History of proposals to fix highway=path ambiguity – and a wayforward?

General talk
,
181

I might be repeating myself here, but consensus about tagging changes is not going to happen any time soon. My summary:

  1. Documenting the range of current usage helps mappers and data users.
  2. A sort of crude major subtype system using a subtag such as path=* with archetype pictures would help mappers and data users.
  3. Adding a usability or difficulty scale subtag might help
  4. Lifting one or two very specific subtypes out of the range could have a chance of successs
  5. Physical descriptive subtags are available, but (overly?) demanding for mappers and data users.

All of the above will not add up to 100% clear path mapping; there still will be many bare highway=path objects in the foreseeable future. So data users will still need to apply fallback values and fallback rendering.

I think it is worth it to appoint the subtype with the highest occurrence as the default type, thereby setting recommended fallback values for all relevant subtags. Not as a definition of what a path is, but as an anchor point for mappers and data users.
If a mapper knows how a bare highway=path will be interpreted, (s)he will be encouraged to verify and map the differences. I know I would.

7 Likes
182

For pedestrians, the profile is fine, but what about cyclists, horses, and motorcycles? I think the basic highway=path should be excluded from these profiles.

But what exactly counts as a ‘non-bare’ path? Does adding only a surface tag, sac_scale, or even just foot=yes make it qualify?

183

Yeah, Thank you to @Peter_Elderson for that reminder and post. I wholeheartedly agree with what you said above!

Also,

Yeah, true that and guilty as charged :slight_smile: .

I do, however, wonder how exactly one should go about interpreting the different paths if we assume that a path with no other additional tags are at least ‘passable’. The ones that are really wide and have excellent conditions would have descriptive tags/an additional path= category, and the ones that are narrow and in a horrible state would similarly have their own descriptive tags/another additional path= category? Ways with only a highway=path tag would represent a middle ground. That might work?

I agree with this, but above we had these responses:

So a path would have to represent a middle-ground between the really good ones and the really bad ones, or @dieterdreist and you won’t re-tag them?

184

That is what I observe too. It covers 100% of what is newly added as bare and naked path. Some are the typical trampling traces over grassy areas besides a parking lot. Some look like that:

Steigspuren
Steigspuren1296×1296 416 KB

I cannot even fault the mappers for not sticking more attributes there, it is such a minor path that nobody wants to spend any more than the base tag on it. I rather not add informal, I do not want my name signed on that. I have no idea which sac_scale grade. Surface ground, that’s certain. Obstacle?

I cannot imagine which routing profile actually would want them – trail-running?

3 Likes
185

Hiking, trailrunning, cyclocross, MTB.

186

I was referring to the other major traditional case for highway=path: urban shared use paths that mappers in some regions have refrained from tagging as a highway=*way out of concern that it would bias renderers in favor of one mode of transportation over another, or that it would imply a different traffic sign. To the extent that the sign has legal force, it probably translates to an access or designation tag, so this wouldn’t be a “bare” highway=path anyways. In regions where we don’t have a strong compunction to map the signs as ways, the advice “choose the best highway=*way” seems to be working well.

For more primitive, non-urban trails, I agree that highway=path is usually better-fitting than highway=*way, but I wasn’t claiming that highway=*way is a solution for everything. Like several others here, I hold out hope for something like path=* to move us away from an overreliance on access keys for classifying paths.

Eventually, you’ll post enough examples of these that British English speakers will come around to something like trace or trod as a classification value. If not, I submit highway=foot_precedent for consideration. :wink:

2 Likes
187

The MTB trail 100 m from the other looks quite different:

MTB-Trail
MTB-Trail960×960 258 KB

It is not mapped yet, the MTBers not in need of? Cycling not allowed there, btw. I will not map it neither. There is a heavy duty downhill bike in the picture. Not on the picture jumps, that some created, so not an informal path, as it shows improvements/construction, trail_visibility excellent too :wink:

For cases like those, nobody maps a footway or a cycleway but a path. It is just a pity, that path is so broad and paths with no attributes commonly considered

Today I wrote a changeset comment to please be more specific, at least add surface=dirt;informal=yes. What more would you recommend?

Here, new paths with no attributes almost always are the bad ones, mostly ephemeral stuff, the good ones long since mapped, few new ones appearing. The ones of today’s just cutting switchbacks of a forestry track, all OSM consumers will route pedestrians and cyclists there. 99+% of their users will condemn them for that.

1 Like
188

Around me, I still see bare highway=path quite often for newly opened paths without signage (especially short connecting paths), and for paths on the edges of urban areas that are quite commonly used to get from A to B. I would want any pedestrian router to use these (perhaps not a cycling router but I’m not a cyclist).

I do try to add surface at least in this situation. But does that help much? If there is a choice between reasonable paved and unpaved alternatives, maybe. But often the alternative is a busy road with no sidewalks, where I’d prefer the unpaved path.

189

I think this was THE major force, not explicitly urban, but shared use. The pictures here are not all the originals, some have been changed due to copyright concerns. But they are also about shared use mostly.

Funny note: the bicycle exclusive picture is in Austria, and on that cycleway there I am very much certain that it is not foot=no. The law is unclear and the high court never touched the issue. Nobody ever will get fined for walking there: On the left side of the carriageway and not obstructing cyclists, that is. Just like on any other road.

190

Yes, I absolutely agree with you! The reaction you quoted from me was a question intended for @julcnx and @dieterdreist who replied that the wouldn’t bother adding any additional tags to ‘passable’ paths. My intuition works the other way around (and, I surmise, agrees with you): I’d suppose that a way with only a highway=path tag is at best a dirtway like the MTB trail your picture shows.

In answer to your question, I’d be content with surface=, smoothness= and width= estimates, as well as a informal=yes/no characterization. As I mentioned earlier in this thread, Duck Tagging cycleways[1] is in my opinion a perilous practice. Duck Tagging paths doubly so, for reasons that are all too obvious.

Many people have argued that no one will ever add the descriptive tags to paths (apparently, even piecemeal). I take issue with this argument. The point is that any demarcation on paths will have to be based mostly on their physical properties. Even if we invent new path= categories, those will have to be decided and evaluated based on physical properties. I fail to see how it would be more onerous to add the physically descriptive tags to all or most of the paths, rather that a category tag which itself is based on physical criteria.[2]

[1] That is, tagging non-designated ways that are suitable for bicycles with only a highway=cycleway tag and no other tags. I know that the plural of ‘an anecdote’ is not ‘data’, but in my vicinity, many Duck Tagged cycleways are not the kind of ways I’d be comfortable to ride my roadie on.
[2] I do see a point for a very limited number (two or maybe at most three) of such categories (and in distinction, the three or four descriptive tags I mentioned earlier will be able to express a bewildering array of different possible combinations). For example the =pathless proposal makes sense to me on ‘paths’ that traverse either bare rock surfaces of among boulders, where the route of the ‘path’ is marked with spray-painted rocks or waymarkers. These ‘paths’ wold not e.g. have an intelligible value for width=, and might need special tools to traverse.

191

The suggested tags aim to break down highway=path into more descriptive categories:

  1. highway=pathless: This tag would cover routes that are often indistinct or non-existent, potentially unsafe, and sometimes marked only by minimal indicators like cairns. Such trails typically require additional tags, like sac_scale for difficulty.
  2. highway=path: This would remain for unpaved or rustic trails mainly used by pedestrians and sometimes by mountain bikes or rugged motorcycles. The emphasis here is on trails that are more informal and not heavily constructed.
  3. highway=two_wheels: Meant for pathways accessible to bicycles, motorcycles, or horses, yet not suitable for cars. This would focus on paved or constructed ways usable by various two-wheeled vehicles, possibly replacing or supplementing highway=motorcycleway.

Transitioning to these new tags would require a lengthy adaptation period due to highway=path’s extensive current usage. This period would allow for gradual re-tagging and encourage the mapping community to adopt the changes, with automated re-tagging considered for specific cases. The proposed tags would provide greater clarity to both mappers and consumers of OSM data by establishing a more granular classification system for non-car pathways.

2 Likes
192

Probably you are true, but compare this with road classification which is based on usage and purpose. Not the physical criteria. Shall we have two different classification systems?

1 Like
193

That’s a good point and question. My understanding is that a =path is already a classification based on usage and purpose. This, of course, does depend on local legislation though. A way that is not reasonably/legally a footway or a cycleway but which permits both, is usually a path (and so it is already a functional category, & further distinctions of it would have to be primarily based on physical criteria).

194

I appreciate that you have given a list - I’ve seen a lot of people say “we should promote secondary tags much more” without being specific about which secondary tags. One issue I’ve mentioned before is that it is currently not always obvious how to apply some of these tags to paths.

In particular, the documentation for smoothness would need a lot of work if we are going to ask mappers to systematically add this tag to paths.

The wiki currently refers to “drivable ways” which I would interpret as roads for motor vehicles, although that might be a language issue. The examples also seem to be heavily influenced by driving with a lot of references to paving, asphalt, and concrete. It says that the best unpaved but compacted roads are “intermediate”. Does that mean that no unpaved path, no matter how smooth, could ever be higher than smoothness=intermediate? Also, for non-cyclists it would be helpful to have more hints about how to distinguish suitability for trekking v city v racing bicycles.

For width, I’d like to see a bit more clarity about when to use width versus est_width - after all, at a sufficiently granular level, every measurement is an estimate! In particular, how to tag unpaved paths where the width may vary a lot over short distances. Perhaps some rules-of-thumb would be useful, e.g. “if two people can easily pass each other at any point without stepping off the path, it is probably at least Xm…”

surface I think is generally OK, although perhaps we could be clearer that often tagging as simply “ground” is good enough - StreetComplete in particular seems to de-emphasise this tag, but I find it is often the only practical tag without splitting up a path into pointlessly small segments.

I think my points above hint at one reason. For a lot of paths I have walked, I could fairly confidently say, for example, “this is a trail” based on my memories and photos. That would be valid for a long section of path. But I don’t have the data to map the surface, width, and smoothness of every little segment. And even if I did, I find smoothness in particular difficult to apply for the reasons given above. For those reasons, I’ve tended to be a bit negative about the “just add secondary tags” approach. But I might change my mind if we could come up with a way to apply “good enough” tagging so that I don’t have to check for changes every 20 metres when hiking.

8 Likes
195

That is probably because you don’t frequent such paths. The path width changes with the terrain. It can be 30cm on a stretch of 10m, then it widens to 50cm for 30m, then 2m for 25m, then you have to climb over a rocky area for about 4m, and so on. If you think it is fun to be mapping these on the spot, while holding to a grass/rock/cable/poles, then do that, by all means. To me, it is not.
The same goes for inclination, surface, or pretty-much any other physical characteristic. Doubly so when they are covered with fallen leaves. Eventually, I could pick a number from the SAC scale and say “this was T3 or T4” based on the most difficult part, but that’s it.

For the rest, it seems that a consensus is building around a few distinct types of paths (whether by all physical characteristics packaged together, along with a purpose, visibility, etc.) and I think that’s good.

6 Likes
196

Please have a look at this discussion on the Wiki page for smoothness, re “description of impassable”. In short, the description currently reads

Ways that used to be passable roads and can still be recognised as such in the landscape (useful for hikers), but have deteriorated so much because of fallen trees, being overgrown, severely eroded, etc. that no 4-wheeled vehicle can pass it. Consider to map narrow parts as highway=path and the whole way as disused: or abandoned.

This suggests that there’s a minimum width requirement for something to even be tagged with smoothness=. This doesn’t reflect how smoothness is actually used (impassable in particular is very often used on paths that were never roads).

I’d like to change this to

Ways, most commonly highway=path, that are not passable by any vehicles, not even mountain bikes or specialised off-road vehicles, but are recognisable in the landscape and may still be useful for hikers. Consider also lifecycle prefixes disused: or abandoned: for example if the way used to be a passable road but has deteriorated.

It might be worth its own thread (that description has been in the Wiki since 2008!) but it would be good to hear from people in this thread if you agree with the proposed change.

5 Likes
197

For my part, I consider the objective “descriptive” tags onerous because they aren’t criteria at all, but rather precise observations. width=* requires measuring equipment, skill, or guesswork. Both width=* and surface=* may require splitting the way an inordinate number of times to accommodate variation. On the other hand, if you’re referring to something more subjective, like trail_visibility=*, smoothness=*, or sac_scale=*, I think the problem becomes that any criteria are essentially arbitrary. Moreover, just as the labyrinth of access keys do a poor job of iteratively refining highway=path, it isn’t clear which combinations of smoothness=*, surface=*, and sac_scale=* map one-to-one with everyday vocabulary.

I haven’t experienced the “duck-tagged” cycleways in your vicinity, but the use of highway=*way in the U.S. is actually grounded in intuition and reinforced by cultural norms, not some robotic decision made in a vacuum. Admittedly, these norms may not be universal; we’ve already basically established the Safir–Whorf hypothesis of highway classification. Still, if an American accidentally opens iD to your neck of the woods and starts tagging highway=cycleway based on American norms, rest assured, you would have no qualms about riding your road bike on these ways. The highway=cycleways over here may not meet Dutch standards of performance, but our fixies aren’t abnormally rugged. If anything, the hapless armchair American would apply highway=cycleway too conservatively without your valuable on-the-ground knowledge.

A new path=* classification system would be an attempt to create new norms that are hopefully consistent with existing common sense. In order for this approach to be a success, each value would need to be distinct, easy to learn, and memorable, and ideally easy to translate to plain language. By analogy, most of us intuitively know the difference between a service=driveway and a service=drive-through without needing to consult a checklist or treatise.

2 Likes
198

Even with the explanations, I fail to see how these are counterarguments. Generally speaking, if a path really changes from a 30cm (about a foot) narrow way to a ten-meter (over 30 feet) wide way, I’d say that that’s pertinent information to convey to people that use it. Two further points:

  1. @alan_gr did point out that e.g. the estimate on width doesn’t need to be granular and exactly precise to be useful. I, and no one else, has never required centimeter exactness. Of course, width does take a real number, and they are in fact uncountably infinite (though, in practice depend on the size of the mantissa of the floating point used to store them). Though infinities and the continuum are mathematically very interesting points, I’d say that an estimate within a meter would suffice for most purposes in OSM. Same goes for the other descriptive tags. I do frequent such paths, and at least from where I’m from they don’t change from asphalt to grass to gravel every few centimeters, or even tens of meters (and if they did, I do think it would be important to tag this).

  2. Furthermore, if further classifications on paths would mostly depend on physical characteristics, the ‘granularity’ argument would just be postponed or transposed to the categories. I.e.: if the physical characteristics of the path would change enough that they would make a difference in the roughly-within-one-meter-gradient estimate (& same for surface and smoothness), presumably, one would have to apply a different category for that part of the path as well (for exactly the same reasons). So even if you’d use the categories approach, in your examples of very varying pathways, you’d still have to use different categories for the very different parts of the paths.

Ceterum censeo, I think the points made by @Antonio_Eljesser_Briceno above are really interesting. I could see the three categories (or something like them) as a possible compromise, or at least a good place to discuss next.

199

In addition, I doubt that this covers all ways currently tagged as path. Which means that highway=path will still be used for a broad spectrum of ways, which may or may not in fact belong to one of the new categories.

I also think these categories overlap with existing types: highway=cycleway and highway=footway.

In general, I think significant change of meaning of a massively used existing tag has little chance of success. Lifting a specific and well recognizable case out of the spectrum to get its own highway value IMO does have a chance, especially if it comes with a full and finite retagging plan and tooling (presets, validations, QA-tools).

However, that does not solve the highway=path ambiguity issue.

200

Perhaps, but that’s your opinion, not a documented consensus. Is it reasonable to expect a new mapper have the confidence to make that judgment? And then why do we have both width and est_width, if it’s OK to use a rough estimate for width?

Not from asphalt, no. But I certainly know paths that change from rock to dirt to grass over short distances as they pass from exposed rock to forest to forest clearing. Which is why, as I said, I tend to default to “ground” for a reasonably long section of trail. But again, can we expect a new mapper to judge that is acceptable?

I don’t think so. There are lots of trails I have walked that basically share the characteristics of trails in mountainous parts of this region: surface depends on surrounding terrain and vegetation, width also depends on terrain and vegetation but usually with significant parts wide enough for only one person, with sections not smooth enough for normal bicycle but probably OK for a mountain bike. If there was something like highway=trail I could use that tag without resurveying, and I think it would provide meaningful information. But I simply don’t know whether any specific point on those trails is dirt or grass, or 0.5m or 1.5m wide.

6 Likes