[Tagging] traffic_signals:direction=* vs. direction=*
baloo at ursamundi.org
Sun Mar 26 07:48:26 UTC 2017
On Thu, Mar 23, 2017 at 7:44 AM, Martin Koppenhoefer <dieterdreist at gmail.com
> 2017-03-23 10:30 GMT+01:00 Jean-Marc Liotier <jm at liotier.org>:
>> As the "complex intersections" section of the highway=traffic_signals
>> page describes a gradation of model complexity
>> maybe coexistence is also possible between simple tagging of a
>> highway=stop+direction=forward node on the way to an intersection and a
>> complex relation-based model for more complex cases: having a simple
>> model that covers most cases has such significant benefit that I would
>> say it more than offsets the cost of having more than one way to do it.
> If we choose to trade off stability / disambiguity / reliability for a
> little less complexity (no relation), the better solution is positioning
> the node for the sign not on the highway but in the driving direction close
> to it to the side (could be sold as "actual sign position" save some
> exceptions). This way you have the direction implicitly stored and don't
> depend on other ways, but there is the risk of someone dragging the node
> with the sign or the highway so far away that another (unrelated) highway
> comes closer.
This assumption would generally mean that on tight-radius corners common in
rural American towns, the implied face of the stop would be towards the
major road on the downstream side of the intersection, when in reality it's
facing the upstream side of the minor intersection, thanks to the stop sign
being closer to the side of the major road than the minor road.
> Both variants require postprocessing anyway (either you have to look for a
> closeby highway or for the direction of the way(s) your node is part of).
> Yet another option would be to use tiny way stubs for the signs, these
> would have a direction (and could even imply a default sign direction if
> defined like that).
Not to sound like a broken record, but what exactly would be so complicated
about adapting the existing and adopted enforcement style relations to stop
and give way devices? This completely removes ambiguity and guesswork and
can already be done using existing tools. Plus, like turn relations,
further developments on the editor end would make this even more trivial.
-------------- next part --------------
An HTML attachment was scrubbed...
More information about the Tagging