Proposal talk:Bicycle use cycleway

Direction dependent

Only cyclist traffic "passing" a traffic sign (like ) can have the "obligation to use a cycleway". The signs can sometimes exist only in one direction on some sections between intersesctions. Alv (talk) 12:39, 18 September 2013 (UTC)

If the back-direction has no sign, cycling is not allowed (on left cycleways, in Germany) or only allowed (not compulsory, on right cycleways). So we can/should use oneway=* for cycleway and/or bicycle:forward/:backward=use_cycleway at the road. --MasiMaster (talk) 01:25, 9 November 2013 (UTC)

Alv's statement is (IMHO) not true where I live and map; a compulsory cycleway here does not stop being a compulsory cycleway if you join it at a point where you do not pass a sign. Even if you (might) get away with claiming that you didn't see a sign (sign not placed, placed incorrectly, vandalised, stolen as a trophy, etc.), you are expected to drive as much on the right as possible (and a cycleway alongside a road is normally considered to be part of the road). Perhaps this is different in different jurisdictions. When mapping in OSM, you can be expected to do due diligence to determine the status of that way; if it is a cycleway at the beginning and at the next junction, it is almost certainly a cycleway in between too if the infrastructure is the same, even if there is a connection for bicycles in the middle without an appropriate sign. (Here, they are generally well-signed.)Frankl2009 (talk) 14:29, 5 December 2013 (UTC)

+1 for "stolen as a trophy" :-)

But, seriously, in which way is this important for tagging or routing? All that has to be added are the correct oneway tags on the cycleways where applicable, right? As far as the router is concerned, it will send you on the cycleway either way. --Emkey08 (talk) 16:28, 5 December 2013 (UTC)

I failed to spell it out that it is not necessarily a "property of the way", but can depend on the driving direction and/or where you're coming from. Alv (talk) 00:06, 8 December 2013 (UTC)

I am not sure I understand this. Are you saying there are situations where in one direction there is no restriction on cyclists using the road, but in the other direction cyclists must use a separate cycleway beside the road?Frankl2009 (talk) 23:28, 8 December 2013 (UTC)

Yes. Like here, in this direction there are no signs so a cyclist coming from the right and turning right, away from the camera may cycle on the road, but some 130 meters later there's a compulsory cycleway sign in the other direction, so a cyclist coming towards the camera must use the cycleway. In fact, this looks like an example of the other point below: anyone coming from the Kuutamotie between these points does not pass a sign at first, so they don't have the obligation to use the cycleway, until they do. Alv (talk) 21:23, 12 December 2013 (UTC)

Here's an example in NL. This is the reason we've added bicycle:backward=use_cycleway etc. in the examples but I must admit this is kind of complicated.( PeeWee32 (talk) 17:33, 13 December 2013 (UTC)

Thanks for the examples. Yes, that makes it clear that more complex tagging is needed in such situations. In the Dutch example (PeeWee32) the red coloured road surfacing (i.e. not the separate bicycle path) is what they call a "suggested lane", which has no legal status. So there is no restriction of any kind on bicycles in the direction of that lane (for the other direction, bicycles must use the cycleway).Such lanes are currently being tagged in the Netherlands as cycleway=shared_lane (and in this case, it would have to be marked as cycleway:left/right:shared_lane, depending on the drawing direction of the road). They are often mistagged as cycleway=lane. Frankl2009 (talk) 20:05, 13 December 2013 (UTC)

OT: In Germany we have also "suggested lane" (called "Schutzstreifen, and not colored). I wonder why shared_lane is used for this? I think it is unsuitably to speak about a lane, and also about shared in this caes. Motorvehicles may use the shared_lane only if it is necessary and put no cyclist in danger. Key:cycleway say: "Cyclists share a lane with motor vehicles, and there are markings indicating that motorists and cyclists should share this lane. The road markings are usually there to highlight a cycle route or to remind drivers that you can cycle there. Also used for the on-road shared-lane marking called a 'sharrow'".

A shared lane/sharrow is a "full" lane, which can also used by cyclists. I see a bigger difference between suggested lane and sharrow than suggested lane and a official cycleway=lane. One big difference for overtaking is: parallel vs. use the reverse-lane. Either we can use a new tag for suggested lanes or we can use the cycleway=lane with suitable access tags. --MasiMaster (talk) 17:34, 14 December 2013 (UTC)

Situation in Poland

In Poland use of cycleways or dedicated lanes is mandatory, assuming that designated cycling route leads in intended direction. ("Kierujący rowerem jest obowiązany korzystać z drogi dla rowerów lub pasa ruchu dla rowerów, jeśli są one wyznaczone dla kierunku, w którym się porusza lub zamierza skręcić."). It is unclear whatever it includes mixed, unsegregated cycleways (marked with sign ) Bulwersator (talk) 09:37, 17 November 2013 (UTC)

Why this tag makes sense (quote from mailing list)

from https://lists.openstreetmap.org/pipermail/tagging/2013-November/015606.html

Instead of this bicycle=use_cycleway we could use bicycle=no an additionally a “my special vehicle/situation=yes”. There are a few reasons why I do not think this is a good idea.

1 Mapper are no legal experts

In OSM we rely on mappers and not legal experts. An ordinary mapper in NL (where I live) does not know what the legal status is of many extraordinary vehicles and there are many. I have a 3 wheel velomobile with such measurements that I am allowed to ride these type 2 roads (but not the typ1). Hardly anyone I speak knows this. I’m also sure they don’t know the legal status of horse carriages, skateboards, sedgeways etc. The problem is that there are no traffic signs for all these exceptional vehicles so how should a mapper know?

2 too many tags

Imagine that all special vehicles and situations would be tagged. How would we see all these tags in the editor. I’m afraid it would be a complete mess in OSM. Simply to many tags so we loose overview which might scare mappers away.

3 Changing law needs changing tags

Imagine that in NL law would change in such a way that groups of race-cyclist of more then 10 are allowed to use the type2 way. Then this would have to be mapped. Who is going to do this? If we would have the bicycle=use_cycleway nothing had to be changed. I think we have to be carefull with mapping legal access in OSM unless the traffic signs are obvious. In fact I think that if we map in such a way that we (and routers) know what traffic sings are present, routing for any vehicle should be fairly easy.

4 country specific

All the exceptional vehicles and situations vary from country to country. Imagine tags like “three wheel bicycle wider then .75m=yes” in NL and “three wheel bicycle with combuston engine > 250Watt=yes” in an other coutry. This is just going to be too much for most mappers. I would not start mapping these exceptions abroad because I just don’t know all legal aspects.

In short: Mapping this way will never happen in such an extend that it will improve routing for bicycles (both ordinary and exceptional ones) .

So all this made me feel it was an illusion to improve bicycle routing by adding different tags for all these exceptional vehicles/situations. This could work in theory but it simply will never work in practice

So, it had to be as simple a possible. Something any mapper could see in reality based on traffic signs and roads and cycleways. That’s the reason why we have added a definition of the tag.

end of quote

Bulwersator (talk) 09:44, 17 November 2013 (UTC)

Why this tag does not help to improve routing

We're facing a routing problem here, not a tagging problem. As such, it has to be solved within the routing software and not within the database. Rationale given below. --Emkey08 (talk) 16:24, 3 December 2013 (UTC)

Are you sure? First, it is a tagging problem. Why we don't should tag access-restrictions? If you say we don't need this tag, because it will not improve routing, I will ask, why we need to tag the "motor_vehicle=destination"-signs in residential 30-zone OverpassTurbo link? The router should route over the mainroads in any case. I think we should tag access-rights! Second, it is also a routing problem. Fastest and shortest route will not work with no bicycle-tag on the road. --MasiMaster (talk) 23:49, 4 December 2013 (UTC)

Well. I already said that I'm not against this tag if it helps mappers to improve data quality (and it probably will). And yes, we should tag access rights, but at the same time we should only introduce new tags if they are necessary. As far as routing is concerned, the tag is actually not essentially necessary. Did you read and understand the rationale given below? If so, please don't just say "it will not work" just because you think so, but give an technical argument instead like Alv did. --Emkey08 (talk) 10:00, 5 December 2013 (UTC)

So if the tags helps improve data quality and access tags are necessary, already this are 2 reasons the tag is necessary. Ok, i thought my routing example was clear... I will explain it a bit more precise: If you need the shortest way, you have to give all highways the same factor (only length is important). So often the road is the shortest way and the router will chose this if it is untagged. Problem: Routing over a "forbidden" road. Other example: If you need the fastest route, the router have to give different speeds for some ways. I.e. you are not allowed to cycle more than 15, 20 or 25 km/h on a cycleway. On roads you can ride faster, there, you have to wait only once to turn left, not twice or more (like on the cycleway). Cycleway trafficlights set earlier to red. So the router will use the road for fastest route if it is untagged. So we need to tag the road, and bicycle=no is simply wrong.

This was for explain. But I think we should not talk (here) about the routing problem and how routing could be improved (agree this is important, but not here), instead if we need this tag / is there are a better value- or tag-name or something like that. --MasiMaster (talk) 18:59, 5 December 2013 (UTC)

Ok, thanks. Right, fastest and shortest route in the strict sense would be problematic without the tag (though this distinction probably isn't very important for cyclists, as they usually don't exceed speed limits).

I'm in favor of the bicycle=restricted tag proposed by Rjw62 below, as this would be a general enhancement for access rights tagging, and not only suitable for bicycles. I'd therefore say, reject this proposal, and propose the introduction of "restricted" as an access right value instead (this should probably be done anyway, since already used a lot in practice). --Emkey08 (talk) 09:29, 6 December 2013 (UTC)

Case 1: routing of "ordinary" bicycles (that is, bicycles which in general will / have to use compulsory cycleways)

Routing software will have to interpret a bicycle=use_cycleway tag as an advice to not use the road (and thus, as an advice to use the cycleway). This is because interpreting a bicycle=use_cycleway tag as bicycle=no (for ordinary bicycles) will result in even worse bicycle routing, as cyclists often have to and / or are allowed to use the main road for short sections even if there is a compulsory cycleway present, e.g. for making a turn. Since the tag can thus only be an advice, the routing software will implement this by giving the road a higher weight, resulting in a preference for the cycleway. --Emkey08 (talk) 16:24, 3 December 2013 (UTC)

There is no provision in Dutch traffic regulations for the exceptions to mandatory use of a compulsory cycleway (or mixed moped/cycleway) which are mentioned above (e.g. turns); so a router on Dutch roads should treat the proposed tag on an adjacent road as a bicycle=no. However, there are specific categories of bicycle which are allowed to use the roads as well as (or instead of) the compulsory cycleways and it should be possible to route for those cases over the road, if desired. Weighting cannot do this and setting a router to 'avoid cycleways' or similar won't work either (since you might want to use the optional cycleways). Frankl2009 (talk) 16:53, 3 December 2013 (UTC)

From a legal point of view, you're of course right. From an algorithmic point of view, a router will be very vigilant against interpreting bicycle=use_cycleway as bicycle=no for ordinary bicycles, since the routing would then heavily depend on very accurate and complete mapping of all road intersections which involve compulsary cycleways (e.g., think about service and residential roads which interconnect with the main road on the opposite side of the cycleway - these then all need to be also connected to the cycleway in order to enable routing as users expect it). So even in countries where bicycle=use_cycleway could be interpreted as bicycle=no from a legal point of view, it won't be done because in practice an weighted algorithm will generally obtain better results due to its ability to tolerate mapping errors to some degree. Furthermore, you'd generally want to implement an algorithm which is universally usable, and not just in some countries. --Emkey08 (talk) 18:01, 3 December 2013 (UTC)

Yes, mapping will have to be accurate, and yes, it is easy to overlook these connections. Fortunately, the infrastructure (kerbs in particular) usually makes it clear (in The Netherlands) where a connection for bicycles (and mopeds) needs to be made opposite side roads, driveways, to ensure correct routing. So it is up to all us mappers in country to work on this (and we do). As for universal applicability: yes, that's why the proposal is being made. It's basically a hint for a router or renderer to ensure that it works properly everywhere. Frankl2009 (talk) 10:10, 4 December 2013 (UTC)

The relevant NL law snippet, poorly translated as "Cyclists use the compulsory bicycle or bicycle / moped path. They use the road if a mandatory bicycle or bicycle / moped path is missing." doesn't really make it unambiguous when a bicycle path is or isn't missing. If you're turning left, and the bicycle path does not turn left, there is no (mandatory) bicycle path turning left, i.e. it is missing. Only previous court cases and other legal sources can make it clearer. One can not ever read the law by looking at just one sentence.

Also, it didn't take long to find a streetview example of another case with an exception that's not directly allowed in the law, but must be: here the two way cycle path on the left has mandatory path signs at both ends, but a cyclist coming from the right - from a signposted mandatory path - has no way of knowing that. For them, there is no mandatory path visible, so if they turn right towards the direction the camera is looking in the link, they should be able to safely assume they can drive on the road until they see the next cycle path traffic sign. Alv (talk) 07:20, 5 December 2013 (UTC)

This is certainly not the way I would view this Streetview example, as someone who lives and is used to cycling in the Netherlands. I would take the cycleway, which in my opinion is clearly an example of a compulsory cycleway (the optional cycleways are almost never adjacent to the road). The cycle node route boards and the cycle nodes information board also suggest it is a cycleway. Although the signs are important, they are not the only consideration in following the traffic regulations. A police car coming up behind you would probably stop you and direct you to the cycleway. Frankl2009 (talk) 14:46, 5 December 2013 (UTC)

We're starting to miss the point. This should be about how the bicycle=use_cycleway tag relates to routing, and about the argument that an weighted algorithm will in general score better results in regard to fault tolerance than an algorithm which just interprets bicycle=use_cycleway as bicycle=no. A (well written) router won't rely on the assumption that everything is 100 % correctly mapped if you have an alternative that is fault tolerant. --Emkey08 (talk) 12:32, 5 December 2013 (UTC)

Thanks for this. I am saying that a router should not route any ordinary cyclist in the Netherlands over the road in the example given. I believe that a hint will be required to ensure this. That issue is now covered below. Frankl2009 (talk) 14:46, 5 December 2013 (UTC)

@ Alv. This is a very nice example indeed. Practically in my bakcyard so I should comment on it ;-) ((coincidence? Nehhh can hardly be)

The cycleway is indeed a compulsory cycleway. On the steetview example there is some reconstruction going on so I don’t know it the cycleway sign is still missing but for the sake of this example let’s just assume it is missing. Believe me that it is a compulsory cycleway ( google proof) and as far as I know there’s no “bicycle forbidden” sign on this road. I’ve been there many many times and as Frankl2009 has pointed out you are not supposed to ride your bike on the main road. I would certainly get a ticket if a policeman would see my riding my bike there. Then why did I never do this? To be honest .. it never crossed my mind. As Frankl2009 pointed out, for a Dutchman it is so obvious that you have to use the cycleway that it just does not cross your mind. But … if you are a foreigner… it’s a different story. In that sense it is very useful if we would indicate that this type of road is a road next to a compulsory cycleway. Of course I could ride my bike and have it go to court if I would get a fine. You never know what the judge is going to decide ;-). But if you want to avoid all this hassle just look at your smartphone /GPS unit with OSM data. It could show this is a road where you’re not supposed to ride your bike indicated by the compulsory cycleway, the “bicycle=use_cycleway” on the main road or both. (instead of the “wrongly” tagged// bicycle=no) For a Dutch mapper there will not be much discussion about whether or not you are supposed to ride your bike there. --[User:PeeWee32|PeeWee32]] (talk) 18:32, 7 December 2013 (UTC)

Anyway, this was not the point, but the fact that the "must not drive on the carriageway" can depend on where the cyclist comes from - if you don't pass a sign, you follow the "normal" rules", and you don't have the obligations to use the cycleway. (Really, I just picked a spot "near Amsterdam" and went looking for a cycle path crossing a road that has cycle tracks but no "regular" intersection there. But I guess NL is "small" relative to the mapper density, so somebody might/is bound to live nearby.) Drivers only need to obey signs they pass, otherwise they follow the "normal" traffic rules. Anything else will be unenforceable - it might take a round of court appeals - unless "they" can prove it's the "way of the land" i.e. virtually everybody knows it. It might well be that in NL everybody is thought to look for the random unofficial route signs and would consider that a cycle path (e.g. here any "path" next to a road is a footway/sidewalk just for pedestrians, unless you see a sign, or have seen a sign)- but not compulsory because they don't pass a "blue sign". The police might well instruct you to take the cycleway - I do hope your's are kind and eager to start by giving guidance vs. comply-or-baton. The traffic environment is hardly ever completely signposted. When the traffic laws were formulated, they never thought the roads would need to be unambiguous to "stupid" computer code - even if that were their ultimate goal - all drivers with a license are taught the basics to come along even when the signs have small ambiguities, even when they have to give up some of their "rights". Alv (talk) 23:59, 7 December 2013 (UTC)

Ok point taken . And also I believe you that the location was just a coincidence . It is handy though because today I went there and took a Picture. As you can see there is a sign (compulsory cycleway) but I can not guarantee that in similar situations there will always be one in NL. The situation in the google link does not exist anymore. Here’s where the picture was taken. The (blocked) cycleway on the west side of the road (google view) is now a footway and a new cycleway a little north was added (see OSM). I know for sure that there is not always a cycleway sign on the other side of the road in Germany. I remember looking for a cycleway but thats not always easy to see if your are in a low velomobile and signs are missing ;-) Anyway.. your point is still clear. All I can say is that Dutch traffic law was simplified in 1990 using a motto. This was “Geef je verstand eens voorrang” which freely translates as “Give priority to your common sence “ . I’m no legal expert but I am sure a policeman could give you a fine if you would ride your bike on the main road. If you would then bring it to court I can imagine a judge saying “you should have used your common sence”. [User:PeeWee32|PeeWee32]] (talk) 11:43, 8 December 2013 (UTC)

I think it is not only a cycleway problem. This problem could only be solved if we tag the position of signs and the direction from where they are visible. This is not practicable. I prefer to tag this shown cycleway as anybody other. If you get to this intersection, you can trust OSM data (which I not prefer) or trust the reality.

It is a problem of the government. They miss to add signs there. BTW: the government also miss to add the right sign to this cycleway. I can't see a parallel road, so the compulsory cycleway sign is wrong and should be replaced by a non-compulsory sign. --MasiMaster (talk) 14:33, 8 December 2013 (UTC)

Funny that you say that "it's not practical" because the missing signs and inconsistencies are exactly the reason why we started mapping traffic signs in Finland (back then only limited set of signs but nowadays we're actually trying to collect them all). With the experience so far, I'd claim it's well possible, and more importantly, useful to map the traffic signs. --Ij (talk) 20:07, 8 December 2013 (UTC)

The compulsory cycleway sign is sometimes used in the Netherlands when there is no adjacent road to indicate that certain categories of moped are allowed. That's because there are three categories of cycleway and the third type, the 'optional cyclepath', does not allow the lowest category of moped; so the path you cite would allow those mopeds (those unable to travel faster than 25 km / hour) whereas the 'optional cyclepath' would not allow them (if they had an internal combustion engine and it was in use; it would allow them if you could switch off the engine and pedal instead; on these optional cyclepaths scooters are therefore not allowed unless electro-motor driven). Frankl2009 (talk) 00:12, 9 December 2013 (UTC)

Well, Alv repeats his statement that "if you don't pass a sign, you follow the 'normal rules', and you don't have the obligations to use the cycleway". Two things here: (1) The "normal" rule in the Netherlands is, as a cyclist (two-wheel, no exceptional size) you do not cycle on the road if there is a compulsory cyclepath or a moped/cycle path (note that there is a third category of cyclepath, an optional cyclepath, which is not dealt with here). So the only question is whether or not you could be expected to see it as a compulsory cyclepath, a moped/cyclepath, an optional cyclepath, a footpath, or a service/parallel road. I already said (and PeeWee32 confimed this) that I believe most people cycling (and mapping) where we live and work would see it as a compulsory cyclepath or a moped/bicycle path (same rules regarding not using the road). Note that IMO (but IANAL) you do not have to pass a sign to make it legally a cyclepath; it only might mitigate any legal action taken against you if you ignore it and cycle on the road (the "I didn't know" defence, which however is usually not very effective). (2) What tells us whether it is a cyclepath or not is the sign at the beginning of the path and at the next road junction with the cyclepath. If a mapper has done 'due diligence' and checked s/he will know whether there is a sign at either end of the path. Frankl2009 (talk) 23:54, 8 December 2013 (UTC)

I don't know about NL but the problem here in Finland is that you could just as well get fined from driving on a sidewalk that is by default foot-only (of course in practice you only can get into trouble if you hit someone but that's besides the point). Most of our cycleways are segregated=no type and those are without any markings because they lack the solid line which reveals all segregated=yes cycleways (usually the width reveals that it's intented to be a cycleway but there are certainly some exceptions to this rule). Obviously the ambiguity should be solved through proper maintainance by placing the appropriate sign which most likely has just gone missing since the installation but in practice there will always be significant latency before the lack of sign is discovered and reinstallation takes place. --Ij (talk) 13:30, 9 December 2013 (UTC)

Yes, same in NL: no cycling on a footpath; but if it is not signed in any way, it is probably not a footpath (and I would tag it as a path and not a cycleway or footway. But the specific point under discussion was the example which was not signed at a left (or right)joining junction. My point is that (a) this does not require a new sign since the path continues and (b) a mapper can be expected to show 'due diligence' and find out whether the way has any signs at the beginning/end (next full junction) and tag accordingly. Again, IANAL, but I very much doubt that the simple lack of a sign at this joining junction changes the status of the way from cycleway to whatever else it might be. There are other situations (now mentioned above) where there ismore uncertainty or where additional tagging is necessary. This is not one of them, IMHO. Frankl2009 (talk) 20:18, 13 December 2013 (UTC)

Thanks for your view on this matter. I've seen you've changed the wiki but as far as I know this is not something to be done during voting.

I think that without a bicycle=use_cycleway it's going to be very difficult for routers to propose a good route for ordinary bicycles in e.g. Germany. The legal differences between the counties can be considdered by a outer. A route for ordinary bikes in NL can be made assuming a use_cycleway is equal to bicycle=no. In DE a router could handle this in an other way.

I understand you doubt the practical use of this tag but please also considder that it is not interfering any other tagging so it will at least not harm OSM.

Please also bear in mind that this tag is not only for routing. If I want to make a map (online/offline) showing where an ordinary bike in NL is allowed to ride it is impossible for me to do. (not considdering that fact that most of these ways are "wrongly" tagged with bicycle=no in NL) For example I can not use an overpass query to extract these roads as long as the use_cycleway tag is not present. In NL many connections on cycleways have been added to OSM. You'd be surprised how detailed this has been mapped.PeeWee32 (talk) 18:58, 3 December 2013 (UTC)

Sorry about the edit on the main page, I just wanted to make sure this issue is sufficiently discussed. Please just undo the edit there if you think it shouldn't be there.

I agree that the tag will not do any harm to OSM. Moreover, if this tag can convince mappers to replace incorrectly used bicycle=no tags with bicycle=use_cycleway tags, then it will actually help to improve the data quality. What I'm trying to say, however, is that the bicycle=use_cycleway tags themselves won't help to improve routing (as reasoned here).

I think you need to be clearer about the exact purpose of the tag. If its purpose is to replace and / or prevent incorrectly used bicycle=no tags, then I'm convinced that it is a good thing. This however unfortunately doesn't really come out from the proposal page IMHO. Maybe this also is the reason why several people have voted with "no". --Emkey08 (talk) 21:13, 3 December 2013 (UTC)

For the record, I am not convinced that the bicycle=no tag is being incorrectly used in the Netherlands (for the legal reasons mentioned above), but it certainly conflicts with tagging practice in Germany, and there are different opinions on the legal question in the Netherlands too. So it does seem to make sense to add this tag for the sake of clarity.Frankl2009 (talk) 10:10, 4 December 2013 (UTC)

For the record, I'm a mapper who has used the bicycle=no many many times (in NL) when there was no "bicycle forbidden" sign. As you see I'm converted now ;-) . Main reasons: this conflicts with many other countries and it has never been proposed and accepted. It just happened because in NL for an ordinary bike it is (almost) the same as a bicyle=no and we never considdered voting for it. I whish someone would propose the bicycle=no on these roads and have it discussed on the tagging list. This way we could get it over with but as you know I'm against and will not propose. But if you (or anyone else) want to propose I would encourage that. I would vote no but it would help the discussion. PeeWee32 (talk) 16:05, 7 December 2013 (UTC)

@ Emkey08. I’ve send a question to the tagging-list and the answer was that it's OK to change as long as it does not change the intention the proposal. Besides it adds to the discussion and the wiki of the proposal needs changing anyway so I won’t bother. On second thought I think I have to thank you because this new page is very useful.PeeWee32 (talk) 07:45, 7 December 2013 (UTC)

This exact thing however can also be done without any bicycle=use_cycleway tags. The routing algorithm must simply implement a general preference for cycleways, as the router should actually always prefer cycleways when routing ordinary bicycles. If you see a route switching from the cycleway to the main road in order to save a few meters, then this simply is because of a poor routing algorithm, and not because of missing bicycle=use_cycleway tags. --Emkey08 (talk) 16:24, 3 December 2013 (UTC)

Perhaps we have different ideas about the location of these compulsory cycleways. They are not all right next to the road so the difference can be more than "a few metres". I agree that algorithms for bicycles need to prefer cycleways; but, at least in our jurisdiction, they must not route over roads when there is a compulsory cycleway nearby; so a hint to get that right would help. I can understand that this seems irrelevant in jurisdictions where compulsory cycleways do not exist or where the infrastructure won't usually cause routing problems if a little weighting is applied. Frankl2009 (talk) 10:10, 4 December 2013 (UTC)

I believe one of the problems without the hints is that the weighting / cost function can not be a constant, because in intersections, the "legal" route for bicycles is easily much longer than the route where the cyclists takes the main carriageway beforehand, even if the actual difference in length is some 10 to 25 meters. Like in the example pic i made up (easier than searching for real world examples, but the idea should be understandable), the cost of taking the road would have to more than 1.25 times that of a cycleway. In the rural areas, where intersections are farther away from each other, that ratio would send a cyclist down a 10 km route along cycleways instead of the legal 8 km long road. It's hard to imagine a case where a cycleway "next to" a road would be that much longer when going the same way - even if they usually have more bends than the main road. Currently known-to-be-used algorithms "work" on the edge level, so it's not even possible to amend them to scale the constants based on length: there's bound to be more than one "graph edge" between the points where the scaling would have to be considered along several edges. Alv (talk) 15:25, 4 December 2013 (UTC)

Good point. You're right, using a constant weight factor alone would not be sufficient. It would also have to include something like this: when switching from a road to a cycleway, use a lower factor for the first X meters of the cycleway. When switching from a cycleway to the road, use a higher factor for the first X meters of the road. Thus, bounty switching to a cycleway and penalty switching to the road. This will prefer short cycleways even if they are much longer in relation to the road (but not in absolute distance). Best values for X and the weighting factors would of course have to be determined in practical tests. When using adequate values, the router will be able to get it right in almost all situations. --Emkey08 (talk) 20:32, 4 December 2013 (UTC)

Very good point indeed. It was not very hard to find a real world examples. Here's an example where the router takes the shortest route and has no way of knowing it should not take these roads (because there is no bicycle access tagged).--PeeWee32 (talk) 15:58, 13 December 2013 (UTC)

Wrong. Correct routing actually does NOT depend on any bicycle access tags here. The problem simply is that the routing algorithm is poor. Improve the routing algorithm like described above and it will be correct without any additional tags. --Emkey08 (talk) 13:39, 15 December 2013 (UTC)

If you need the shortest route, you have to give the same factor to all ways! But if you like the "shortest, but prefer cycleway"-route, then you are able to increase the weight-factors of cycleways. In first case, and in fastest route the prefer of cycleways are not allowed. --MasiMaster (talk) 01:19, 17 December 2013 (UTC)

Case 2: routing of "special" bicycles (that is, bicycles which don't have to use any cycleways)

We're seeing many exceptions for compulsory cycle ways, depending on local law. First, note that it is de facto impossible, but at least impractical, to map them all. We won't map the fact that you're allowed to cycle on the road in a group of >= 15 cyclists in Germany, with a racing bicycle which has a rim width of <= 23 mm in Austria, if the cycle way is on the left side and you ride on the main road for a short distance in Finland, with a tricycle in the Netherlands, and ...

I simply disagree that it is 'de facto impossible' and I doubt that it is 'impractical' to map jurisdictional differences. We are surely doing this all the time. In the legal systems, they have already worked out how to make the distinctions; what we need to do in OSM is find ways of translating that to tagging which will work. This does require routing software to respect national access rules.Frankl2009 (talk) 11:00, 4 December 2013 (UTC)

The point is, the user has to tell the routing software that he/she is using a special bicycle which is not obligated to use compulsory cycleways. If the router knows this, it can simply apply a user preference for determining road and cycleway routing weights (a route for a special bicycle may either prefer the road or the cycleway - it's the user's choice).

Note that we also don't need any bicycle=use_cycleway tags here, as they don't apply for special bicycles anyway. The only precondition here is that the bicycle=no tag must not be present on the road. Since special bicycles are allowed, it would be incorrect to have them anyway. --Emkey08 (talk) 16:24, 3 December 2013 (UTC)

The assumption is that because a very small number of 'special bicycles' are allowed on a road, it is incorrect to tag that road as being unavailable for the vast majority of 'standard' bicycles (because they need to use the adjacent compulsory cycleway). People who have 'special' bicycles usually know what they can and cannot do. Frankl2009 (talk) 11:00, 4 December 2013 (UTC)

Exactly. Since those people know, all they have to do is to tell the router that they use a special bike, and all the router has to do is to ignore any bicycle=use_cycleway tags. Thus, the tag is not required for routing special bikes, which is what I wanted to show here in case 2. As far as the bicycle=no tag is concerned, this is not a question of majority or minority. If there are bicycles which are allowed, bicycle=no (alone) simply is incorrect. Yes, it would be a good thing to replace them with bicycle=use_cycleway. But this is just because the bicycle=no tags will be removed then, and not because the router depends on the bicycle=use_cycleway tags. --Emkey08 (talk) 10:24, 5 December 2013 (UTC)

I agree. Basically the proposal uses a tag to introduce a third roadtype. Currently we have 'no restriction' and 'explicit resctriction'. The tag introduces 'implied restriction'. Any of these can be linked to groups of vehicles. The types 'explicit' and 'implied' link to different groups. The use of bicycle=no introduces the need to add a second tag to that road to omit special vehicles from this key. In that sense the proposal is good: it reduces the number of tags in OSM, while maintaining the difference between explicit and implied. It does not interfere with other tagging. But it does not solve routing problems, it just introduces other problems: I have yet to see a router where I can set my vehicle to 'cargo bike' or '10+ cyclists'. So while it arguably solves the routing for a specific group of cyclists it is not a generic solution for all road users. Such a generic solution can only be found by adapting the routing software. In a more strict sense: adapting OSM to accommodate less ideal routing software is simply wrong. Which makes it a difficult proposal to vote for. I approve the tag because it gets rid of bicycle=no where needed, but oppose it for the reasoning behind it Noordfiets (talk) 10:41, 7 December 2013 (UTC)

This proposal was/is meant to

a) to add extra access information for cyclist to OSM

b) as a consequence of this extra info to improve routing

We now have NO-voters that emphasize on the routing for the next reasons:

1 It should not be tagged because it has routing in mind (Noordfiets)

2 It should not be tagged because it will not improve routing (Simon Poole, Emkey08, Giardia, Madeco)

I really think this is a shame because it seems to me that the these voters forget about the prime goal and that is adding access information to OSM. Even if you think routers can not use this information it should not be a problem because a) is still valid. It won’t harm routers either.

I am hesitant to address the routing benefits because it will lead away from the main goal but I will give it a try.

In NL the use_cycleway roads should not be accessed by ordinary bikes. For ordinary bikes it is practically the same as a bicycle=no. The router for ordinary bikes can use this info and as a consequence exclude these roads from routing. As a consequence you will never be routed over these ways. Openfietsmap (garmin based OSM map) has already said to consider a bicycle=use_cycleway the same as a bicycle=no for its routing. Still not convinced ? Please consider a) and forget about routing ;-) PeeWee32 (talk) 13:49, 7 December 2013 (UTC)

For the record, I'm (also) voting against this proposal because I think we should use bicycle=restricted instead (i.e., I think we should introduce a new general access value category instead of just having a special value for bicycles only). --Emkey08 (talk) 12:38, 8 December 2013 (UTC)

OK. Are you then also saying we should retag all the "vehicle"= destination like e.g. motorcar=destination? If so, do you think that is realistic? --PeeWee32 (talk) 15:30, 8 December 2013 (UTC)

No, I don't say it should be retagged, since vehicle=destination doesn't conflict with vehicle=restricted. One could use access=destination for a "destination" access restriction, access=restricted for a generic access restriction (due to whatever reason, depending on applicable law), and access=no to generally disallow access. Since bicycle=restricted isn't used so far, it can be interpreted as "restricted access because there is a compulsory cycleway" (but, admittedly, I'm not sure whether there could be any other "restricted" bicycle access situations except compulsory cycleways). If bicycle=restricted can be used, it would be coherent with the currently used access=restricted tags, which is what I'd prefer. In either case, I'd want to have something different than bicycle=use_cycleway since this would be too specific IMO. --Emkey08 (talk) 18:04, 8 December 2013 (UTC)

I'm not in favor for bicycle=restricted, because it could be used for other cases. But what we need is a own tag to show the specific right. We have the same thing with footways. So if there is an footway, in you have to use it instead the road. foot=use_cycleway don't work, foot=use_footway can work, but what about bicycle/foot=use_sideway, use_sidepath or something? --MasiMaster (talk) 21:14, 8 December 2013 (UTC)

@ Emkey08. If I understand you correctly you just want the bicycle=restricted and no additional k/v (like bicycle:restriction=use_cycleway) for situations where there's a compulsary cycleway. One could also considder a "destination" as some sort of restriction and as you say there might be other restrictions for bicycle (like) vehicles. You say use_cycleway is too specific but I think "restriction" is too vague.--PeeWee32 (talk) 20:40, 9 December 2013 (UTC)

Conclusion

The demand for the bicycle=use_cycleway tag arises from the wish to improve bicycle routing algorithms. However, the approach to solve the problem is wrong. There is nothing which has to be fixed within the OSM database. It's the routing algorithms which have to be fixed and improved instead. --Emkey08 (talk) 16:24, 3 December 2013 (UTC)

In Germany it's often better to avoid compulsary cycleways (for safety) as well as their normal roads (for legal reasons). How would you do that without bicycle=use_cycleway? --rayquaza (talk) 20:57, 3 December 2013 (UTC)

Note that the bicycle=use_cycleway tag alone won't be sufficient for this, as the cycleways themselves also need to be tagged as compulsory then (otherwise the router can indeed avoid the road, but not the cycleway). But I'm not sure if we really want bicycle routing algorithms to avoid cycleways - sounds a little strange, no?

I agree the tag can help to improve the data quality, if used as a replacement for incorrectly used bicycle=no tags (also see my other comment above). But bicycle=use_cycleway is not necessary to improve routing. --Emkey08 (talk) 21:40, 3 December 2013 (UTC)

Alternative Suggestion: use bicycle=restricted + a second tag to describe the restriction

My problem with the proposal is that the proposed access tag is rather nebulous in meaning and has different interpretations in different places. It's also likely to get misused by mappers who want to make routers prefer neighbouring cycle paths over the road for reasons other than legal restrictions. I think we should avoid having a top-level access tag that attempts to capture such restrictions, without a way to record what the precise restrictions are. I would therefore like to suggest an alternative, which makes use of the existing access=restricted tag (currently the 11th most popular value of access=*).

While not documented in the wiki, this tag presumably expresses the fact that use of a way by a particular mode of transport is usually not allowed but there may be some exceptions. This seems to fit the bill perfectly here. You can't normally cycle on the road, but you may be able to in certain circumstances. This is all that most general-purpose routers will want to know -- they'll avoid routing bikes along the road where possible, and if they need to produce a route including some of the road, they can warn users that use of the road is 'restricted' and they may need to dismount and/or seek an alternative route. So my suggestion is to use bicycle=restricted on the roads in question.

The missing piece is then how to document the precise type of restriction that applies. For this, I think we need to use a separate tag. I'd suggest something like bicycle:restriction=* where the values come from a finite set of machine-readable strings to denote the different restrictions that can apply in each country. For instance if there's a single German "you must use the cycleway when appropriate" rule, we could have bicycle:restriction=DE:use_cycleway. The wiki would then contain a list of all the bicycle:restriction values, together with explanations of their precise rules, how to spot the signs on the ground, and any exceptions that apply (e.g. for trikes, large groups, etc). Routers that wish to attempt to implement these rules then have the option of doing so.

I believe that this approach offers the best of both worlds. We have a more standard access tag value on the bicycle key, that general routers are more likely to interpret in a sensible fashion, and that is likely to be more appealing to those who think that bicycle=no would suffice. And we have a machine-readable tag to record the precise restrictions in effect for users and routers that do want to know about the details. The dual-tag approach and forcing people to record the precise restriction also makes it less likely for the tags to be mis-used by people adding them for non-legal reasons.

--Rjw62 (talk) 12:20, 4 December 2013 (UTC)

I would suggest bicycle:restriction:foobar=yes type of tags, so ways with multiple restrictions may be tagged (see shops - good luck with tagging something that fits multiple categories at once). So we could have bicycle:restriction:use_cycleway=yes --Bulwersator (talk) 18:21, 4 December 2013 (UTC)

Rjw62, I support your suggestion. Except that I think we don't need to "document the precise type of restriction that applies" on tag level. Since those access rights usually depend on country-specific law, why not define them on the meta level instead, as done for the default access restrictions? --Emkey08 (talk) 10:46, 5 December 2013 (UTC)

@ Rjw62. Thanks for your view on the matter. I must say I think this proposal is second best to the bicycle=use_cycleway  ;-) . That needs a little explanation. We agree on the fact that we have to use the bicycle key. If we would not then we could have contradicting information like bicycle=yes (or no) and and a tag like “bicycle:use_cycleway=yes” . I like a tagging scheme that makes it impossible to add contradicting information (by accident). (thats why I prefer the oneway=no to express a two way street and not a twoway=yes ;-) ) So far we agree on using the bicycle key.

I also like a scheme that makes it easy for mappers. One key/value combination is easier then 2. I think we can use one key/value combination if the keys are complementary and do not contradict each other (or could be both valid at the same time?).

I think that bicycle access (generalised) could be split in 3 different options.

A. Bicycle=yes (including more specific values like designated & official)

B. Bicycle=no ( ban sign)

C. Bicyle=somewhere in between yes and no (use_cycleway, destination)

Option C in your proposal woud be : bicycle=restriction, bicycle:restriction=use_cycleway (or similar). But if we want to draw one line the “destination” tagging would become : bicycle=restriction & bicycle:restriction=destination if I understand correctly. That would need changing the bicycle=destination to the new tagging. Is this how you look at it? Just curious not cynical.

Maybe even go so far as to split the “yes” options in : bicycle=yes & bicycle:yes=compulsory? I admit that cycleways could be both designated, compulsary and official but I don’t think this will lead to many practical complications. From a logical point of view I can understand your proposal but from a practical point of view I would say it is not necessary to use 2 key/value combinations. (and maybe retagging the bicycle=destination?). So I’m not saying your proposal is wrong (far from it) but I just prefer the original proposal.

I must admit that the name “use_cycleway” is arbitrary and not 100% clear but I’m in for a better name to express that access implications on the road is due to the fact that there is a parallel compulsory cycleway.

I agree with Emkey08 in the sence that I also hope the we’re not going to tag all the exceptions but use default country specific access restrictions. This makes life for mappers easier. -- PeeWee32 (talk) 10:28, 7 December 2013 (UTC)

foot=use_footway

If this proposal happens to be approved, then we are in for foot=use_footway. --Fkv (talk) 20:36, 4 December 2013 (UTC)

Why? What it would mean? --Bulwersator (talk) 10:12, 5 December 2013 (UTC)

It means that pedestrians have to walk on the sidewalk if there is one present. This is usually also dictated by law in most countries. --Emkey08 (talk) 10:49, 5 December 2013 (UTC)

I hesitate to say this, but we would also need to tackle the (two categories of) mopeds too.Frankl2009 (talk) 14:53, 5 December 2013 (UTC)