[RFC] Feature Proposal - One-way for pedestrians

[osmuser63783]

Thanks for weighing in - I always enjoy reading your insightful comments. It’s a shame we’re in the voting period now, otherwise it could have been possible to take them into account in the proposal.

@Minh_Nguyen

At least in the routing engines I’m familiar with, this is actually rather natural. For example, OSRM has separate profiles for walking, cycling, and driving. A separate routing graph is computed for each, so the mere fact that a road is represented by an edge in the routing graph means that it’s relevant to pedestrians. At this point, the router has no context about other modes of transportation, but if the edge has a one-way flag based on oneway=yes, then the router could feasibly assume it applies to pedestrians.

Sorry, I don’t understand this point. My point was that with the current tagging mess, the best chance that a pedestrian router has to correctly interpret the data is to use an algorithm like:

If the footway has a oneway tag then assume it applies to pedestrians (= don’t route pedestrians against the direction of the way), except if it has a tag like bicycle=designated, then assume the oneway doesn’t apply to pedestrians (= route pedestrians both ways)

This will result in the correct decision most of the time, but people were arguing that it’s counterintuitive for pedestrian routers to have to look at vehicle access tags to work out who is meant by the oneway tag, and @easbar (GraphHopper) was one of them if I remember correctly. Do you disagree with this?

@Minh_Nguyen

To me, the concept is simple. oneway=* on a roadway applies to everyone in the street, but the effect on the shoulder or sidewalk is undefined. On the other hand, oneway=* on a footpath applies to everyone on the path, because footpaths are for feet. If an occasional maintenance vehicle is also allowed on the path and can go in either direction, there’s a oneway:motor_vehicle=* key for that. If bicycles are allowed on the path and can go in either direction, there’s a oneway:bicycle=* key for that.

Personally that makes sense to me: I think on a shared-use path, oneway:bicycle is much clearer than oneway if what you’re trying to say is that it’s one-way for bicycles. But I suspect you may be underestimating how well established the consensus is in the German community (and maybe also in some other European countries?) that the oneway tag never applies to pedestrians.

For an example, see this discussion on the StreetComplete issue tracker on the question “which tag should SC set if the user determines that a cycleway or a path a footway with bicycle=designated is oneway [for bicycles but not for pedestrians]”. The consensus in that discussion seems to be that SC should set the oneway= tag.