I still think that for the overwhelming majority of STOP and YIELD (give_way) signs, the tagging scheme of a node either on the intersection or near the intersection on the approaching way makes sense.
And in fact, I think that a fairly simple rule of interpretation actually answers those who complain that the approach of tagging a node makes no graph-theoretic sense. I contend that it does. A vehicle finding its way in our navigation model traces a path from edge (a segment of a way between two nodes) to vertex (a node) to edge to vertex ... from its origin to its destination. It is true that nodes have no direction, but edges do. I'd say that the simplest interpretation that could possibly work for 'stop' or 'give_way' is: (1) If there is no direction, the regulation applies to any vehicle approaching the node from any incident way. (2) If there is a direction (forward or backward), the regulation applies to a vehicle approaching the node in the corresponding direction on the way. For the simplest case of an ordinary traffic light or a four-way STOP, a simple tag on the intersection suffices. For the case of a STOP or YIELD (give_way) sign on a side street, place a node near the intersection, tag it with the sign, and mark it with the appropriate direction. For the case of a STOP sign or traffic signal guarding a narrow bridge or single-lane way, the node can still be on the way and have the appropriate direction tag. Obviously, if we are using the node for other purposes, we may need to disambiguate with stop:direction=* or give_way:direction=*, just as we do with other ambiguously-named tags. I'm having a hard time picturing any case where this couldn't work. It doesn't involve measuring distances to intersections, trying to divine from a sign placement beside a way what it means for traffic on the way, or any other weird preprocessing. I could explain it to a novice mapper in two minutes. For the occasionally truly weird case (e.g., traffic moving ahead or turning left must stop; traffic turning right gives way to left turns from the oncoming way, or something like that) we still have the proposed relation schemata available, but I really don't get why we feel we have to impose those on the bog-standard two- or four-way STOP. On Mon, Mar 27, 2017 at 8:30 AM, Jean-Marc Liotier <j...@liotier.org> wrote: > On Mon, 27 Mar 2017 14:08:55 +0200 > Topographe Fou <letopographe...@gmail.com> wrote: > > > > When you say direction=forward I assume it is forward/backward and > > not other direction values. > > Yes. > > _______________________________________________ > Tagging mailing list > Tagging@openstreetmap.org > https://lists.openstreetmap.org/listinfo/tagging >
_______________________________________________ Tagging mailing list Tagging@openstreetmap.org https://lists.openstreetmap.org/listinfo/tagging
