[OSM-talk] Turn restrictions ambiguity

[Tobias Knerr]

SteveC schrieb:
> Ok so in that case... why don't we make best practice to split your way
> A in to two directions, rather than hundreds of little ways?

You mean something like that

    ^A1   |A2
    |     |
    |     |
    |     |     B
 ---*-----*------
    |     |
    |     |
    |     v

with both A1 and A2 being oneways? It's possible, but should probably be
done only if the two directions are separated in reality.

Otherwise, this will affect the possibility of turning. It also isn't
great that the user sees two roads where only one exists in reality. You
also have to deal with navigation software announcing two junctions
instead of one, and so on.

If you then consider that applications don't interpret anything except
the no_/only_-prefix and aren't expected to care about the rest of the
value ("left", "right" and "straight on" being nontrivial concepts for
software), you'd have to create two directions for B, too. At that
point, it's probably best to just split A at junctions and be done with it.

Tobias Knerr

PS: I'd really love to see a feature to select ways between *click1* and
*click2* in editors. Would make all that way-splitting much less of a
problem.