RfC: Have pathless hiking routes (no trail_visibility) be some manner of related nodes instead of paths

Overview & Issue:

There’s a number of paths that are… shockingly straight lines that completely ignore contours. I haven’t been on all of them, but I’ve been on some and they feel much more like a route in the western US sense of the term - a cross country path (think Roper or Skurka) where users are expected to use their own discretion routefinding but make their way from described locations or waypoints to the next ones. Drawing lines in between widely spaced nodes to make a path is a poor experience.

Having routes be a set of numbered waypoints would have some clear upsides:

  • less impact created by people focusing on an arbitrary track and creating a trail where none existed
  • lless confusion by people that expect a trail when it shows up in a mapping client
  • clarity that it is meant to be an XC route leading to some location (usually a peak or named feature)

One thing that I’ve found useful and has come up in the #trails channel in slack is using node networks for these routes that are pathless. I think node networks cover the limited use cases while being more true to the spirit and skills required.

A major use of them would be for commonly used but informal routes, but these formally exist in Austria as “alpine routes” and they’re just being added as paths at the moment.

Proposed General Solution

After some discussion, come up with a node_network for such off-trail pathless routes ala Climbing Routes, which was also suggested by the OSM Trails working group.

This is already a commonly accepted practice in the general alpine off-trail community in the USA at least (don’t share tracks but waypoints, don’t create new trails, etc). While I personally think the proliferation of acronym routes is a bit overdone and I think defeat the purpose of XC, this image is an example of how a “route” can be loosely described by waypoints/nodes

Aside: I’ve done portions of that independently of any guide, and the marker saying to keep on the W shore of South Guard lake is poor advice early season when there’s steep snowbanks along it, but that’s the nature of routes and why people should be able to improvise based on conditions. If “connecting the dots” is too hard for them, they should stick to paths or do simpler routes until they have that skillset.

I have a solid mind for systems, but am pretty new to the details of OSM, so feel free to suggest major changes to this section!

Have the overall map object be something like:


Climbing - OpenStreetMap Wiki already exists and can be used for inspiration. Some suggested nodes:

pathless-hiking=top would generally be a peak or pass. note that a lot of off trail passes & cols are already somewhat arbitrary and are not natural saddles, so these are really standalone nodes for pathless routes that exist because they are a “named place” if informally.

pathless-hiking=bottom would sometimes/often not exist, a lot of times a pathless_route up to a pass starts from the shoreline of a lake, or the side of a trail and is up to individual routefinding preferences until the terrain gets steeper/tighter. this could be a way as well as a node, indicating a general starting area?

pathless-hiking=important could be a part of the route where more attention to routefinding / orienteering is necessary to avoid having to backtrack and/or adding extra risk. this could also be pathless-hiking=decision_point, pathless-hiking=crux, etc with a description on it. In the Sierra Nevada, the Class 2 ledge (SAC 2-3) which is the only way around a Class 4 drop (UAII 3) for Finger Col, or the small pond along the Cirque Pass route with a deceptive grassy shoot leading to exposed wet cliffs to the left, or straightforward terrain on the right when dropping.

pathless-hiking=key-marker a location of a key cairn or marker on the route. should be used sparingly as to not just create a node_network of hundreds of cairns essentially creating a path

Not having these be rendered in mapping clients off the bat seems fine to me, IMO they should be an optional map layer (Gaia GPS), or rendered very softly (Caltopo).


So far everyone has been supportive of the general idea (or abstained) in the #trails channel on OSM Slack and in a related thread here on trail_visbility. I can link to those in a follow up comment for visibility.

Some relevant comments from Tag trail_visibility: Proposed Improvements for this Descriptive Tag

I started a discussion on this topic a week ago on the #trails channel on OSM slack, the general idea of these “pathless paths” either not existing or being a node_network has been unanimous. I’ll include some of my thoughts from there as well.

You can join this slack domain here https://slack.openstreetmap.us

It’s worth pointing out that trail_visibility=no as originally intended was probably meant to be used for paths that had some markers on them, not just for trails that don’t have any visibility. That seems like a case where the shift in meaning of the key has led to unintended consequences. IMO those should be redone as node_networks with minimal key nodes vs “paths”.

I’m personally against creating paths out of random off-trail GPX recordings. A lot of times there might be a key section to an off-trail route, but the rest is pretty open to personal route finding and can change based on snow levels, etc. Having them there as a path just promotes the creation of more informal trails/erosion and people staring at their phone instead of reading terrain. Probably around half my backpacking is pathless and markerless, but I don’t think it’s appropriate to create an OSM path out of the routes I’ve taken even if they worked well. I’ve deleted a few paths that were pathless and markerless (and happened to have poor routefinding as well) that I’ve come across.

In regards to the principle of reproducibility there’s a big difference between seeing a possible route, and seeing a trail which has been caused by repeated traffic over the same location - that’s the “ground truth” where I would draw the line at making something an informal path vs node network of loosely spread out markers.

I can easily see this getting out of hand - if people create node networks for every possible navigable path in the backcountry (like how some people create huge cairns to regularly mark a strong flowing outlet to a lake in a remote backcountry basin) the map becomes a cluttered mess of people “finding” routes that have been used for many decades in a fairly similar fashion.

IMO most off trail passes in the Sierra only need a top, a small minority need a bottom… but where is the bottom? it’s hiking up from a lake shore or something, or would that be the nearest trail, or would it be a final push to the ridge where things are steepest and it’s most critical to be “on route”? I would put the passes that need a middle node on one hand, but I’m also experienced at routefinding and others might want a node every couple hundred of feet.

For informal routes I’d be wary of marking cairns (there’s some XC cuts, like Simpson Meadow to East Pinnacles I’ve done three times, differently, none terrible, and encountered a few cairns each time lol) - I feel like there could be a good mentality of putting a few extra nodes on a route with a description field (after this pond turn right to stay on Class 2, don’t go down the grassy ramp for Class 4) but leave someone getting from the top of the pass to that midway point on their own - they can pinball from side to side looking for easier drops on shelves, or drop down more directly if they’re comfortable with simple unexposed climbing

I destroy hundreds of (non-digital) cairns a year and make less than 5. It has to be one weak area of an otherwise strong abandoned trail, or something informal where missing that one non-intuitive spot is a genuine issue of safety. Finger Col ledge being the only Class 2 option out of a Class 4 area felt fair to me (though I knocked down random branching ones below it I built up the existing cairn by the ledge) or the two Class 3 options out of Class 4 drops on Valor Pass I chose to leave the cairns undisturbed as I didn’t encounter any for what up to that point is “walk up, nearly anywhere, from nearly anywhere on the shore of this lake”. There could literally be hundreds of specific routes for the same off trail pass.

While this isn’t what I would put up for a public pathless_route, but it illustrates the node network idea (I took a screenshot of personal waypoints to share with a local). I’d drop some of them, and clarify others for public use, but this feels much more “XC” than having a path rendered.

My own intuition (as a computer scientist more than as a hiker) is that we’d be better off treating the types of routes and the types of (non-)paths as orthogonal issues. Both topics are complex and ill defined but maybe there is some complexity that can be avoided by keeping something like a “physical surface vs information space” distinction. There is so much to sort out between node networks and other routing systems that I fear that adding non-paths may confuse the issue beyond recognition.

I remember that a few months ago I tried to steer the debate about scrambles towards the more general issues of non-paths: beaches, rocky places, some meadows, etc. There is a similar debate among cross-country skiers too: map a fake path or a surface? Could it be that there is something to gain from considering “generalized paths” as a combination of nodes, ways and areas?

A question on terminology - not a native English speaker, telling the obvious highly appreciated:

Path and Trail are synonyms and cover what is there on the ground. The imprint of human traffic on the surface of the earth.

Route covers the idea, commonly expressed as a verbal account of how to get from here to there, e.g. “Follow that path up until … look out for bootprints pointing West; Head straight up the streambed; Just keep on the arete, pass the pillar midway on the South side; &c.”

PS: Thank you for “bootprints”, all the online dictionaries failed me on “Steigspuren”, a very popular term in German literature.

Another question on terminology: Nodes do not make a network, when there are no edges connecting them. In the hiking domain, highway=path is the most used tag for such edges. Drawing an OSM way turns a mere collection of nodes into an ordered set. Bonus: Makes a software routable graph. All fine, where there is a path on the ground too.

Looking at the picture how to route around the gully, while getting sweaty palms from the dark shadows lurking in the lower end, I observe, that the nodes have clear succession mostly imposed by terrain only. Do you suggest to mirror that in data? The method then should be as easy to maintain as a highway=*, if it wants to be successful, needn’t it?

1 Like

Was reading a little while ago about node networks in the Netherlands and Belgium


Maybe some idea can be pinched from this.

A clarification question. Are you familiar with node networks in Belgium and the Netherlands? They’re a very specific thing: they consist of numbered signposts at junctions like this that tell you which way to go for which neighbouring node.

They solve a particular problem: in a country full of criss crossing paths and roads, they answer the question of “which way do I go at a junction” much more flexibly than the traditional approach of signposting linear long-distance footpaths. You can now describe a route as 54-43-05-21-54 and follow it without even needing a map.

Are you suggesting to re-use the node network tagging scheme (that is, relations with network:type=node_network) for the purposes you have in mind? What you’re describing has some superficial similarities (navigating from node to node) but the existing node networks are not for hiking in pathless terrain. The Belgian and Dutch nodes are connected by ordinary paths and roads. Using it for off-trail hiking could really confuse data consumers.


I have the same understanding as you, but I’ve often seen the word ‘trail’ used (improperly?) to designate routes. In French we also see the same pattern: “sentier” and “chemin” mean “path” but they are often used to casually designate routes.

As for your definition of “route”, once again I have the same understanding but the original message of this thread has me wondering. What lies between two cairns in a rocky area? a path, a non-path, a particular kind of route that fills the void where a path should have been if the ground was different?

This seems to be trying to address an issue I’ve alluded too a few times, but I did write it up for back-country ski routes (not marked, visible only in certain circumstances, often described in guidebooks).

I was pondering this last night and was thinking that a pathless route would almost work better as an area - it could be wide at the base, then narrowing at chokepoints and then widening back up again. I’m not sure how much value it has that point, but another approach to pathless routes is for people to draw very wide lines over a map.

We already have people adding non-paths as paths, which to me is less than ideal.

I’d also honestly be fine with just not allowing pathless routes - we already have some passes and cols which are not natural saddles, so are essentially the top node of a route that get away with being an arbitrary “feature”. I think for some of them having a bottom node would be good (Little Joe Pass can be hard to locate without GPS). My drawn up notes for a community elsewhere as an example (though I’d cut the vast majority of waypoints here, I’ve done two variations from the bottom and would consider getting to the base of the chute from Lake Reflection up to the person attempting the route with no nodes to guide them - I kept them in this photo for the context of how to find the proper avalanche chute to go up). Pardon the language at the top, it was quite a slog getting up there and I was tired lol. I marked in blue where I would have networked nodes for this route - four nodes, top, bottom, and the two main forks in the chute.

I’d say that in OSM terms a path can be a trail, and a trail can be a mix of imprint of human traffic on the earth and markers (where such imprints aren’t possible). A route can contain a path or trail, but may not. There’s also the distinction of routing or route relations, which is how to decide which path(s) to take to get to your desired destination.

So a route can have a path or be pathless (orienteering / routefinding), if there is a path it can take the form of trail or road or whatever.

Having a sequence would also be necessary to avoid ambiguity, sometimes one has to ascend or descend a certain area and traverse then go up again and it’d be hard to tell which horizontal marker to aim for. My screenshot here has a lot of extraneous points in it (things I found interesting, extra notes to myself, etc), I think in general there’d be 3-5 markers per side of a feature to minimally define a route, which shouldn’t be too hard of a burden to create.

We were about 80-90% sure it’d work out and definitely scouted it out visually and talked it over before trying it out and were willing to turn back and just head back via trail instead of creating a loop. No direct exposure despite the steep shadows.

I’m not familiar with them, it just came up as a suggestion that seems more sense to me than drawing a set path between nodes. I don’t think it should have the exact same rendering - I think having it follow more in the climbing route example where you have a top, bottom, and bolts in the middle makes more sense though I can’t think of any big wall climbs where they’d be easily visible on a map.

I do think that taking that idea to work off of makes more sense than modeling it after a road or trail as a distinct path. Having numbered nodes of a route (Cirque Pass Top, Cirque Pass 3, Cirque Pass 2, Cirque Pass 1, Cirque Pass bottom) etc with some descriptions on them or a short sentence if zoom the map in would make sense.

What lies between two cairns in a rocky area, according to OSM norms, would be a path unless the cairns are so far apart that it’s hard to see them. I’d personally view that more as a route (there is the ability to routefind your own way to the next cairn, vs following on top of bootprints on a surface) but since there are many formally maintained and recognized trails that consist of regular markers over terrain that won’t hold imprints/wear on a surface I think just saying that there is a path there is fine. If there’s a marker every 10m or 30ft in open areas I think that’s “pathy” enough, vs say a cairn every quarter kilometer or cairns that random people put up on their own versions of a route in terrain that has no formalized path on it.

The “Lost Canyon” trail in the Needles district of Canyonlands National Park is a nice single track dirt trail when it is in valley floors, then switches to being markers only (cairns) for going over a ridge, then switches back to single track again. If the conservative (in terms of not wanting tourists to die and have to medivac them out) national park service calls it a trail who am I to argue?

And looking down at the same “trail” earlier that day where it is clearly visible single track.

If there’s just a cairn here or there it’d be trail_visibility=horrible and/or trailblazed:visbiilty=horrible trailblazed=cairn. At the point where we have a trail_visibility=no which is entirely or mostly pathless (or IMO a route which is bad or horrible visibility with only markers and no path) I feel like it should just be rendered as a route, or as fragments of a path where there is a use trail with blank spots in between for people to figure out for themselves. There’s a lot of broken “fragments” of paths in the Sierra Nevada mountains which seems appropriate to me in terms of ground truth.

That link isn’t working for me. :frowning:

Works for me both in the quote & original post. Wonder what the issue is.

Ah, it blocks my VPN. Loaded it now.

Routes. To a certain extent showing routes spoils some of the key aspects of being away from the piste. Picking ones own line, both in the light of one’s ability and safety of surroundings is part and parcel of developing the skills and awareness necessary. In many places (Albona at Stuben, most of Les Grands Montets below 2800m, Pavi in A-basin) all that is needed is to show that the area is free-ride territory. In others entry lines and exit routes may not be immediately obvious and are worth showing (it is not uncommon for these to be the least pleasant part of a route), particularly if there are cliffs or other hazards lower down (e.g., above Le Fornet, Val d’Isere).

Very similar thoughts here. Hazards seem to be more snow orientated - trying to map every area with loose talus or exposure doesn’t seem feasible, and there’s less surprise due to going at a slower pace and not (generally) having terrain buried under a potentially thin layer of snow.

For your purpose, having a general area with optional entry/exit lines seems to make sense, and that’s something I considered in my comment above (though more a node for top/bottom than lines/ways). It’d be a bit noisy for a general map, but could work if rendered as a separate map layer by clients that would be opt-in.

“Trail” has a distinct meaning in American English (and to some extent Australian/NZ English) which is closer to “Route”. It is not often used that way in British English. I would therefore generally advise against using it in OSM to avoid confusion. I realise this is a bit of a horse-bolted[1] scenario in a thread about trail_visibility but there you go.

[1] es ist ein roß entsprungen, as the hymnodist wrote.

1 Like

To a certain extent, though that’s regional. I always thought of trails as the compacted/scarred surface over terrain growing up in California, after spending more time in the southwest where trails can consist just of cairns for periods of time I’ve had that narrow definition expanded but still think of a trail as a nice single track in my head vs a route. The NE has a lot of tree blazes, Acadia has cairns, etc.

There have been some calls for the new trail_visibility to have an alias path_visibility that would be used going forward with trail_visibility being deprecated and eventually ported over, but it hasn’t caught on. IMO that’s a bit clearer if it encompasses both surface and markers. You can chime in at Tag trail_visibility: Proposed Improvements for this Descriptive Tag where this thread/issue was forked from to help keep it on topic.

1 Like

I like the diplomatic answer, in OSM a trail can be a path or it can be a route :wink: (Thereby ignoring the where clause. But who is going to honour that?)

In my area, trails where on the ground there is not much bootprints, but where there is a steel cable running all along (not of the via_ferrata kind, but the assisted_**trail** kind), get tagged trail_visibility=horrible. If trail_visibility is to start meaning any clues, we have to change some mappings, some dating back ten years.

The name of the song, the Christmas carol, is “Es ist ein Ros’ entsprungen” - a rose, not a horse :wink: It is a very nice song, I never managed to sing it correctly.

I’d just appreciate if debate here kept to one meaning, so I do not have to guess every mention of the term. At least path seems unambiguous?

A bit of digging found this, birth of the idea for the sac_scale tag - Re: [Talk-de] Wanderwege, Klettersteige, Kletterfelsen

Here Re: [Talk-de] Wanderwege, Klettersteige, Kletterfelsen it was voiced, that a new base tag highway = trail might make sense, and one highway = via_ferrata just as well. But this immediately got voted down, to not create more highway-inflation, and use highway=footway instead. Discussion then turned on to cycleways.

So, that should explain, why its called trail_visibility and not path visibility, because the brand new path only came little later, but just in time to make it the preferred base.


1 Like

maybe I’m becoming too conceptual here but could we say that, in the universe of routes (people advising other people where to walk), a path is a special case of route? That is, it has its own way of conveying information whereas other routes rely for that on cairns, on signs, on lists of pre-existing routes, etc.

This would open a way to situations where a route can be made of routes that are either paths, composite routes themselves made of other routes… and other objects (non-paths) made for instance of ordered lists of nodes ?

Obviously @Richard has been entertaining some unusual ideas about this song. :smile: :smile_cat:

I have been looking for some suitable examples and this precisely relates to one I have found.

The Westgrat of the Muttler is mapped as a path on OSM, but is not marked at all on the SwissTopo maps. There are some good pictures in this account on Hikr showing that as the rock band is reached there is a cable anchored to the rock, implying some existence of a route

I’ll add some more to the thread later, for instance Fluela Wisshorn (GR, CH), Piz Minschun (GR, CH) and the Bliistock (GL, CH)

It’s a famous mispronunciation by non-German speaking choristers!

(Though the finest version is of course the one set to the English translation ‘A Spotless Rose’: https://www.youtube.com/watch?v=QQnApR9PNV8)