[OSM-talk] Re: One way streets

[Laurence Penney]

The problem here is IMO that the basic model of assigning the onewayness 
to the way is flawed, even allowing for exception definitions. It is 
harder to explain, harder to remember proper methods of working.

In the real world it is parts of roads between junctions that have 
directionality. Topologically a sequence of vertices, each with just an 
incoming and an outgoing edge, is equivalent to a single edge. And an 
edge is directed or not. Therefore we should treat sequences of segments 
between junctions as single entities in a UI. A curved (multi-segment) 
road portion should work the same as a straight (1-segment) road. This 
is all irrespective of any way assignments.

[That said, if - in terms of data compression or usability - it seems 
beneficial to allow entire ways to be defined as oneway, then that's 
fine for a UI.]

It seems odd that node or segment order is important when anything is 
presented visually. Also, being able to set a "reverse" direction (the 
"oneway=-1" that's been mentioned) seems essential to me. Therefore the 
idea of typing "oneway=yes" seems very hacky. Can the UI not show the 
direction visually with a little arrow, and include a 'flip' button? It 
seems no more sensible to offer a text interface to this than to have a 
text interface to point coordinates or colours.

-- Laurence