Some worthwhile comments were posted. From what I have learned, I propose do add an ambox like message to consumers of openstreetmap data to the documentation:
Consumers beware: Highway=path is a very broad tag that can represent anything from a several meters wide paved construction – over what the heading picture shows – up to the idea of where some unspecific terrain can be passed through, without any markings on the ground.
General purpose consumers might want to check if they are interested in pathways with additional tags such as sac_scale!=hiking, mtb:scale!=0 or trail_visibility!=excellent. This list not exhaustive, just a first stab.
Consuming OSM ways with just a bare naked highway=path tag might require deep research into what is the regionally common meaning of such.
Please share ideas, suggest how to better word, or explain, why general purpose consumers shall use, what above excludes.
It may be generally a good idea to add a paragraph “Advice for data users” to certain wiki pages for keys and tags that need it. highway=path is a prime candidate.
I think it is better just to put the actual use in clear, unambiguous text, including the world wide access defaults. Then proceed with some often encountered subtypes, suggesting which extra tags will help tell them apart. If possible, using best practices.
Taginfo numbers may help.
Data users, mappers and tool builders can then determine their handling accordingly. A pointer here and there might help. An extra warning box, I think not.
I think if we are going to have a warning especially for highway=path, we would need to be clear about what makes this tag different from all the other highway tags that don’t have this warning.
The warning here emphasises physical characteristics. But highway=trunk, for example, can be a multi-lane asphalt divided highway or a gravel track with just enough space for two vehicles to pass (maybe there are even narrower ones somewhere). Even looking at a single road, the physical characteristics of the Carretera Austral in Chile are very different between the outskirts of Puerto Montt and more remote southern regions. There is no way a router could sensibly estimate travel speed for that road without looking at secondary tags.
Instinctively I agree that highway=path is “worse” than this highway=trunk example, but I find it hard to pin down exactly why.
I’ve been silently following this thread (& the Take 1, too), as well as other posts covering the path-controversy. I find these discussions very interesting and enlightening, even though (or precisely because) there are no clear cut answers available to the general problem.
As I said, I of course, also have no solution to the path-problem, but I think that @alan_gr 's reaction above is on the nose, and an excellent point.
For example, a highway=cyclewaycan also, absolutely, be anything between an unpaved narrow pathway to a multi-lane, strictly segregated, silk-smoothly paved, bicycle-privileged and completely unambiguously signposted road. That’s why we need extra tags like width, surface, smoothness, segregated, designated etc to all the roads in all the non-trivial cases. I think this is fundamentally no different for paths, though they may need more of these specific tags, since there are less obviously inferable properties on them.
I think anyone who considers one picture of a single way or road to be “the one true meaning” of anything represented by OSM is clearly mistaken or delusional, or both.
I suspect that you’re trying to make a joke, but would suggest staying away from that in wiki pages. Instead say something like “highway=path has been used by OSM mappers to describe all sorts of features, from this extreme to this”.
I think if we are going to have a warning especially for highway=path, we would need to be clear about what makes this tag different from all the other highway tags that don’t have this warning.
I agree, while your trunk example doesn’t work in the German or Italian context, there is a similar huge range of different roads tagged as unclassified or residential road. Or tertiary. What actually to expect (in terms of surface, width, smoothness, etc.) depends a lot on the context (rural mountain area or city for example)
Yes all the road & pathhighway tags can have a wide range of physical characteristics. However, most are specific enough to provide at least some useful meaning. If a way is tagged only with highway=trunk, a data consumer should be able to assume:
This is a road intended for and used by the common motor vehicles in this region
It is more important to the transportation network than most of the other roads nearby
If a way is tagged only with highway=cycleway, a data consumer should be able to assume:
This is a path intended for, and used by bicycles
Even though both of these situations are still rather ambiguous, they at least provide some specific information that gives the data consumer a general idea of what the thing is.
Lets look at a way tagged only with highway=path for comparison. There are a number things a data consumer can assume but they are all broad and unspecific:
This might be a path or it might be a pathless route with little to no visible evidence of existence on the ground
It is used by one or more of the following
Pedestrians
Bicycles
Horses (or other similar domestic animals)
Motorcycles
Small off-road vehicles (dirt bikes, ATVs)
Golf carts
The terrain might be too rough for any wheeled vehicles or horses
The terrain might be too rough for a human walking, requiring scrambling instead
These assumptions paint an extremely vague picture and cover a much bigger range of things. The only recommendation I could make to data consumers is to treat plain highway=path as a rough, faint path that is hard to follow and with suitability for anything other than pedestrians unknown. Go ahead and interpret it more specifically when combined with specifying tags, but without those you really can’t assume much. Best to assume the lowest quality infrastructure that the tag might represent (in this case none).
The difference with the climbing route path extreme, is that there is nothing on the ground. It is simply a set of segments whose coordinates are given by an organisation.
That doesn’t reflect the actual use, I think. Most paths are clear, and much easier to follow, and accessible for (suitable, but not exotic) bicycles (and probably horses). I consider mapping a scramble as highway=path with no other tags a mapper error.
I can live with a scramble mapped as highway=path + path=scramble (even if the scramble is 20 m wide and you have to find your microway between the rocks), or a key descriptive physical tag, or an appropriate sac scale.
highway=trunk is in general passable. And while it sometimes may not be in some jungle/tundra, it is in area where such trouble can be expected.
While highway=path trap may happen nearly anywhere, including city centers.
I would not combine effort with documenting actual use with yet another attempt of trying to succeed with tricky and complicated proposal that was attempted numerous times and has not succeeded and it is enormously complex and confused project. (including some people believing that it already succeeded)
A data consumer, in order to consume hghway=path without any tags, has to assume some kind of access. I think access for pedestrians, bicycles and horses to ways mapped as highway=path, is world wide most common, and the database reflects that. And =no for all other vehicles. Do you disagree?
Feel free to ping me in a new thread about this. But in short, I disagree.
I would not combine also discussion effort about documenting actual use with discussing yet another attempt of trying to succeed with tricky and complicated proposal that was attempted numerous times and has not succeeded and it is enormously complex and confused project. (including some people believing that it already succeeded).
Would you say, path without other tags, as it is used today, is only for pedestrians? Meaning that all other uses have to be tagged explicitly? That is a valid preference, of course, but I’m trying to assess if current usage supports this.
You will have this problem more or less for each highway type. Even with the warning, data consumer still need to make assumptions and will base them on something in average. The only solution in my point of view is having more data describing the highway.
