Looking at some cities there is a considerable amount of highway=footway+oneway=yes without a bicycle tag that are very likely not meant to be pedestrian oneways, yes. In some cases these are probably mapping errors (missing the bicycle tag) for example, but just looking at their frequency and how often there is an actual pedestrian oneway in real life (in cities, not bike or amusement parks, not mountains) it seems unlikely that restricting the direction for pedestrians was intended. More importantly, I do not think that the meaning of oneway should depend on other tags like bicycle in the first place (c.f. this post).
Apologies if this is a radical suggestion, but perhaps Go Out and Have A Look?
There still will be issues where whatâs on the ground is inconsistent or just silly. For example, here there are segregated / non-segregated cycleway signs pointing in both directions, even though youâd be mad to cycle it south to north. Also, here the signage implies that cyclists have to join a lane on the road and that the sidewalk to the south beyond is a non-segregated cycleway.
I explained to you before, that this is not about going out and having a look. It is about the OSM database being machine-readable or not. If I just went out to have a look I would not need a database at all 
1 Like
at some point you would need a lawyer, but these are at very least enforced (if you run zoo and put âone way onlyâ sign and security enforces it - is it legal restriction or not?)
in such situations like zoos, these are typically just recommendations, and actually enforced only at points, e.g. if you are queuing for an entrance or exit only âonewayâ gate, it would usually be possible to just turn round, leave the queue and walk back before you reach the gate. Many such oneway segments are actually punctual restrictions that were extended by the mappers to where it seemed to make sense, because points donât have a direction
Even if, it sounds like youâre trying to debate whether actual oneway-ways for pedestrians exist at all. I think itâs pretty clear that they do. There are really only 2 questions:
- Should
oneway=yesapply to pedestrians as well in specific circumstances? - How to tag a way to be usable in one direction only by pedestrians, riders, skiers, given that some/many people donât like the recommended tagging from the wiki?
2 Likes
Iâm suggesting that if you do go out and have a look you will be able to update the OSM tagging to make it both match reality and make it impossible to be interpreted incorrectly. Its generally not possible to fix inconsistent keys without survey because you donât know which one is wrong.
The wiki section https://wiki.openstreetmap.org/wiki/Key:oneway#Pedestrians is still problematic (it contains untruths such as âas oneway in OSM only applies to vehiclesâ**), but the bigger problem is that what is in OSM just does not properly match reality, often for several reasons. With that changeset of mine yesterday, the oneway issue was the least of the problems ***.
** the wiki page editor should perhaps have written something like âI believe that the oneway tag in OSM should only apply to vehiclesâ.
*** What has actually happened in this case is that (i real life) the provision for cyclists and pedestrians there has evolved over time; a previous âshared sidewalkâ has become more formal and parts are now segregated, and the entrance to the cycleway for cyclists from the road has been moved. OSM tagging has evolved alongside this, and certain elements of the old infrastructure were still around in OSM.
So you agree that we have oneway for pedestrians in such cases?
And I linked exactly such cases. Suggested route of tour across zoo is different thing, not mappable as oneway and is not included in my examples.
And for hiking trails it applies across entire way and seems enforced primarily by higher risk of dying (Orla PerÄ averages more than one death per year since it was opened, IIRC). See say
from https://www.youtube.com/watch?v=chn6c41_vls I have not checked how much you can be fined.
Yes, and I think a lot of that is from ambiguity in what the tag itself means, rather than the oneway tag. For mappers who think a footway is for walking on foot, there is no more ambiguity about footway than about steps. Those mappers may well be unaware that some countries have large scale use of highway=footway on ways that not only allow cyclists but even have formal signage directed at them.
But at some point we have to draw a line and decide how a previously unclear, poorly documented and haphazardly applied situation should be resolved. We cannot just say âsome mappers think that oneway never refers to pedestrians, others disagree when the highway is pedestrian-only, screw your routing softwareâ.
Once upon a time there was Wild West where everyone with a gun could do as they wanted. Despite the anarchy, that was not that bad, since migrants still flocked to the Wild West, attracted by its vast open space, liberty and opportunity. However, step by step, Law and Order has been introduced in the Wild West, property protected, and gunmen sent to history. I obviously apply an imperfect analogy to OSM situation, but the bottom line is: we cannot be Wild West forever.
The 20th birthday is a good occasion to bring a binding decision how the (rare) one-way pedestrian ways should be tagged, have a decisive vote if necessary, and then document the convention in Wiki, and finally implement it in iD and JOSM as needed. (For example, iD could warn you that âOneway=yes has no effect on pedestrian-only highwaysâ, or have the âOne wayâ checkbox do some magic on those, orâŚ)
6 Likes
Not sure is it the best metaphor and has own not really needed baggageâŚ
which tagging scheme you propose for that? Do you think that any discussed here or used already fulfils that requirement? Because it seems to me that lack of consensus how to do this is the main problem here.
One of them. It doesnât matter which. However, the edit I referred to above didnât need one tagging scheme or another to remove ambiguity; instead to match what was on the ground I removed the oneway tag.
Out of curiosity, are you the same easbar who is a contributor to GraphHopper? Are you trying to figure out how to tell, as a data consumer, if a way is one-way for pedestrians or not, because someone complained that your foot profile ignores one-way footways?
As a data consumer, given the current tagging mess, I think that to guess the mapperâs most likely intentions when tagging oneway on a footway, you would need to look at other tags. If some form of vehicle access has been explicitly tagged, e.g. bicycle=designated / motor_vehicle=private / ⌠then it is likely that the mapper meant the way to be oneway for vehicles only. If not, they probably meant pedestrians. If the mapper used tags like oneway:foot, foot:backward or oneway:bicycle, then you can be sure.
Yes, I hope as mappers we can agree on a better approach going forward, but for now, this approach should give you a fairly accurate interpretation of most waysâŚ
2 Likes
So you agree that we have oneway for pedestrians in such cases?
actually the situations are various and divers, I do not question that real oneway situations exist (e.g. in the mountains in exceptional conditions), but I have also seen queue lines mapped like this, where I would say you could legally at any time leave the queue and walk in the other direction and away, itâs not a legal restriction in this case, just the result of entrance and exit being separate (but the actual restriction is at the point, not on the way leading to it).
1 Like
For the wiki, could the following work as a neutral summary?
Under âhow to mapâ (new section)
There is some ambiguity related to the
onewaykey and pedestrian access (and routing). There is a clear consensus in the OSM community that when the tagonewayis used on a street, e.g.highway=unclassified,highway=residentialorhighway=pedestrian, it only applies to vehicles. However, the tagoneway=yeshas also been used in some situations where the mapper who put the tag clearly meant pedestrians, such ashighway=stepsand=via_ferratawithoneway=yes. This makes the use of the tag onhighway=footwayand=pathsomewhat awkward and the data hard to interpret (for more information, see below under âData consumersâ). There is ongoing discussion in the community about the best way forward. Tags such asoneway:foot=yes,foot:backward=no, andoneway:bicycle=yesdo not suffer from the same ambiguity.
Under âdata consumersâ (to replace existing section âPedestriansâ):
On a street, such as
highway=unclassified,highway=residentialorhighway=pedestrian,onewaycan safely be interpreted as applying only to vehicles. (It should not affect pedestrian routing.)On
highway=stepsand=via_ferrata, the tagonewaycan safely be interpreted as applying to pedestrians.On
highway=footwayandhighway=path, theonewaytag can be harder to interpret: did the mapper mean pedestrians or only vehicles (e.g. bicycles)? Answering this question is often possible for a human by looking at the context (e.g. on a shared sidewalk where bicycles are allowed, the mapper probably meant foronewayto apply only to bicycles; on a narrow path in the mountains, they probably meant hikers). However, for a data consumer (e.g. a routing algorithm) it is hard to take such context into account.The more precise tags e.g.
oneway:foot,foot:backwardandoneway:bicycleare unambiguous.
The current section âPedestriansâ contains a lot of detail, some of it is wrong, some is contradictory, and some is just too detailed for the overview page and could go on the oneway:foot/foot:backward pages.
7 Likes
Yes, Iâm also contributing to GraphHopper. Generally, Iâm not trying to fix things because someone is complaining, but rather because I want to make OSM routing (with or without GraphHopper) as useful as possible. Therefore Iâm not only thinking about changes in routing software, but also trying to comment on the usability of OSM data for routing purposes.
In this case the issue you mentioned made me aware of the oneway tagâs ambiguity for pedestrians, indeed. Of course it is possible to make a guess about its meaning by considering other tags. But as pointed out several times here already this is not a very reliable way to interpret the tag, so I donât think it is a âsolutionâ to the overall problem.
5 Likes
Iâm happy, someone from team routing gives some insights about the usability of the data, and the way it is repreented in the database. Thank you for this! Most of the time, we we tag things, hoping the routing engines will magically understand what we mean, without knowing whether the data is usable for the purpose at all.
Do yiou see an problem withs any of the 2 currently proposed solutions (foot:backward=no / oneway:foot=yes), and what exactly are they, and would they only apply to GraphHopper?
5 Likes
I didnât mean to sound rude, or like it was a bad thing that you respond to user feedback! I think itâs great to get the perspective of a data consumer here. I just meant that itâs helpful to know where you are coming from.
2 Likes
Sorry for the delay.
No, both these tags unambiguously state that a way can only be travelled by pedestrians in one direction. What is problematic on the other hand are ways missing these tags, because something like highway=footway&oneway=yes could or could not mean the mapper intended to say it is a oneway for pedestrians, a routing engine like GraphHopper can only guess in this case (meaning it will often take the wrong guess).
No worries, I did not take it like this.
1 Like
@osmuser63783 Thanks a lot for your changes in the Wiki. I think the issue is explained more clearly now. Just a few more suggestions:
The introduction begins with:
The oneway tag is used to indicate the access restriction on highways and other linear features for vehicles as appropriate. This means that this tag should be used when this way can only be used in one direction by vehicles.
So it does say oneway applies only to vehicles, but unless one is aware of the issues we discussed here this might be still easy to miss. Maybe there could be a short reference to the section (Oneway and pedestrians) already here?
The later sections state:
Tags such as
oneway:foot=yes,foot:backward=noandoneway:bicycle=yesdo not suffer from the same ambiguity.
The more precise tags e.g.
oneway:foot=*,foot:backward=*andoneway:bicycle=*are unambiguous.
It might be obvious to some, but it could help for others to state that these tags are not only unambiguous, but mappers should be encouraged to use them, because otherwise data consumers (like routing engines) wonât be able to interpret the data reliably.
something like
highway=footway&oneway=yescould or could not mean the mapper intended to say it is a oneway for pedestrians, a routing engine like GraphHopper can only guess in this case (meaning it will often take the wrong guess
yes, this tagging should be discouraged and people invited to survey and fix it with less ambiguous tags
2 Likes