To build a consensus, could anybody propose a photo of an asphalted path in an urban environment without cycle-related signs and mapped as a highway=path and not highway=footway ?
It’s also about legal access. Data users should be enabled to consider a. whether one can legally access the way with a particular mode of transport, and b. how practical/feasable/comfortable/hazardous this is, what to expect. Routers can discard ways without legal access, and penalize discomfort/obstacles/danger.
In OSM, these characteristics can be implicit for the object, e.g. a highway=cycleway is implicitly bicycle=designated. Or they can be OSM-default, e.g. default bicycle access on highway=footway is bicycle=no. Else, they have to be tagged explicitly.
The path case shows that almost anything goes. This is why, to me, it is important to
a. describe the full spectrum of actual use of highway=path;
b. describe implicit tags (not very many, I’m afraid);
c. agree on how to tag the characteristics of the many variants of path (with options, because different schemes can be valid);
d. agree on a set of OSM-wide default tags, that can be assumed by data users if not tagged explicitly.
What’s the goal here again? Warning data consumers that the range of what can be mapped as a highway=path in OSM is extremely wide (and that therefore they need to consider tags like surface and sac_scale for routing), or giving examples to mappers of paths that the community agrees should be tagged as highway=path? Because these are two very different goals.
I thought the goal was the former, and if so then the shared-use path (first picture in this post) is fine, but like I said somewhere else, the whole collage should be accompanied by a warning that not everyone would tag all of these examples as paths.
To me at least, it seems that your point d:
contradicts something you (I think correctly and importantly) said earlier in this thread, namely:
As I noted previously, most highways in OSM would benefit greatly from the additional tags, not just paths.
Duck Tagging any highways is in my opinion a perilous convention precisely because of the points raised in these threads. It is, of course, handy to have a way tagged only with a highway=cycleway/footway/path tag if and only if the only other option is having no way at all.
I nevertheless think we should insist (e.g. by trying to build consensus on the additional tags needed, more text and example photos in the wiki pages as suggested in this thread, and after that using QA and QC tools) that such ways are by definition highly imperfectly tagged and in dire need of further descriptive tags. This is precisely because there are so few things one can infer in a truth-preserving manner from the highway tag alone (at least for cycleways, footways and paths).
I don’t see a contradiction here. What I am after is acknowledging the wide variations, not fixating things on narrow or country-specific definitions, and at the same time establishing a clear set of suggested OSM-wide fallback values for data users.
I’m not trying to nitpick here, but precisely that enterprise sounds contradictory to me. That is, there are very few fallback values for any of these highways. Mostly access-oriented, I suppose.
I think acknowledging (& documenting, etc.) the wide variation sound excellent, but I think we should insist on further descriptive tags rather than finding some nebulous fallback values and encouraging Duck Tagging further.
Exactly, and that is what underlies most of the issues.
A tagging scheme to describe all the variations, excellent. But that does not answer the data user’s question: if further descriptive tags are lacking, what do I assume, or do I leave it out altogether? Do I take country-specific notions of what a path/cycleway/footway is, and if so, which countries do I support?
I think the way to go is to define a fixed set of OSM-wide defaults for the descriptive keys.
2 Likes
I also do not think the two are contradictory. It seems to me there actually is a lot of implicit meaning assumed anyway, it is just sometimes wrong.
I think that cycleway and footpath have implicit surface=paved or at least compacted. Bridleway and path have surface=ground.
For all, foot=yes.
For cycleway, bicycle=designated.
For path, foot=designated.
The position of the wiki, in my opinion would be “these are often the case but not always, please do fill them up” and tools should complain at least about surface (designation I think is more solidly implicit).
In practice, for path, sac_scale=hiking and trail_visibility=good/excellent are somehow implicit but given the ongoing discussion, should not be treated as implicit and tools should actively complain about them.
BTW: Tag:access=designated - OpenStreetMap Wiki is heavily based on “everything is a path” principle. Even for UK, where I think this is discouraged (?). It would make sense to me to change the paths there to bridleways/footpaths/cycleways with the exception that I do not know what a footpath in the UK means.
I still do not understand if highway=path was intended those 15 years ago to deprecate cycleway/footpath/bridleway and it just never was universally adopted or if it has always been meant to exist alongside those. To me it would make sense to try to define them so there is as little overlap as possible.
2 Likes
Okay, now I understood your point! Thank you for responding! True that, though again: I would assume that there are probably very few (mostly access-oriented?) defaults that one can safely assume from the highwaytag alone.
All the other descriptive tags (=designated, surface=, smoothness=, width=, sac_scale=, mtb*=, etc.) will have to be decided and added on a case-by-case basis.
1 Like
Yeah, I admit that the differences are sometimes subtle, but:
Why should we assume things, if we at the same time acknowledge that those very assumptions are, as you said, “just sometimes wrong”?
A good case in point! I’d argue that the distinction between paved (or hopefully the more specific, e.g. asphalt/paving_stones) and compacted is an important one, and not to be brushed aside as a nebulous assumption on just ‘some kind of surface other than ground’.
Well, we have two contradictory demands:
- data consumers need to decide on something when they see a highway=* tag without any additional tags
- ideally we want all highway=* tags have further tags on them
Possibly these can be at least used as criteria for deciding between path/footway/cycleway. However I think saying “you may assume this but it might be incorrect sometimes to consumers” and “please add additional tags” to mappers can be done at the same time.
1 Like
Yes! Thank you very much for pointing this out! I can certainly empathize with the data consumer’s need, and I do agree that the two demands can be filled at the same time. That is, we needn’t just wait for all the ways to be perfectly tagged with all the descriptive tags before we can consume the data. I mentioned earlier that I joined OSM in 2008, and I fully expect it to take another decade and a half for most of the highways to get the extra descriptive tags. Clearly, the the data consumers can’t just wait this out.
However I do think that there is an interplay at work with these contradictory demands, too. I think we need to also boldly emphasize that the assumptions we derive from incompletely tagged ways are very tentative, and that the answer to the nebulousness of our assumptions and the tentativeness of our inferences, is that we need more of the descriptive tags.
I fear that if we (or the data consumers) only concentrate on finding and documenting the nebulous and fuzzy assumptions, and fail to emphasize the dire need for the extra tags, we’ll just feel complacent with the fuzzy assumptions and never have an impetus to correct the data.
1 Like
Yes, the goal is to document and warn about the wide scope of the tag. But admittedly, this picture triggers lot of discussion, so I think it’s good to seek for another one with less bicycle signs 
2 Likes
From reading along here, the only picture that is unanimously not contested is the centre picture in the collage. The one currently in use in the documentation. To me this says, the picture in the documentation not in need of replacement.
The third one (mine, with an adult and a kid on a mountain path) is not really contested neither.
Designated without path only makes half the sense. It also leans highly to the paved municipal infrastructure path. And that right from the start. In 2010 the traffic signs got added to make that even more final.
Curiously, I learned in this topic, that a number of contributors here rather see path the best fit for something, where it is unclear what the designation. The dirt path in the wiki a very good fit for that 
The first picture is not contested for illustrating that these ways are often tagged as path. The discussion is about whether it should be tagged as highway=path, but I think we have agreement that it is often tagged like that in practice because it’s a JOSM preset.
1 Like
Yes, disagreement is in the same vein as this poll Alpinist routes marked as footpaths - #39 by Hungerburg - now closed for good – 25 no vs 20 yes.
I feel like starting a new poll, with the cycleway/footway/promenade picture and asking, whether that should be tagged path. In the data it is a cycleway. I created the section from a split, with JOSM. I cannot remember if I got a warning.
I also feel like starting a Take 4 topic in this series. A mountain bike trail that I mapped with mtb=designated promptly got retagged bicycle=desginated. I have an idea about designation!
Making it a JOSM preset must have been discussed somewhere…
Please try to keep discussions on teh same subject together in one thread. If the subject changes, that is a reason to start a new one.
An mtb is a bicycle, after all
In my mind, retagging from mtb=designated to bicycle=designated incurs information loss.
3 Likes