[OSM-talk] colour fills in osmarender

[Richard Fairhurst]

Andy wrote:
> Probably 50% of the ways I have created in residential Sutton Coldfield have
> been non-linear, in that they contain loops, multiple branches and the like.
> This is because I define my way as one which represents the contiguous
> physical street reference, normally its street name. If the named street
> forks then so do my ways.

Hmm, interesting.

My approach has been is that a way comprises a linear run of segments  
each of which have the same attributes (key/value pairs).

So for Ticknell Piece Road in Charlbury, which has plenty of little  
branches off it (in the way that modern housing estates tend to), each  
branch is an individual way - albeit one also called Ticknell Piece  
Road.

The danger of choosing a single "contiguous physical street reference"  
is, well - which one do you choose? Streets have attributes popping in  
and out at regular intervals: road numbers change, National Cycle  
Network routes arrive and then depart, and so on.

IMHO the above definition of a way avoids this problem; makes lots of  
applications easier including styled map-drawing and route-planning;  
avoids the problem with segments having their own keys/values; and  
(perhaps most importantly) lends itself to a way-centric UI as Tom  
described earlier.

cheers
Richard