node_network has dedicated connection points, connected with spoke-like routes. Like the numbers below. A base_network would be similar, just without those nodes.
Exactly what I was going to write after reading your list: being networked or not is another dimension of the subject. All the more because there are routes that are half-network and half-linear route, or half-network and half-loop. See for instance the V22 regional biking route in France.
I believe that the 0%-to-100%-networked dimension can be modeled without any new tag or even any new tag value: just put all the main/alternative/approach/node2node parts of the route in a superrelation and let consumers decide how they want to represent it. Adding nodes to the superrelation, as is currently done for node networks, can help.
EDIT: what is funny about V22 mentioned above is that not only is it half networked, it is also national or regional depending on what criteria you use (length and notoriety vs operator) ; even its reference is ambiguous: it is the only regionally-managed route that has a reference with two digits in the French naming scheme.
Adding anonymous nodes to the superrelation would just help consumers decide whether they are facing a network of some sort, or just a continuity error in the route.
If weâre expanding this conversation to include bike routes or other kinds of routes, then this approach might fit some regional route marking practices, but it doesnât work universally. For example, in New York, State Bicycle Route 25A is an alternative route for State Bicycle Route 25, but the two routes are distinct.
In this case, they donât even quite connect to each other. Disconnected alternative routes can also occur when a route is established on one bank of a river, with the trail on the opposite bank as an alternative route. Normally this isnât remarkable, since many bike routes are discontiguous, either temporarily or by design.
In the U.S. Bicycle Route System, many one- and two-digit routes have three-digit auxiliary routes. U.S. Bicycle Route 201 is a 105.1-mile-long (169 km) spur of USBR 1. They do connect, but theyâre essentially related only numerically for planning purposes. Route 201 is national in the sense that itâs part of the national network, though spatially itâs confined to a smaller region of the country, spanning two states (which are regions for the purpose of network=rcn). Have I mentioned how ambiguous the term âregionâ can be? 
âRoles for recreational route relationsâ suggests creating a superrelation to hold the route and any related special routes. Some U.S. mappers occasionally make these superrelations for U.S. Routes (which are highway routes), thinking they might be useful, but we delete them because theyâre basically categories and a bit of a maintenance headache.
Sorry, I was too lazy yesterday to find a hiking route to exemplify my point.
Definitely. If an operator decides that an alternative has its own reference, we can map it as a route of its own. The âstrange beastsâ I was mentioning are collections of routes that their operators define under a single reference. Creating them at separate routes is confusing to end-users and does not respect the designerâs intent; putting them in the same superrelation, even with alternative or approach roles, is confusing to QA tools if we donât help them to find where sub-relations are meant to connect.
Those would be 2 separate routes, alternatives are the same route, but for example you can use left or right side of the river.
Yes, Iâm aware. Some networks simply give you the option, others would pair the route number with a modifier like âAlternateâ, and still others would distinguish â25Aâ as a separate route. We should have tagging solutions for all these cases, not necessarily the same solution for all of them.
I donât get your point here.
base_network and node_network are not the same. In a node network you have usually a node number signposted where in a base network you have general signposts to the train station, next main road, city center,âŠ
In general I would think, if the âidentifier on the signâ is the same, just create a dedicated sub-route for that alternative and put all of the sub-routes into a superrelation. I would do the same for the case, if the identifier got the addition âAlternateâ. At most I would distinguish both cases in the role of the sub-route in the superrelation.
If the identifier differs (like USBR 25 vs. USBR 25A), create a new independent route relation.
I did not say or imply that they were the same. What I implied is that we do not need network:type at all. The difference can be established with tags in the nodes or with roles given to them or with an additional tag in the superrelation if really needed (but I suspect that we can avoid it).
Please donât break established tagging for the node networks. They represent a route planning structure/system that is different from standard routing and different from following long routes. The network:type key was established specifically for this purpose.
1 Like
This suggestion is for a route with one overall name and overall ref, including all side branches, shortcuts, variants etc. Not for the route collection of one provider/operator/administration, which they usually call their ânetworkâ. I would suggest using a network relation for these collections, although in my experience network relations are bags of objects not used by anyone, just requiring maintenance while providing nothing useful. *
- This opinion would change the minute that a data user, any data user, starts to use a network relation to help their end users.
This opinion would change the minute that a data user, any data user, starts to use a network relation to help their end users.
how would that work, what is the information for an end user from a network relation that is not present in the individual routes?
A route that participates in a base_network usually doesnât make much sense on its own. Users will rarely choose a route in a base_network to follow from start to finish. Instead, trips will be planned along parts of the different routes of the network and the routes themselves basically only serve as hints for what the waymarking will be.
A base_network usually only has simple guideposts at the crossroads that will give destinations (same as the destination signs for cars) and the waymarking to be expected along the route. A node_network will have named or numbered intersection markers which point only to the next name or numbered intersection. (And before people try to prove a point: the necessary condition here is the presents of node markers, not the absence of guideposts with conventional directions.)
Some countries have planned their pedestrian or bicycle infrastructure as base networks, which makes them really easily identifiable. Switzerland is the most prominent example. Youâll notice that the network covers every corner of the country not just the parts that are interesting for tourists. The node network in the Netherlands is a similar prototypic example.
In other regions, base networks have usually slowly developed out of a system of routes that became more and more dense. At some point, the routes themselves become less important and people just start planning their own trips using existing routes.
The German black forest region recently realised that their collection of routes is really a network and they are now in the process of switching to the Swiss system.
The Harz mountaineering club also realised that their route collection is a mess and decided to solve it by reorganising it into something that is an even bigger mess and as far away from a linear route as possible. Still, at the end of the day it is a base network.
The Czech republic started out as a collection of routes and they still organise their system with routes and different colors (which sometimes also overlap on the same path). Still, the resulting set of paths is clearly a base network that connects all the country and meant to be used as such.
For the US, Iâve already pointed out the Aldo Leopold Wilderness an example. Just take a random route there: Waymarked Trails - Hiking You clearly wonât go and say âI do the South Fork Mimbres River Trail todayâ. You are more like to use it while planning a trip to McKnight mountain. Hence, base network. Contrast this with the routes along Mount Whistler which are really separate day trips along the Highway. Not a base network.
From a data-user point of view, the distinction between a linear route and a route that is part of a network is really important. For a network route the only interesting information is where the route goes along. It doesnât matter where it starts and where it ends. Youâd want to know the waymarking for a single linear section but if you cannot cartographically follow the route, it doesnât matter. A linear route should be marked in a way that you can clearly see where it start, where it ends and how exactly it goes along in between.
The distinction is also important for route planning. Linear routes are something Iâd want to present to the user as a whole. Network routes only matter as a preference indicator for the router.
1 Like
How would you classify the two examples I gave above (V22 and Sentier des Mines)?
Exactly.
Some mappers have voiced the idea to record information that is currently present in all the routes belonging to a particular network, only in a parent network relation. Such as operator data, or a shield description common to all the routes in this network. Data users could then in theory retrieve this common information from the parent network relation. I donât see this as a viable way forward in the OSM ecosystem. (Every time I say such a thing somebody is bound to prove me wrongâŠ)
1 Like
V22 is a inter-regional route with a lot of alternatives. The alternatives are currently badly mapped in OSM. They need appropriate roles in the main relation.
The Sentier des Mines I remember well having lived right next to it for a couple of years. It was before my OSM mapping time but even then I was wondering what the local hiking club was thinking how this thing is to be used. As a general rule, Iâd consider all the local hiking routes (including Sentier des Mines) in the south-east of France as a base network. If you read guide books from the area, theyâll usual describe a tour as âfollow the yellow markings for a bit, then switch to blueâ etc. Going up Sainte-Victoire, youâd choose between red, yellow and green route like you choose if you want to take Main Street or Redwood Boulevard to get to the supermarket.
I share your perception. But who are we to second-guess the choice of the operators? We have various configurations where the current radical split between networks and linear routes forced us to such devilâs choices. For instance take E2 (West and East). Is it one route or two routes? we force ourselves to make that devilâs choice.
2 Likes
Agree.
If you look at this relation, a lot of sub routes like " Loire Ă VĂ©lo 3e - Liaison Cadre Noir Saumur » have âdescriptivesâ names containing the word âliaisonâ (meaning âlinkâ).
This one shouldnât have the tag name, but the role approach and a tag from=Cadre Noir Saumur.
But what @StC mean is that route form implicitly a real network.
A software like Knooppuntnet route planner could help the cycler/hiker to create a route from A to B inside this ânetworkâ.
For me this is one route.
An hiker could choose to cross Europe by the Est, another by the West, some prefer to follow a loop in this ânetworkâ.