> > after this transition was made in the db, an automated client 
> could maybe scan
> > the data for ways, and alter them to be a group of paths if 
> they contained a
> > branch, or just a path if they did not contain branches. I 
> would also suggest
> > that tags were again then put directly on the paths, rather 
> than on the new
> > groups, unless necessary.
> 
> No, tags must be put on groups. All roads should be a group, some with
> one path some with more. All road names must be at the group level - it
> would be a complete mess to have some roads consisting of only one path
> having a road name at the path level, but probably most roads that have
> a bit more complexity having road names at a different level. That would
> just be the worst of all possible worlds.

This seems to defeat one of the advantages of the node-path-group structure, namely that *parts* of the group can have different properties even though they are all part of the same entity. Without this, we still have to end up splitting naturally grouped paths in order to represent a property which changes along its length (or branch).

While the name of the street would normally be on the group, I'd have thought, it may be that parts are residential and parts unclassified; or that one of the paths is a bridge on layer 1, etc:

<group>
  <tag k='name' v='High Street' />
  <path>
    <tag k='highway' v='residential' />   
    <noderef id='...' />
    <noderef id='...' />
    ...
  </path>
  <path>
    <tag k='highway' v='unclassified' />   
    <tag k='bridge' v='yes' />
    <tag k='layer' v='1' />
    <noderef id='...' />
    <noderef id='...' />
  </path>
  <path>
    <tag k='highway' v='unclassified' />   
    <noderef id='...' />
    <noderef id='...' />
    ...
  </path>
</group>

David





More information about the talk mailing list