Consuming highway=path

ezekielf · 2024-08-30T18:05:11.703Z

GreekMountains:

An alpine route would only be in the alps???

Aah yes this is confusing, but the English word “alpine” can refer to large mountains in general, not just specifically the Alps.

n76 · 2024-08-30T18:08:54.457Z

GreekMountains:

I agree with what you are saying. In addition many Dirt roads are not updated for understandable reasons- only a few people go down them and they are not map editors!
I ride a dirt bike as my main sport/ hobby and seek out these things all the time.
I am also a little confused as what an “ alpine route” is as there is only “ path” and “ dirt road “ available to render in this situation.

An alpine route would only be in the alps??? Surely we do not want another tag for one specific mountain range

“Alpine” may have originated in the Alps but the term is used outside of that area. At least in the US high elevation mountains, usually those that extend above the upper timberline, are described as having an alpine environment. I believe that term is used in most of the world but cannot be sure.

And there are definitely routes, some might hesitate to call them trails, that would be well described by the definition given for sac_scale=alpine_hiking in the wiki.

Hungerburg · 2024-08-30T21:28:01.710Z

bagis:

the Swedish word “stig” which is strongly connected with a path in the woods

This reminds me of the German word “Steig”, which connects to steep paths in the mountains. The term “Pfad” I only know from pathfinder clubs where it is used colloquially, else it reminds of poetry. Generally I think it connects to single-trail walking paths. I would be stunned if the German locale of StreetComplete used that term to refer to cycleways.

This further reminds me about my amazement, why this topic draws no attention at all from the people that use path in the sense of “shared use 2m wide paved urban infrastructure”. I think, they are just too complacent in being right and pathless routes only a niche nobody cares about.

I guess most consumers of OSM data belong to latter group?

Peter_Elderson · 2024-08-30T21:50:22.674Z

I get the feeling that we agree that a highway=path can vary widely in access and in physical characteristics.

If access and physical characteristics are tagged on the feature, data consumers can use the extra tags for their target groups and applications, such as maps and routing profiles.
If they don’t, they are not optimally serving their users.

For highway=path without any other tags, data consumers who do use the differentiating tags, will have to decide on default values for those tags.

The access tags are easy: world wide defaults are: general access=no, foot=yes, bicycle=yes, horse=yes. I would add: oneway=no.
What, if any, are sensible world wide defaults for width, height, visibility, incline/decline, smoothness, surface, various skill scales et cetera?

e.g. Could a minimum width of 50 cm and a maximum width of 200 cm be assumed, with a tendency toward the narrow end, meaning that in the absence of any width tag, a path could be rendered as 1.5 m wide, which physically allows foot, bicycle, and horse access.

dieterdreist · 2024-08-30T23:25:33.000Z

e.g. Could a minimum width of 50 cm and a maximum width of 200 cm be assumed, with a tendency toward the narrow end, meaning that in the absence of any width tag, a path could be rendered as 1.5 m wide, which physically allows foot, bicycle, and horse access.

there is no minimum width, but width for unpaved ways is generally more difficult to estimate. For a path on bare ground, being just 20-30 cm wide on the ground is no problem if there is space above for the body

Msiipola · 2024-08-31T06:39:58.052Z

Peter_Elderson:

horse=yes

Does “horse” mean all animals which are used for riding? Camels, elephants etc.?

Peter_Elderson · 2024-08-31T08:23:17.016Z

I was considering what would be a suitable default when width-tags are missing. This could be reflected on a map, or be considered by a router for a specific transport profile.
A default is not ‘what it can be’, but ‘what is the best guess’.

Peter_Elderson · 2024-08-31T08:36:36.402Z

Msiipola:

Does “horse” mean all animals which are used for riding? Camels, elephants etc.?

Would an elephant need an existing path?

SomeoneElse · 2024-08-31T09:01:26.599Z

Peter_Elderson:

or be considered by a router for a specific transport profile.
A default is not ‘what it can be’, but ‘what is the best guess’.

While “best guess” might work in some situations (a blog post about how wonderful OSM coverage is, for example), but I’d suggest that a router should only suggest a route that it knows will be OK. The access tag “best guesses” that you suggested above are just plain wrong in the countries I live in and visit most.

If a particular tag (such as highway=path) has such broad usage - even if you exclude frankly silly examples like the “scuba path” mentioned elsewhere - that it can’t be relied upon for a particular mode of transport without other clarifying tags, then a router shouldn’t suggest it.

Peter_Elderson · 2024-08-31T10:18:07.740Z

SomeoneElse:

I’d suggest that a router should only suggest a route that it knows will be OK.

That would mean highway=path without other tags is not routed at all. Is that a realistic approach?

Msiipola · 2024-08-31T10:27:10.426Z

I don’t own a elephant, so I don know!

SomeoneElse · 2024-08-31T11:34:08.557Z

Peter_Elderson:

That would mean highway=path without other tags is not routed at all. Is that a realistic approach?

For any mode other than on foot, absolutely. If the mapping in a particular country is so poor than most highway=path don’t have other tags, then a sensible approach would be to penalise “unknown quality” routes as if they were a muddy, boggy, path (which of course it’s entirely possible that they might be).

ezekielf · 2024-08-31T18:50:42.726Z

Peter_Elderson:

That would mean highway=path without other tags is not routed at all. Is that a realistic approach?

In my opinion, yes. To me it feels like highway=path on its own is nearly on the same level as highway=road in terms of what data consumers can glean from it. i.e. very little.

Here is an overpass query attempting to identify instances of highway=path without any more specific descriptive tags. I’ve not been able to run it worldwide without the request timing out, but in every area I’ve run it, there are many results. They represent all sorts of different path types as well as things that aren’t paths at all. Don’t trust this tag.

Duja · 2024-08-31T19:33:34.218Z

On the contrary, in my country and a good part of the world highway=path is pretty well-defined as a countryside trail, with implied surface=ground and access foot=yes; vehicle=no; bicycle=perhaps_but_bad_idea. I’m willing to bet a considerable amount of money that over 95% of highway=paths in the Balkans are like in the leading photo of the Wiki page:

commons.wikimedia.org

File:Forest path and trees.jpg

(And I’ve just edited a path in the Dolomites where I hiked today). I really don’t know the history of the tag, but in many places it just emerged spontaneously to denote a “hiking trail” or something resembling it. I was surprised when I found out usages in Germany and the Netherlands describing a paved, urban, combined foot & cycleway. In my view, that was an abuse of a tag intended for something else.

(I don’t have a very strong opinion about whether path should be used for SAC-scale alpine trails – but that’s another topic).

ezekielf · 2024-08-31T20:32:42.129Z

Duja:

On the contrary, in my country and a good part of the world highway=path is pretty well-defined as a countryside trail, with implied surface=ground and access foot=yes; vehicle=no; bicycle=perhaps_but_bad_idea. … I was surprised when I found out usages in Germany and the Netherlands describing a paved, urban, combined foot & cycleway. In my view, that was an abuse of a tag intended for something else.

I share this point of view. If highway=path were universally used in alignment with your definition I’d be much happier. That is how I’ve always used the tag. Unfortunately, there are many uses that don’t align with this narrower definition. The seeds of this are right in the original proposal which aimed to potentially replace highway=footway, highway=cycleway, and highway=bridleway. This didn’t happen but the idea that highway=path can be for anything called a path took hold anyway.

Hungerburg · 2024-08-31T22:15:45.689Z

Duja:

describing a paved, urban, combined foot & cycleway. In my view, that was an abuse of a tag intended for something else.

A counterpart of yours in Austria tagged a UIAA VII climbing route as a path and fights viciously for it to remain a path and get rendered by any and all consumers so it gets the variety it deserves.

Needless to say, there are also other mappers that insist that highway=cycleway is feature of the past and the canonical mapping is highway=path+bicycle=designated.

Woazboat · 2024-09-01T12:39:56.246Z

bagis:

For me highway=cycleway and highway=footway just adds confusion, deprecating them would direct mappers and consumers towards handling highway=path together with other tags.

ezekielf:

The seeds of this are right in the original proposal which aimed to potentially replace highway=footway, highway=cycleway, and highway=bridleway. This didn’t happen but the idea that highway=path can be for anything called a path took hold anyway.

A big reason why that approach was bound to fail and leaves us with the mess we currently have (even though it does make sense in a way) is that we do not really have any tagging schemes that describe the practicality of using a path with a bicycle, etc…
The access tagging scheme is only concerned with legal matters (which is a good thing, don’t dilute it by mixing other things into it).
Tags that describe the physical properties such as surface, smoothness, width … can help to make a judgement whether a path is suitable to ride a bike on, but they are very time consuming and difficult to tag. Picking the right values can be hard even when you are right there, standing directly on the path you are trying to describe. Doing it accurately using only aerial images, even after having actually walked a path is bordering on impossible.

If I see a mountain hiking path on the map or on aerial imagery, I know that using it with a bicycle (not talking about MTBs) is completely impractical, but I have no way to actually add that information.

An approach like that would probably have worked for ‘normal’ roads for cars. (Car-)Roads have far less variability than paths for pedestrians/cyclists/horseback riders/hikers.
The difference between highway classifications (highway=primary/secondary/tertiary/…) is often not really distinguishable on the ground and doesn’t change the main conclusion that can be drawn from the tag: ‘car goes brrr’.
The importance/classification of a road could just as well be added as as sub-tag without massively changing its meaning.

Peter_Elderson · 2024-09-01T14:46:40.507Z

Looks like defining “what a path is” is not a viable option.

I think that a lot of issues would be solved by agreeing on
a. Acceptance of the range of real-world features tagged as highway=path (and describing this as fact-of-life);
b. Establishing a set of tags that adequately describe specific cases or country-specific rules
c. Establishing a world-wide set of implicit and default tags;

This would be the anchor for mapping, validation, QA, rendering and routing, regardless of which tags are chosen. It would be a compromise, so everybody will be unhappy about some aspects, but also everybody should be happy about other aspects.

I think a. and b. are not that problematic. Current practice is well underway.

c. is a subset of b.
Establishing this would require mappers and mapper (sub)communities to look beyond “what is a path in my neighbourhood/region/country” to “what is a viable world wide default”.

Example, not for path but for cycleway: in Nederland a cycleway along a road is by default oneway, but the world wide default is two-way cycle traffic. So we explicitly tag oneway=yes on all oneway cycleways along roads, AND oneway=no on all two-way cycleways along roads. That’s a bore, but we mostly use presets for this, and that is the thing to do once world wide tags and defaults have been established.

Hungerburg · 2024-09-01T21:20:40.801Z

Peter_Elderson:

describing this as fact-of-life

I was so obnoxious to tag the “path” where @Superfebs got lost a highway=route, because the mapper said, there is nothing on the ground and they made it a “grey way” not a “red one”, as if openstreetmap was a paint-book. Not a day after it got retagged a highway=path and will again be rendered red in Tracestack Topo. See Way History: ‪Sentiero 30‬ (‪1307520308‬) | OpenStreetMap and Changeset: 155205011 | OpenStreetMap for discussion.

Mind you, @Superfebs managed to go down a pathless UIAA III route without taking harm. How many of the participants in this topic can claim doing that?

So, yes, highway=path can mean most anything, even immaterial stuff, and that is a fact, as put forward by CAI Salò by reinstating a path where there is none.

Curiously, efforts to tell consumers of this data about that get ridiculed Consuming highway=path, Episode 2 - #3 by Peter_Elderson – All the while consumers are expected to honour these elite tags and for ages they fail to do that. Who is more stubborn, the community or the consumers?

Peter_Elderson · 2024-09-02T07:47:23.423Z

Hungerburg:

Curiously, efforts to tell consumers of this data about that get ridiculed Consuming highway=path, Episode 2 - #3 by Peter_Elderson –

That was not the intetion, sorry if I gave the wrong impression.