It turns into a snorkeling path. 
5 Likes
Do you mean @julcnx’s repeated references to motorcycles and 4x4s on highway=path? These (and snowmobiles) definitely are considered motor vehicles for the purpose of access restrictions where I’m from, but in many parts of the world, including some places where I map, practicality can be the overriding concern.
Incidentally, whether highway=track fits probably depends on the situation. A great many of the highway=paths that Meta mapped in rural southern Vietnam would be obvious candidates for highway=track. But according to the current documentation, that would be based on function and usage rather than the vehicle form factors that happen to fit. The same tag would definitely be inappropriate for tight urban alleys in the same country.
Most OSM-based routers don’t distinguish between the various kinds of motor vehicles. This is partly because, in North America and Europe, a motorcyclist would use the same roads as a full-size car. The only difference is that motorcycles might be allowed to use a freeway’s HOV lane as an exception (and might lane-split through stopped traffic in some places like California). This lane-level detail is irrelevant to the current generation of routers.
By contrast, in Southeast Asia, cars face many more legal and practical restrictions than motorized two-wheelers. For this reason, of the OSM-based Vietnamese mapping startups I’ve noticed over the years, every single one defaults their router to a motorcycle profile that doesn’t imply suitability for cars – if they even offer car routing at all. Unfortunately, I don’t have a current example, because my previous go-to example, Goong, seems to have switched back to government GIS data.[1] However, I recall that these motorcycle profiles always avoided bare highway=paths, as Kurviger does.
In other words, a tag for “motorcycleways” wouldn’t have an advantage over existing road tags in highway=* and motorroad=*. Unsuitability for cars is essentially a region-wide caveat, sort of like american_monster_truck=no in most of Europe. Meanwhile, either highway=path or a new highway=motorcycleway would prevent existing motorcycle routers from routing on ways meant for them.
Don’t shed a tear for them. They never attributed OSM anyways. ↩︎
2 Likes
The current wiki documentation states:
highway=trackare “mostly used for agriculture, forestry, outdoor recreation, and similar activities on open land”highway=pathis “a trail or path that isn’t wide enough for a typical four-wheeled motor vehicle.”- `highway=path is not meant for use by four-wheeled (two-track) vehicles.
If you’re suggesting that any narrow trails used for ‘agricultural’, ‘forestry’, ‘outdoor recreation’ purposes should be tagged as highway=track, that contradicts both the established usage and community consensus I’ve gathered before.
As you mentioned, motorcycle routing profiles are rare, but with clearer guidelines, the existing and future ones can be enhanced. Kurviger, for example, is primarily used for on-road/touring motorcycling, which requires excellent or good paved roads. This is likely why highway=path is excluded, as many of these paths might be unpaved or lack surface tags.
But why? A new tag would imply motor_vehicle=no + motorcycle=yes, which would automatically exclude it from car profiles while ensuring it’s included in any future motorcycle profiles.
1 Like
I think it’s very good to have this discussion here, away from current hot battles, but on the basis of specific problems or problem zones documented with photos. Would it help to name the goals of entering or not entering a highway and attributing a highway this way or another way? Maybe we could agree on a set of goals and how to balance goals in case of goal conflicts.
For instance, “I” referring to the grade C scramble up Nursery Buttres of Table Mountain might state, the goal is not have any dangerous paths in OSM which irresponsible apps, regardless of the attribution of the highway, might present to users as walking paths.
Or with reference to rock climbing gardens, especially with near vertical walls, one might say, the goal is not to create a dense fabric of lines on OSM which nobody would be able to read. And someone else may argue, the goal for rock climbing gardens is to help climbers especially with high-resolution GPS function on their mobile phone to locate the start of the climbing route (which indeed is often a challenge even if there are so-called topos around for the climbing garden.
If we could generally agree on goals and how to solve goal conflicts, mappers could still judge differently in concrete challenging situations. But they could refer to agreed goals and conflict solving principles rather than having a fundamental discussion of the basics, emotionally charged.
3 Likes
Yes. I think we should have a few more related threads, resulting from the current discussions. This one is a reference for issues, since it seems that some are not aware of them.
One thread would be for solution proposals. These are, however, currently discussed in various other threads, mainly
and
This would happen only in the climbing gardens. But that’s not such a bad thing, after all. The routes are, then, densely spaced and it would be even more useful to have them on a map, in order to identify the ones with appropriate difficulty grades, then when they are sparsely distributed.
In cases when they are too sparse, like some nice multi-pitch routes, it becomes really useful to have them on the map. The beginning is usually critical but also traverses along the route are useful to have, especially in the areas that are not well-marked. The preparation, including the topo study, is important but it is still useful to have some assistance for main and emergency exits, descent routes, etc. Many descent routes are also not mapped because they are not used as frequently or at the same spot to create a visible path.
Cross country skiing use the piste:type=nordic for ways. If they’re also accessible during summer they get an additional highway=path tag (or other value)
I would think snowmobile trails also should have a different tag than highway, that optionally can be tagged with highway if used during summer?
3 Likes
To be clear, I don’t intend to make a blanket statement; I’m only characterizing what I’ve encountered in my mapping so far. Contrary to some of your posts about Southeast Asia generally, I haven’t encountered many clearly justified examples that are currently tagged as highway=path in Vietnam, but I acknowledge that different situations may call for different solutions. As this topic is all about documenting problems, it would be helpful if you could also give some examples of what you consider to be problematic.
It’s true that the wiki’s labyrinthine definition for highway=track excludes paths for two-wheelers and focuses on access to the land. The examples I gave seem representative of the majority of highway=path in Vietnam, were all clearly intended for accessing farmland. Whether they are wide enough for two or four wheels is ambiguous and sort of a moot point. It’s not as though you can simply count the number of ruts and declare a vehicle type. What about two-wheeled carts pulled by four-legged beasts of burden?
I suspect it isn’t possible to make generalizations about Southeast Asia because of how differently each country has been mapped. More than anything, what I encounter is blatant mistagging. In fact, the classification of a rural way in Vietnam seems to reflect which corporate mapping team added it more than any real-world criteria. Unfortunately, not all of these teams documented their rubrics.
This track of ambiguous width, clearly intended for farm access but probably too narrow for an oxcart, let alone a pickup truck, was mapped as highway=service by Grab:
This path for pedestrians and small vehicles through a coconut plantation was classified as highway=residential by Meta:
But Meta also mapped the approaches to this ford as highway=track:
There are some bright spots. They also tagged all the narrow, path-like roads lining the canals in this commune as highway=residential, reflecting the fact that they do lead past thick vegetation to residences:
But in an overwhelming number of cases, I see highway=path used for something like this going through agricultural fields, virtually always by Meta. It’s wide enough for a car, and there may be two ruts, but who knows if they were hewn by four-wheeled motor vehicles? Regardless, a motorcycle could certainly go here.
Again, maybe this is very different than what you’re focused on. Please show your hand. 
In theory, but I’m failing to find a clear case on the ground for such a tag, at least in Vietnam. In urban areas, suitability for motorcycles is determined by regulation, and suitability for cars is determined by regulation and width. But we have discrete tags for that already, and regardless, a street is a street and an alley is an alley. Meanwhile, in rural areas off the main road network, there are no ways purpose-made for motorcycles specifically; motorcycles would use the same infrastructure as bicycles, pedestrians, and carts, as seen above. So the half-dozen Vietnam-specific motorcycle routers I’ve seen were justified in routing over the same motorway-to-track scale as global car routers.
1 Like
I didn’t see this on the wiki, but if it does say this somewhere that should be changed. The highway=path page says:
A highway=path is not for use by four-wheeled (two-track) vehicles. Often the path is simply too narrow. Other times it is wide enough but such vehicles are prohibited (except for official maintenance or emergency use)
Note that “is not for use by…” and “isn’t wide enough for…” do not mean exactly the same thing though they may effectively overlap somewhat.
2 Likes
Just to clarify, that line is from the highway=track wiki under the ‘How to decide’ section:
I just wanted to point out that width is one of the factors that help distinguish a path from a track, and that’s not going to change.
Note that the quoted line is not a definition of highway=path, but an example of something that should generally not be tagged as highway=track. highway=path is mentioned as an alternative to highway=track in this case, but that does not mean that highway=path isn’t also used on wider ways. It very much is. It’s not like everything above a certain width must be tagged highway=track and everything below that width must be tagged highway=path.
3 Likes
I’m totally with you on this—it’s not a clear-cut definition. highway=footway/path/cycleway can definitely be wide when general motorcar access is restricted.
On the flip side, the wiki notes that very narrow paths shouldn’t be tagged as highway=track. That aligns with the global consensus I’ve seen before and the widespread usage I’m familiar with in Southeast Asia.
1 Like
These companies have a history of ignoring both global and OSM guidelines, so I wouldn’t consider their past contributions as valid usage examples. For instance, during Grab’s controversial mass edits during COVID, remote paid mappers in India were specifically instructed to map anything giving access to visible roofs as highway=residential or highway=service.
In Thailand, the term ‘alley’ doesn’t have a clear definition and has caused conflicts for over a decade. So, as a community, and based on global feedback we received, we decided to use highway=residential and highway=path to make that distinction.
Grab, which the Thailand community has been working with since the controversial remote mass edits, has even participated in one of our vote (service=alley vs highway=residential). I have been also in regular contact with their manager, who told me once that the current Thailand schema for narrow urban roads is what they use across their SE Asia coverage, except in two countries—Vietnam and another one I can’t remember. I will get more info on that.
Also to some extent Microsoft in the UK. We’re still working out the scale of the problem, but hopefully it won’t be as bad as you saw
1 Like
I brought up Meta because their work accounts for the vast majority of all highway=path usage in Vietnam and most of the navigation data in rural Vietnam. Grab refined much of that work, but no one followed Meta’s lead in terms of using highway=path, corporate or volunteer. We would need to consider this existing body of work as part of any sweeping effort to reclassify or redefine.
By contrast, both Kaart and HOT documented their approach on the wiki in collaboration with the local volunteer community.[1] The resulting tagging guidelines don’t preclude the use of highway=path but clearly don’t envision it for urban alleys. In order for the Vietnamese community to go along with the approach in Thailand, I think we would need some specific examples to compare. (Feel free to link to another thread where you’ve already provided those examples.)
Rest in peace, Tony.
↩︎
Vietnam can totally have its own approach—I’m not here to change that . I just want to point out that using highway=path for narrow urban “alleys” already has some established use, and it’s worth documenting it. Just like service=alley, it might not be ideal solution for everyone, but it could still be a helpful option for other communities.
I will when I find time. I’m actually in Hanoi right now, about to ride a motorcycle for 2 weeks, so I’ll also hopefully be able to draw some direct comparisons
1 Like
From what happened today: I got asked by an elderly neighbour, how to reach a place (a hospital) with public transport. I read today in weekly gazette about a new OSM based multi-modal pedestrian/PT (public transport) router.
So I consulted MOTIS: While this was searching, I did consult the quasi administrative router for the same journey.
I was surprised to see, that both did select the same PT route. The overall route of MOTIS though was 47 minutes, while the official PT router came out 1:06 hours.
No wonder, MOTIS had a path on record that the other did not. The path being an abandoned funicular, where it is possible to make a shortcut.
Also today, when sitting in a train that passed the very route both routers would take, I overheard some 60+ Ladies talking: When they were much younger, they did take the MOTIS route, one even remembered, 180 steps are to be there.
Personally, I’d only take the MOTIS route when I am in a hurry, or when I am out for a trail-run. The path is mapped 30% incline, no sac_scale supplied. That is true. It is not really steps, it is not mountain hiking neither. There is nothing apart from incline in the OS-Mappers arsenal to make that path less attractive for routers, and the incline in this very case can even be gathered from SRTM quite accurately.
PS: Not just in order to make @Tolstoi21 happy, nothing would be different in case the path was mapped a footway.
1 Like
My thought was that there are no issues as such without goals that are not achieved or are thwarted. Unfortunately, we don’t make our goals clear in the many heated discussions and perhaps we don’t even clearly realize what our goals are. I think this is why many disputes end without consensus being reached.
Perhaps it would be possible to use the issues we perceive to make us aware of the relevant goals and discuss them using the examples. Perhaps we could agree on a common set of goals.
An example was the grade C scramble up Nursery Buttres of Table Mountain mentioned by I (IdleLayabout), tagged “demanding alpine hiking” and presented by the app Geocaching.com as hiking route. I agree that it is a problem when average hikers go there, get stuck and need to be rescued. But removing the demanding alpine hike from OSM would require a goal setting like this: not to have any dangerous paths in OSM which irresponsible apps, regardless of the attribution of the highway, might present to users as walking path. I suspect that such a goal would not be capable of gaining a majority. Maybe also IdleLayabout would not stand behind such a goal.
1 Like
I think very few people have proposed such a goal. The implied goal behind many proposals is more like “avoid tagging dangerous paths in a way that data consumers can easily confuse with ordinary footpaths”. It’s not about removing these paths, only about ensuring that data consumers make a conscious choice about how they use them.
2 Likes
“dangerous paths”, yes…
There are several ways why paths can be dangerous. If they are paths at all. So there’s a few problems already. The solutions that can be drawn just from those two sentences:
- mark ways that are not clearly visible on the ground as something else (pathless?)
- apply a difficulty rating (used to be the SAC scale, so at least there some international standard) so that paths can be displayed differently on a map and people know what to expect
- use additional tags (visibility, smoothness, incline, surface) to indicate what is it makes the path difficult
As Alan noted, the point is not to remove these from OSM but to have them distinct, just like there is a clear distinction between a 6-lane highway and a dirt track full of stones in the middle of nowhere.
But definitely we need to set a list of goals and get into shaping them. There were a few recent attempts at that, too:
Perhaps we could refine the goals based on the feedback since then.
The issues I see so far is:
- the paths that are not visible
- difficulty scale (0-grade - Adding T0/strolling to sac_scale, SAC scale)
- shared-use paths (pedestrians and different vehicles)
- motorcycleways, snowmobiles, snorkelling
I think this issue needs clarifying what the issue is. So far, I think the issue is that the legal access for different vehicles can be specified quite specifically, but the purpose or practical usage cannot.
1 Like