Have a tag to denote fictional "pathless paths" that exist only for routing purposes?

tl;dr, there can be real world impacts from showing paths where there are none, ranging from environmental impact to personal safety. If it makes sense to have a fictional one for routing, have some sort of tag for it along the lines of pathless_for_routing or fictional_for_routing.

I’m sure a better name exists. :slight_smile:


I had some questions on an area with a lot of issues in the US trails slack channel, and amongst practical advice we got into a little bit of philosophy and there was a great suggestion I didn’t think about. I’ll include anonymous quotes from others as well as my own comments and let them know this is here so they can contribute.

There are the routes in question - they are mostly pathless, but occasionally have semi-maintained official paths where headlands are either temporarily or permanently impassible.

There is also a long distance thru-hiking trail that goes along parts of these routes.

The really high level trip planner national parks trail map shows the “route” along the beach, but only had dashed lines for trails for the proper overland trails. This matches how they are treated in USGS maps as well as the current maps endorsed by the park service and sold at their visitor center (someone put a scan of one online, note this violates copyright).

Conversation elsewhere edited for clarity

Comments not by me are in blockquotes.

One of the core principles of OSM that I really like is that of ground truth - it keeps things simple. However “ease of creating a route between two points” can supersede that.

It seems weird to have a set line for a path in a spot like the above. Some kind of node at either end of the beach that are related then a “figure it out yourself” feels more appropriate with more nodes whenever there is a little headland or whatever that could cause issues.

Yeah, I agree with you about ground truth. The example image you shared (the one with the flat expanse of wet beach sand) clearly has no physical path on the ground, so mapping one feels like fiction. OTOH, routers won’t be able to connect the dots (literally) between a trail that ends at the south end of a beach and another one that starts again at the north end of the same beach. So without a fictitious highway=path along the beach, things like measuring distance of the route in a hiking app probably won’t work.

This is an unusual and really clearcut instance of a pathless but nonetheless well-traveled route, so maybe it’s a good case study for coming up with a better way to tag such cases. Maybe we need a tag for highways that are fictitious and only serve to help routers? A tag like that could also be used in pedestrian plazas which are often criss-crossed in OSM by fictitious highway=footway ways whose only purpose is to allow naive routers to route across the space. Or maybe these beach segments of the route really ought to just be mapped as a pair of nodes and a route relation? I’m unfamiliar with the tagging conventions for pathless point-to-point routes but as you already mentioned earlier, other mappers have explored tagging these in the past. This puts the burden on routing software to handle such node-to-node routes if it’s a use case they care about.


You do have a lot of people that just stare at GPX tracks these days and assume they have to follow them

This is definitely valid, and something that I hear a lot from land managers that I talk to. Most people assume the map on their phone is correct and accurate, and they will go to great lengths to “put the blue dot on the dashed line”, in the process trampling vegetation or putting themselves in physical danger.

None of the official maps (or ones recommended by NPS etc) for this route show paths on the beach.

That’s an interesting data point. It implies to me that NPS considers these on-beach routes to be a distinct type of thing from other nearby trails (I assume because they are pathless).


The route can be fatal if you don’t pay attention to tides and tidal restrictions. Likewise there is often many ways to do something that isn’t a hard overland bypass trail.

A common but made up example of a spot: at 2-4 feet you can walk below some rocks on sand, at 5-6 feet you can walk between them on some rougher terrain, at 6+ feet you either need to wait or take an overland bypass if there is one. Trying to come up with a proper difficulty rating would be problematic for routeless routes like this, as at low tide it’s T1, but at higher tides it’s often at least T2-T3 up on rocks above where the sand is.

Things also change not just season to season but after storms more than most terrain - a landslide can cause a 3 foot restriction, sand can fill in or get washed out, trees can get washed up, etc. Some of that can get mapped, but most of it not, or at not it in a timely manner.

PS, insightful side comment from someone

This brings up a minor nitpick on the definition of trail_visiblity – beaches and meadows are “pathless” but “orientation unproblematic”, so don’t fit any of the categories in the wiki.

PPS, some trolling :stuck_out_tongue:

Or we just say that because it’s not in the mountains or alpine it’s all T1 as some of the more hardcore europeans advocate for. That stance makes something like this which starts a few meters above sea level (and is inaccessible except at low tide) odd. I -think- it should be T5, it felt like very easy Class 4.

That seastack an hour and a half later as the tide was coming in:


footway=link is used for places where a connection should clearly exist for routing purposes, but there is no physical footway on the ground. For example when a sidewalk ends abruptly without physically connecting to the adjacent road, but pedestrians can clearly walk across the verge and onto the side of the roadway to continue. The “link” term shows that it was originally intended for short distances, but perhaps it could apply to longer ones as well.


Not directly related, but with your skis you can also route through a Proposal:Piste:type=connection - OpenStreetMap Wiki

1 Like

I prefer the other idea of virtual=yes over footway=link Talk:Key:informal - OpenStreetMap Wiki
I want to keep “link” for connections between parallel (may have many redundancies) or perpendicular (less important than actual main pathways) =footway , =cycleway , and =path . This keeps it similar to =*_link for roads. Such criticism had appeared on the newer proposal. https://wiki.openstreetmap.org/wiki/Proposal_talk:Highway_link#no_real_world_feature_-_non_consistent_with_“link”_concept_for_other_highway=* (although I do think existing use could be considered as a “pathless” “link”)

1 Like

Any such proposal would have to answer the question: What if two people disagree on where exactly the “link” is supposed to be? How can the dispute be settled on the ground? – Because “go there and look at things” is our standard way of solving disputes.

Perhaps it might make sense to contemplate such “links” as a relation object that really only says “routing is possible between these two nodes” without explaining how, and routers could then process that accordingly. Anything that is a way and not a relation WILL show up as a straight line in an editor and then if that straight line doesn’t match the terrain, people will start adapting it…


Perhaps it might make sense to contemplate such “links” as a relation
object that really only says “routing is possible between these two
nodes” without explaining how,

This would break some tens of thousands crossings between two dual
carriageways. These are often a big slap of asphalt that may or may not
have road markings or partial or contradicting ones.

I don’t think that many people contemplate if we break major arterial
roads for principled reasons when there is a clear common sense that
they are connected.

Why exactly should that be different for other modes of transport?

OpenStreetMap takes pride in having topology, and that means: if it is
passable then it should be connected. If there is no further information
then Occam’s razor would suggest a straight line and done.


Whatever the tagging of the connection, I do feel like it is possible and appropriate to use a linear way that approximates the centerline of the area that might be traveled.

Take for example, a route along a curved bay like this example from the South Coast Route linked in the original post:

Even if a “precise” placement on the beach is impossible to determine, a rough estimate is going to allow routers to be less wrong in determining distance to be traveled than simply a relation identifying the two end-points that are always above water.

I do think it makes sense to tag such a connecting way differently than paths that can be seen on the ground, but I have not formulated a strong opinion as to whether this should be a new highway=* value or a separate tag used on highway=footway|path|etc.


I’m using virtual:highway=pedestrian to map connections (for routing) on a pedestrian area.

Usage count is moderate because it’s not supported by any router.
Implementation would be trivial:

  • renderers: just ignore objects tagged as virtual:highway
  • routers: handle virtual:highway=xyz just like highway=xyz
1 Like

I’d say that a map should most of all be useful, and derived principles such as “on the ground” and verifiability should only be applied when they enhance usefulness. A hiker will probably appreciate seeing a line on his OSM-based map indicating in which compass direction he can expect the route he is following to re-appear after fading into a meadow, as long as it is clear from the map that he should not expect to see a path. The tag trail_visibility=no expresses that very well (when rendered as a thinly dotted line) and I don’t think we need to invent a new tag for it. Pathless paths should be mapped as straight as possible, but go around less walkable areas such as steep rocks, dense forest and areas below the low tide line.


Here is the same area in the Strava global heatmap:

It’s often possible to see where people actually walk on the heatmap (which is available for mapping) even when it isn’t possible to see this on the ground. The traces on the heatmap can be “averaged” to draw the approximate line of travel.

Mapping everything off the heatmap has its dangers. I’ve seen pathless routes over glaciers that need special gear mapped as highway=path with no other tags. I wouldn’t recommend this!

But for somewhere you have walked yourself and where you can add appropriate tags such as surface, sac_scale and trail_visibility, cross checking with the heatmap can give better results than just using your own GPX track.


I agree that footway=link is the closet existing thing, but I also think that something else is necessary. Keeping link to its original purpose of bridging short gaps also avoids any secondary issues.

I personally wouldn’t mind just having a path on one side of the meadow and a path on another side, then I know where to aim for but know I’ll have to route find on my own. In such cases I’ll try and find where the trail exits the meadow visually vs constantly checking how far I am from a “path” in the middle of it on my phone.

Having a fake path to link the two together would works, but in that case it would be better to have more deviation to reduce the impact of people walking over the meadow. It also tends to give a false impression that such a path exists, but that it’s just faint or poorly maintained etc.

Yeah a true trail_visibility=no path is complete fiction used for routing or to show an approximation of a known route.

This is outside of the scope of tagging, but perhaps suggesting to routers that trail_visibility=no should be rendered as a translucent wide path (or wavy path, etc) to indicate uncertainty vs just a thinly stippled one would be more accurate if there isn’t some other tag associated with routes such as this.

This is the best way to approximate a linear path, but can also lead people to think that they “need to be on the path”. Land managers are well aware of people blindly following paths on their phones with GPS assuming that they indicate where they “should” be as people have pointed out in comments I copied on the original post.

In regards to surface, sac_scale, and visibility I’m not sure that would be addressed (aside from visibiility=no).

A common but made up example of a spot: at a tide of 2-4 feet you can walk below some rocks on sand, at 5-6 feet you can walk between them on some rougher terrain, at 6+ feet you either need to wait or take an overland bypass if there is one. Trying to come up with a proper difficulty rating would be problematic for routeless routes like this, as at low tide it’s T1, but at higher tides it’s often at least T2-T3 up on rocks above where the sand is.

1 Like

A suggestion in the US Trails group chat was to mark walkable areas - this to me feels more accurate but probably nearly impossible to accurately convey as well. From a previous thread on pathless paths. :stuck_out_tongue:

I don’t know where exactly the borders of this would be, and I’m sure there are small sections in between of higher difficulty that would have to be routed around. No “pathless path” for this pass exists, and I would delete it if there was one as the community norm in the Sierra Nevada is to not map pathless paths - having labeled waypoints of critical decision points is as far as most people go for XC terrain. I did move the pass location slightly since to be more accurate to my GPS survey of the two YDS Class 3 options (SAC T4) to gain the ridge that is otherwise SAC T5-6.

For Valor Pass I know of people that hiked up the ridge on the south side, then dropped as that is a little simpler. We felt perfectly comfortable doing a more direct approach, but stayed above cliff bands (which doesn’t really show as well on opentopo, but with USGS or slope angle shading the map gets noisier for this purpose). This is a pretty vague representation based on my memory - and isn’t entirely accurate as there might be some 2-3m cliffs or ledges etc that could pose a navigational hazard in the middle of that area.

This tells the story pretty well, but is probably a bit overkill and prone to small errors in terrain. In this case just having a node at the top of the pass seems fine to me, the rest doesn’t have any needles that need to be threaded (aside from perhaps the chutes dropping from Valor lake to Ambition, but at some point you have to consider the route to the pass done).

Thinking on this some more, there are some T5-T6 " pathless paths" where there is a narrow route that goes through more technical terrain. Granted trying to show a fairly vertical route on a top down map is of limited use, but there IS a difference between “this path is hard to see / you cannot see it at all” and “there is no path through this area and where you should go will be constantly changing”. The two ideas are closely related, but not always the same.

One personal anecdote is the east drop down Gemini Pass, while the top and bottom are trivial the middle is fairly steep (35+deg) and loose. My partner chose to step plunge down a soft snow field in the middle which I was not comfortable doing but she thought was straightforward. Once she was done safely and moved out of the way, I boot skied down scree off to the side which was collapsing a meter above and below me. I thought was great fun, she was terrified watching me. Trying to just put a trail_visibility=no in pathless terrain indicates that there is a set “path” that people should try to orient on - but even with the same conditions different people will choose different paths!

I don’t think we can assume that trail_visibility=no by itself communicates such ambiguity - in the case of the south coast where someone should actually hike can vary depending on tides, logs washed in, whether they are prioritizing looking at tide pools or making time higher up etc.

To use the strava heat map above, where is the exact pathless path in this location that just isn’t visible?

I think using the average of paths taken in heat maps to generate the actual way that would go into OSM as a sort of centered zone of ambiguity makes total sense. However just rendering it as a lightly stippled line implies that that averaged path IS THE PATH and people should try to follow it even if they cannot see it, which is not the case.

1 Like

What about using an Area to denote the general “walkable” space, with a Line passing through for routing purposes. That would basically be the same as a pedestrian zone in a city.


Mind you that in such pedestrian areas very often footways get mapped. So-called plaza-routing still in infancy.

1 Like

Using an area to denote a walkable space works and was a thought that came up earlier - linking the beaches with the actual overland paths that go over impassible or tidally restricted headlands feels more accurate than trying to force a set “path” on them.

In the case of the hike here (which is probably an extreme if not unique example) the walkable space is constantly changing however. The area around diamond head is an intertidal blockfield that is only walkable for a few hours a day, and can be unwalkable for days. I suppose it could be tagged as intermittently walkable?

The line passing through a walkable area for routing purposes ideally would be styled differently from a non-fictional path, at which point some way to tag something as fictional seems useful.

update: This would make surface and sac_scale keys unusable, but I don’t think that’s a dealbreaker.

Many will disagree, saying this is a solved problem, we have “trail_visibility=no”, that is all it takes. And it has been approved long since.

Trying to second guess what you are after: Pathless paths I never heard of that, pathless routes I am quite aware of. Problem: Openstreetmap has no concept of that. Routes always are ordered collections of paths.

Adding to the confusion: path is physical, route is relational.

1 Like

Yeah I use pathless paths in the OSM sense of “path” and consider these routes as well. NPS describes the area in question as the “North Coast Route”.

Showing a very precise path with light stippling for trail_visibility=no (which is the conventional wisdom) indicates that there is an actual path where that line is, just that it’s impossible to see (or for say bad/intermediate it’s hard to see or is the commonly used path between two sections that are easy to see). The actual best route could be a hundred meters away depending on conditions and personal preference.

While it’d be confusing to have the OSM definition of a route overlapping the alpinist one, perhaps there could be a path=off_path_route or path=shifting_route something. path=approximate_route is probably clearer, but feels more like it needs a FIXME vs there not being an actual path. path=pathless would please me.

A recommendation to just render all trail_visibility=no differently (a faint translucent wide line, a wavy one, etc) would work overall, but there are some technical routes that are more precise despite not being visible.

Only a handful of countries in the world choose to render trail_visibility=no routes as paths on their official maps, it’s unclear why that should be the default for OSM.

Someone made a pathless path consisting of two nodes near Yosemite Valley which gets millions of visits a year. Search and Rescue (I assume) put this in the FIXME:

This is not a trail. It frequently lures hikers from the North Dome Trail towards the north rim of Yosemite Valley and subsequently into technical climbing terrain causing many search and rescue incidents. Please remove this trail.