Is what you want to express that this network can only change it’s highway at an intersection?
If so, I would disagree with you. The difference between highway=track and highway=service in rural areas is only about whether it’s rather used by normal traffic or agricultural vehicles, so obviously that can change anywhere along the way at the “last” driveway and a little further the same way can narrow further down to a highway-path after the “last” farmland access.
Sometimes people ask an innocent question to a global audience before we discover that reality and norms differ among cultures. We learn something new every day! At that point, we can do two things:
Sharing cases from elsewhere can inform both discussions, especially if we’re open to discussing the merits and drawbacks of our solutions to those cases. On the other hand, airing unrelated grievances and imposing one region’s ill-fitting solution upon another region can be recipes for talking in circles, even if the solution turns out to be right.
Since this thread is in General talk, feel free to explain that highway=pedestrian was coined and continues to be interpreted as fundamentally an indicator of the precise location of Zeichen 242.1, that this interpretation didn’t arrive a decade later, by which point half of the highway=pedestrian ways had already been mapped. Explain how to apply this approach to the countries that lack an analogue to Vienna Convention sign D-04.3,[1] through no fault of the mappers there. Assure us that this adapted solution doesn’t seem completely arbitrary to a layperson.
Yeah no. Maybe let’s agree to just map the signs. 
To be clear, I’m talking about a network in the analytical sense, not a road numbering system that we’d tag as network=*. Yes, in a coherent network, nearby edges will generally connect rather than leaving arbitrary tiny gaps.
Most North American cities developed according to some sort of planned street network. Usually a grid is involved. Pedestrian streets are part of this network. In a coarse sense, pedestrians use the street network. In a more detailed sense, they use the pedestrian network, which includes the streets’ sidewalks as well as the pedestrian streets.
The U.S. community has adopted functional classification for most highway=* classes. Functional classification focuses on a road’s access and mobility, pragmatically balancing these criteria with a desire for continuity. In OSM, we’ve expressed that desire with guidelines for minimizing “motorway islands” and “dead-ending” classifications. (Canadians have expressed similar expectations.)
I’m not exactly saying that highway=* values within the street network can’t change midblock. Rather, they shouldn’t change a mere foot or two from the intersection due to some street furniture that can’t realistically come any closer to the intersection. There’s a word for that: pedantry.
By definition, neither highway=track nor highway=service are components of the public street network.
What about a dead-end street?
Yeah, it’s part of the street network. As I said…
The alternative, if we’re still talking about intersections, is that the tiny gaps isolate every highway=pedestrian into an island in the street network.
If sidewalks are mapped separately, they will connect to the pedestrian street segments without issues, won’t they?
If there is a crossing segment, its presence will correctly represent that there is no contiguous pedestrian network, just like when two sidewalks are separated by a crossing?
I don’t see how any of this is incorrect.
After all, the street network you seem to be referring to is only contiguous for users of motor vehicles. For everyone else, it IS an archipelago.
BTW: I’m not taking a position on the specific crossing that started this thread. It seems to be a grey area.
I’m not talking about routing per se, which is affected by lots of attributes like access restrictions. This is more about the division of the space of a city. A typical city has multiple transportation networks that contribute to the overall routing graph. Think for instance of the trunk network, truck network, all the way down to the lowly residential network, all overlaid on each other.
A street in the sense of the street network is not just the traffic lanes on the roadway but rather the overall thoroughfare. So conceptually a pedestrian can follow a street even if they don’t step foot in it. It’s a different level of detail. Functional classification is easier to grasp when looking at a map than when using turn-by-turn directions.
These concepts are the subject of extensive research that underlies a lot of urban planning and transportation planning around the world. I’m far from an expert in this stuff. Just want to raise awareness that there’s a lot of hidden structure to the world when you zoom out a bit.
The meaning in France is not a pedestrian street, but that pedestrians have to go that way. D-04.3 and D-04.1 are not incompatible.
In fact, that’s what the Vienna convention says: COMPULSORY FOOTPATH (your link).
What you meant is:
FR:B54
In Germany:
DE:237=D-04.1
DE:239=D-04.3
What you meant is:
DE:242.1
With an equivalent for cyclists:
DE:244.1
In Germany, like in some other countries they make no difference by access between footway and pedestrian, contrary to other countries OSM tags for routing/Access restrictions - OpenStreetMap Wiki
these are not similar, one is a zone the other a street
Literally right, but on the German Wikipedia, the example for Dänemark is Cykelgade Zone. I don’t speak Danish, but my educated guess…
Not sure to agree with you. The linear part or the crossing is tagged highway=footway, footway=crossing, so the network is continuous.
BTW priority differs between countries: In France, pedestrians have (yes sometimes it’s more theory than practice) priority, in UK they don’t.
If you believe this is correct, I think someone suggested that pedestrian=crossing would do the same thing. But that means consumers will need to interpret pedestrian=crossing to mean exactly the same as footway=crossing (given that what is on the ground is a marked/signed crossing), which is a bit weird.
There is a similar conflict when a cycleway ends in a pedestrian crossing (absolutely only a pedestrian crossing, no cycle crossing, legally, technically, how ever you like to think about it) and resumes as a cycleway on the other side.
You shouldn’t tag cycleway=crossing because it’s not a cycle crossing. You shouldn’t tag highway=cycleway + footway=crossing, because it could easily be misunderstood as a cycle crossing, or as a crossing that somehow caters to cyclists, which a pedestrian crossing does not – and because there’s no way of knowing how consumers will even interpret the spaghetti tagging.
The cycle network is definitely not contiguous at this point, so why pretend it is? And the pedestrian street also isn’t contiguous if it is interrupted by a foot crossing, so why pretend it is?
I was speaking about the foot network. That the cycle network in your mentioned case is interrupted, with a bicycle=dismount, Do you pretend in this case the cyclists can’t cross as pedestrian? Uncomfortable, with routing penality, OK.
This may be a controversial opinion, but I see no reason for the highway classification to change from cycleway to footway just for a short way through an intersection. highway=* is a macro-mapping tag and on a macro level there is an intersection between a cycleway and a street. On a micro level the litte piece that actually crosses the street is for pedestrians only, but it’s easy for cyclists to become pedestrians.
around here, “highway” is not a macro-mapping tag. If there are a few meters with a different property or a different kind of way, we split the way and check tags accordingly. Yes, you can become e pedestrian when cycling, but you should see in the data that in this spot you actually have to become a pedestrian. If a cycleway is interrupted, I want to see it in the data, regardless how long the interruption is.
I don’t pretend cyclists can’t use the pedestrian crossing as pedestrians. I also don’t pretend that the cycle network is contiguous at these crossings. 
Still, I think there are many examples of highway=* changing value for detailed mapping purposes. Mappers represents steps with highway=steps instead of using highway=footway throughout. They use highway=motorway_link to capture minute details about motorway junctions, like this one.
So when infrastructure changes from being cyclist-oriented to being pedestrian-oriented, I don’t see any issues whatsoever with changing the main tag. We’re just accurately representing the real world, and it’s the real world that’s not making sense in this case.
Although depending on the country it might also be allowed to cross a pedestrian crossing on a bike as long as you can get to it legally (as it is in Germany), in which case the network wouldn’t be interrupted.
What is your reasoning here?
That therefore a ped crossing on a cycleroute should not be interpreted as bicycle=dismount
That’s certainly the case where I live in the US.
