Talk:Lane assist/Examples/Motorway exit

From OpenStreetMap Wiki
Jump to: navigation, search

Setting the split (location of the motorway_junction node)

In October 2012 two discussions have been going on @ tagging and 2nd discussionabout sections 4-6 of this image:

Lanes Example 2.png

The first discussion (legally separated ways) got into two subdiscussions. One one for divided highways, the other one was for tagging the situation of a motorway_exit. The opinion on divided highways was on average to ignore solid lines on the highway, a legal separation should not lead to two separate ways in OSM. A special tag, for example the divider tag, should be used to tag the legal separation.

The average opinion on motorway exit tagging was that the split should be early enough to get you off the exit. Remarks made on the motorway exit tagging: '...and an early indication is better than a late one' and 'in all cases of motorway junctions probably option b) is fine too' .

The second discussion, legally separated ways (2nd part), had a mixed conclusion. In France (joedalton85) it's a custom to set the split (the motorway_junction node) at the start of the deceleration lane. In Italy (Simone) it depends on the situation. In several situations the split is set at the middle of the deceleration lane. In Germany (Martin) the split is set where the restricted zone starts.


In the used motorway_exit example you can see the split being set just a bit before the restricted zone. Why? 1) it gets you off the exit in time (no-one likes to have a satnav which does not show the exits in time) 2) Observing exits in Openstreetmap throughout the world, this seems the be the major tagging style for the split 3) it's consistent with the major map manufacturers, which means the millions+ users of PND's and other satnavs are used to their devices showing remaing distances to the motorway_exit in order to leave the highway in time.

The full discussions on legally separated ways

Martin Vonwald Same image as before: http://wiki.openstreetmap.org/wiki/File:Lanes_Example_2.png

Now consider part 4 to 6. At what point would you split the way coming from part 4? a) Before the end of part 4 b) At the end of part 4/start of part 5 c) In the middle of part 5 d) At the end of part 5/start of part 6 e) Tell me!


Two more questions to slip lanes: Where would you start the way of a deceleration lane? a) At the beginning of the lane b) In the middle of the lane c) At that point where it is legally not possible to leave/enter the deceleration lane d) At that point where the lane is physically separated from the motorway e) Tell me!

Where would you join the way of a acceleration lane with the way of the motorway? a) Where the physical separation ends b) Where the legal separation ends c) In the middle of the acceleration lane d) At the end of the acceleration lane e) Tell me!

If possible reason you decisions.

Martin Koppenhoefer 2012/10/17 Martin Vonwald <imagic.osm at gmail.com>: > Same image as before: > http://wiki.openstreetmap.org/wiki/File:Lanes_Example_2.png > > Now consider part 4 to 6. At what point would you split the way coming > from part 4? > a) Before the end of part 4 > b) At the end of part 4/start of part 5 > c) In the middle of part 5 > d) At the end of part 5/start of part 6 > e) Tell me!


d) isn't this almost the same question? If you don't split the way for part 5 you cannot do a) or b) and given how short part 5 is, there is no real difference between c) and d) (but splitting "in the middle" doesn't seem good advice at anytime).


> Two more questions to slip lanes: > Where would you start the way of a deceleration lane? > a) At the beginning of the lane > b) In the middle of the lane > c) At that point where it is legally not possible to leave/enter the > deceleration lane > d) At that point where the lane is physically separated from the motorway > e) Tell me!


don't understand c) currently you should do it at point d) to be consistent with what we have said before.


> Where would you join the way of a acceleration lane with the way of > the motorway? > a) Where the physical separation ends > b) Where the legal separation ends > c) In the middle of the acceleration lane > d) At the end of the acceleration lane > e) Tell me!


a) according to what we have said

actually the by far most often practised mapping according to my experience is b) and for the first question c) which also make certain sense in the case of a motorway. In practical use there is mostly never a difference, as on the motorway sat navs tend to tell you already kilometers before what your next steps will be.

Martin Vonwald 2012/10/17 Martin Koppenhoefer <dieterdreist at gmail.com>: >> Now consider part 4 to 6. At what point would you split the way coming >> from part 4? >> a) Before the end of part 4 >> b) At the end of part 4/start of part 5 >> c) In the middle of part 5 >> d) At the end of part 5/start of part 6 >> e) Tell me! > > > d) > isn't this almost the same question? If you don't split the way for > part 5 you cannot do a) or b) and given how short part 5 is, there is > no real difference between c) and d) (but splitting "in the middle" > doesn't seem good advice at anytime).

I just summarized all variants I've seen in the last days. Well - not all. I omitted the variant where the split was after part 6. Same for the other two questions. ;-)


> actually the by far most often practised mapping according to my > experience is b) and for the first question c) which also make certain > sense in the case of a motorway. In practical use there is mostly > never a difference, as on the motorway sat navs tend to tell you > already kilometers before what your next steps will be.

Correct. But what I've seen on motorways lately it is obvious that we need some more documentation here. And I want a documentation where most (all is impossible) agree on.

Simone Saviolo 2012/10/17 Martin Koppenhoefer <dieterdreist at gmail.com>

> 2012/10/17 Martin Vonwald <imagic.osm at gmail.com>: > > Same image as before: > > http://wiki.openstreetmap.org/wiki/File:Lanes_Example_2.png > > > > Now consider part 4 to 6. At what point would you split the way coming > > from part 4? > > a) Before the end of part 4 > > b) At the end of part 4/start of part 5 > > c) In the middle of part 5 > > d) At the end of part 5/start of part 6 > > e) Tell me! > > > d) > isn't this almost the same question? If you don't split the way for > part 5 you cannot do a) or b) and given how short part 5 is, there is > no real difference between c) and d) (but splitting "in the middle" > doesn't seem good advice at anytime). >

Just to be clear: I agree with Martin Koppenhoefer on splitting the ways only if there is a physical division. However, for the sake of geometry, I prefer to anticipate the split a bit, which is why I would put the split in section 5 or maybe even earlier - the actual position depends on how long section 5 is and how harsh the split is.

Simone Saviolo Just like before, it depends. I tend to join/split lanes a bit early, so either in the middle of the acceleration/deceleration lane or earlier, trying to make a corner that is sharp enough to be acknowledged as an actual "route split" but not so much that it might be considered dangerous or tight.

Paul Johnson On Wed, Oct 17, 2012 at 2:06 AM, Martin Vonwald <imagic.osm at gmail.com>wrote:

> Hi (again)! > > Thanks for all the answers. I would like to ask three more (the last > one for this week - promised!): > > Same image as before: > http://wiki.openstreetmap.org/wiki/File:Lanes_Example_2.png > > Now consider part 4 to 6. At what point would you split the way coming > from part 4? > a) Before the end of part 4 > b) At the end of part 4/start of part 5 > c) In the middle of part 5 > d) At the end of part 5/start of part 6 >

a) 4 lanes. b) Split to two ways of two lanes. c) Two ways of two lanes. d) Two ways of two lanes.


> Two more questions to slip lanes: > Where would you start the way of a deceleration lane? > a) At the beginning of the lane > b) In the middle of the lane > c) At that point where it is legally not possible to leave/enter the > deceleration lane > d) At that point where the lane is physically separated from the motorway > e) Tell me! >

The centerlines should follow the centerlines, but where you have the fork, I'll taper from just before the solid lines start to the start of the physical separation. The //// hashmarks on the shoulder indicating a no travel/no stopping area make a handy reference.


> Where would you join the way of a acceleration lane with the way of > the motorway? > a) Where the physical separation ends > b) Where the legal separation ends > c) In the middle of the acceleration lane > d) At the end of the acceleration lane > e) Tell me! >

I usually join it in where the //// hashmarks are on the shoulder just after the end of the acceleration lane, tapering in from the end of physical seperation.

Johan C good thing to have this discussion. Too often I've seen OSM discussions end up in 'everything is possible' which in the long run will prevent OSM to ever grow-up and eventually become competitive to the commercial boys and girls. (why the f... are millions of Android users using G.. maps and not OSMAND, Navfree or Mapfactor?)

The example is not very clear, but I interpret it as follows: a. it's a motorway example (not an example of divided highways where 5 shows the divided highway) b. 5 is a single or double continuous (mostly white) line c. the point between 5 and 6 is the restricted zone of the motorway exit with the /// stripes on it and/or painted fully white. It's the same point where the y-shape starts. Since OSM is not only for experts but also for beginners, it might be better to improve the example.

Question 1 Now consider part 4 to 6. At what point would you split the way coming from part 4? a) Before the end of part 4 b) At the end of part 4/start of part 5 c) In the middle of part 5 d) At the end of part 5/start of part 6

Answer 1 on motorways b), but it's no problem if the motorway_link (with the motorway_junction node) starts a little bit earlier to avoid a sharp bend in the motorway_link (example: http://tools.geofabrik.de/mc/?mt0=mapnik&mt1=googlesat&lon=11.66657&lat=48.01287&zoom=18 ) Arguments: consistent with the professional tagging schemes by Navteq, TeleAtlas (and G..) / helps me to get off the motorway in time / I really hate missing an exit :-) / is (very probably) the most used tagging method in OSM

Question 2 Where would you start the way of a deceleration lane? a) At the beginning of the lane b) In the middle of the lane c) At that point where it is legally not possible to leave/enter the deceleration lane d) At that point where the lane is physically separated from the motorway

Answer 2 on motorways: a) Argument: a lane starts where it starts, as that's not in the middle nor at the end.

Question 3 Where would you join the way of a acceleration lane with the way of the motorway? a) Where the physical separation ends b) Where the legal separation ends c) In the middle of the acceleration lane d) At the end of the acceleration lane

Answer 3 on motorways: b), but it may a bit later to avoid a sharp bend in the motorway_link. Argument: consistent to deceleration tagging. Though it's less important: in practice it's not possible to miss an entry once you started driving on the motorway_link up to the motorway

Three other things: 1. One question is missing, but i'm already giving the answer to it: consistency involves tagging the full acceleration lane too with the extended lanes scheme, from start (answer 3) till end 2. in my opinion a motorway_link is not an example of a divided highway as described here: http://wiki.openstreetmap.org/wiki/Editing_Standards_and_Conventions 3. in my opinion a motorway (that is: about 99% of all motorways in Europe) is an example of a divided highway

Johan C I today got the following response by joedalton85 (tags a lot of motorways in France):

pour ma part, je prefere mettre les _link dés qu'elle commence ou finnissent pour plusieurs raisons.

  - Ca permet d'avoir une longueur de voirie plus proche de la réalité
  - Ca permet d'avoir une occupation du sol plus proche de la réalité
  - Ca permet au GPS de prevenir " prenez à droite" au bon moment
  - En france on doit se tourner à droite des que possible

Translated (and based on viewing some of joedalton's edits):

For my part, I prefer to put the _link where the deceleration lane starts or where the acceleration lane finishes for several reasons.

  - a long road is closer to reality
  - It provides a land closer to reality
  - It allows the GPS to show "turn right" at the right time
  - In France one must turn to the right when possible.

(translations welcome, both French and English are not my native language)

Alberto Ferraris >Just to be clear: I agree with Martin Koppenhoefer on splitting the ways only if there is a physical division. However, for the sake of geometry, I prefer to anticipate the split a bit, >which is why I would put the split in section 5 or maybe even earlier - the actual position depends on how long section 5 is and how harsh the split is.

+1 I agree with Simone and Martin K.




Legally separated ways

Martin Vonwald Some kind of short how-would-you-tag-this-survey. Have a look at part five of this motorway: http://wiki.openstreetmap.org/wiki/File:Lanes_Example_2.png Only part 5 is relevant. Assume there is no physical separation just a double line between the upper and lower two lanes. How would you tag this: a) One way with lanes=4 b) Two separate ways with lanes=2 each c) Tell me!

Please also reason your decision.

Colin Smale I would choose option b).

Even if all four lanes are one piece of carriageway, it is useful for routing directions etc to be able to make a distinction between the left and right parts of the road. Normal mortals are supposed to treat the solid white lines as if they were a brick wall anyway, and be on the correct part of the road before part 5 starts. If we start tagging exceptions to traffic laws for emergency vehicles then we have an awful lot of work to do... I am not sure I would interpret the diagram in that way though; the fact that there are no arrows on the road from part 5 onwards suggests to me that there is no chance of changing your mind. I think the intention is that part 5 is the start of physical separation, but I might be wrong here.

Martin Vonwald > I am not sure I would interpret the diagram in that way though; the fact that there are no arrows on the road from part 5 onwards suggests to me that there is no chance of changing your mind.

There is no chance of (legally) changing your mind, because there are double solid lines ;-)

> I think the intention is that part 5 is the start of physical separation, but I might be wrong here.

Sorry, you are wrong - I drew the image ;-) But because the image is not 100% clear I added a note in the related article and now I have to make sure the note is correct and clear.

Janko Mihelić One way with lanes=4. If firemen ever want to use osm, i want them to have good data.

Pieren > Sorry, you are wrong - I drew the image ;-) But because the image is not 100% clear I added a note in the related article and now I have to make sure the note is correct and clear.

Your image and note are not really helping. The figure tend to represent the physical layout but a footnote says that it is just a legal separation. And even worse, you say that some people are violating the general guideline... So please, for the newcomers and sake of clarity, improve your picture and draw a single solid line only or any thing that do not confuse readers. Because what is "2 solid lines" in section 5 becomes a real physical seperation on section 6. And remove the footnote comment admitting mistakes. This is happening daily in OSM but it is not a reason to accept them in the wiki doc.

Tobias Johansson I created a new picture very much based on the old one. Just made the road gray to try to make it more clearer? http://minkarta.no-ip.org/Lanes_Example_2.svg Im not sure how to upload it so if anyone thinks this is better please so, otherwise I atleast learned a little what I can do in inkscape :).

Volker Schmidt I would opt for (b) even though I know that this is not the "offcial" way f tagging. The reason: in section (4) the driver can still change lanes, at least on the middle lanes, whereas in section (5) he cannot (legally) change lanes any more between the middle lanes. This example clearly illustrates the - known - limitations of the presently used tagging scheme.

Tobias Knerr I would choose (a) and reserve separate ways for _actual_ physical separation. One practical reason for doing so is that there would simply be no possibility to distinguish legally separate ways from physically separate ways if we used the same mapping (splitting the way) for both. Splitting the way at that point could be considered "tagging for the router", in my opinion, and ignores the needs of applications that do use OSM for something else than navigation.

Martin Koppenhoeffer > I would choose (a) and reserve separate ways for _actual_ physical separation. One practical reason for doing so is that there would simply be no possibility to distinguish legally separate ways from physically separate ways if we used the same mapping (splitting the way) for both. Splitting the way at that point could be considered "tagging for the router", in my opinion, and ignores the needs of applications that do use OSM for something else than navigation.


+1 to all of this. If we want to be able to distinguish physical and legal separations we have to stick to our own rules. Anyway there will always remain a slightly unsatisfactory geometry situation on points like these, because you have to do the transition from 1 way (in the middle of 4 lanes) to 2 ways (each in the middle of 2 lanes). There is simply no really elegant way to do this. Additionally to the lanes=4 and oneway=yes you could put a divider-tag on the way http://wiki.openstreetmap.org/wiki/Proposed_features/Divider even if it doesn't explicitly tell you where the divider is placed you might be able to infer it from the following ways (at least in this case).

Tobias Johansson I use b. Because thats the only way I was/am aware of to tag whats on the ground. A divider. But yes I would like there to be an easy way to tag it in the same way.. I even think there should be a possibility to tag a physical divider in the same way, because some roads with physical dividers really dont need to be drawn as two seperate ways.

Simone Saviolo > I would choose (a) and reserve separate ways for _actual_ physical separation. One practical reason for doing so is that there would simply be no possibility to distinguish legally separate ways from physically separate ways if we used the same mapping (splitting the way) for both. Splitting the way at that point could be considered "tagging for the router", in my opinion, and ignores the needs of applications that do use OSM for something else than navigation.

+1, BUT...

I agree with you that a long section of road with double solid lines would have to be tagged as a single way. If not, an ambulance would be routed to go to the end of the double solid lines and then go back, even though in reality it just needs to cross the road (even though such an example sounds silly, please consider that a route involving this situation would come out all wrong because of weird paths).

However, if we found ourselves with the situation depicted in the image (I'm talking about the full image here, not just section 5), then I'd say that we start the way splitting as soon as it is topologically correct. If the way splitting happens fifty metres ahead of where an actual physical barrier begins, it should be no problem for any consumer, and we would still get an acceptable model of reality based on our base concept of drawing road by their centerline.

Bill Ricker > I would choose (a) [One way with lanes=4] and reserve separate ways for _actual_ physical separation.

+1. Agree, different reasoning. Paint is not asphalt. (Maybe this legal separation should be like a turn restriction.) Early notification of coming turn is build into GPS so you can pick the right lane - possibly before lane-must-turn solid lines trap you. Around here, all too often the solid line appears BEFORE any arrows. Good lick with that.) +1 Alberto also.

Markus Lindhlm The answer is b. But as I'm sure you've noticed there's some divided opinion about this

Martin Vonwald > But as I'm sure you've noticed there's some divided opinion about this.

That's why I asked! Actually I don't think that we see any consensus about this soon. But then I can document at least that there are two variants under discussion. If I claim in the wiki that a) is the ultimate solution it will be "fixed" by supporters of b) and vice versa. As I don't like edit wars I prefer to write the truth: both are used. I don't claim this is perfect (or at least good) but it is the current status-quo. As soon as there is a consensus about this issue I'm happy to update all affected wiki articles. But I'm a little afraid that I won't live long enough ;-)

Alberto Ferraris a) because distinction between physical and legal barriers is important. Ok in that picture there is no much difference, but as Simone pointed out, for long roads there is a big difference: if any router can't distinguish between physical and legal barrier, it will not suggest to emergency vehicle to cross the line, and it will tell them to do a long alternative trip instead. This is a big problem, because when you are driving the GPS shows you only a little portion of the route and you may not understand (especially in emergency) that you can shorten the route simply crossing the line. Moreover if we accept solution b) we should tag every road with continuous line in the middle as two separate roads, one for each direction.

>Additionally to the lanes=4 and oneway=yes you could put a divider-tag on the way http://wiki.openstreetmap.org/wiki/Proposed_features/Divider even if it doesn't explicitly tell you where the divider is placed you might be able to infer it from the following ways (at least in this case).

+1 We can improve this proposal, to make clear where the divider is placed.

Colin Smale I don't understand why emergency vehicles are so important in this discussion. In the first place they have wide-ranging exemptions from traffic rules, which (let's be honest) we are never going to tag in OSM. Secondly they are never going to be relying on OSM data (or indeed any normal sat-nav) for lane-precise routing. They are trained to use their eyes and brains to make split-second decisions on what is safe and an acceptable risk under the circumstances of that moment. Thirdly, they will be about 0.0000000001% of the potential users of OSM data - why should we compromise "service" to the vast majority of real users for the hypothetical benefit of the very few.

Eckhart Wörner I fully agree with you; if we were going to map for emergency vehicles, we'd probably have to add neway:conditional = no @ emergency for almost all oneway roads first.

Svavar Kjarrval I think most laws require that even emergency vehicles observe restrictions like oneway streets. If there are any restrictions which can be broken in case of emergency vehicles, I think they'd program their routing software to them.

Colin Smale >I think most laws require that even emergency vehicles observe restrictions like oneway streets. If there are any restrictions which can be broken in case of emergency vehicles, I think they'd program their routing software to them.

Not in the UK or the Netherlands at least. They can do whatever they see fit in the course of duty, especially with lights and sirens. Of course they can be called to account if anything goes wrong. But a policeman chasing a criminal who turns the wrong way up a one way street is going to follow them and is not going to take a longer route just because of a little no-entry sign. Maxspeed, turn restrictions, oneways, waiting/stopping/parking restrictions, traffic signals - you name it, they can (and do) ignore it when required (sometimes only under certain circumstances such as life-and-death). Having said that there are certain restrictions which they would do well to follow, such as maxheight and maxwidth. If it don't fit, it don't fit, with or without blue lights!

Svavar Kjarrval > Not in the UK or the Netherlands at least... Then it could be easy for them to tailor those requirements to ignore restrictions into the routing software.

Philip Barnes The law varies from country to country. In the UK, it is legal to cross a solid white line to turn into a side road, or driveway. You can also cross one to overtake a slow moving vehicle, such as a cyclist or tractor. In France, where it is illegal to cross a solid line even to enter a driveway. It is common to see short gaps in the solid line to allow traffic to turn. This would be a lot of work.

Simone Saviolo In Italy the law is similar to the French one. If turning or U-turning is allowed at a specific spot (maybe because of a lateral road), then a break in the solid line is painted. This is formally most correct: you can't trespass the continuous solid line, but *in that point* the line is not solid.

If we were to separate ways where the solid line is painted, we would have horrible maps: 1) we would need to draw a non-existing way wherever the line is broken (for example to allow a left turn into a lateral road); 2) we would need to keep splitting/joining/splitting again/joining again the ways if a long road has a section with solid line followed by a section with dashed line followed by a section with solid line and so on; 3) we would make routers incapable of understanding what kind of way splitting is happening. What would they say? Turn right? Keep right? Why should you tell me to keep right, the road is straight! Sure, we must not map for the router, but we must also try not to map AGAINST the router.

Johan C I think there's some confusion here. Imagic's question was on a motorway example.

Three things on this.

1. I've noticed these OSM'ers in favour of option a: Junker, Tobias K., Martin, Simone, Alberto and Eckhart. Could any of these OSM'ers please put an example (older than today [?]) of this tagging on this list? That's because i've never seen one anywhere in OSM yet.

I've got a lot of examples of option b on motorways. Netherlands: http://tools.geofabrik.de/mc/?mt0=mapnik&mt1=googlesat&lon=5.05512&lat=52.06882&zoom=18 Germany: http://tools.geofabrik.de/mc/?mt0=mapnik&mt1=googlesat&lon=8.71883&lat=50.06418&zoom=18 Luxembourg: http://tools.geofabrik.de/mc/?mt0=mapnik&mt1=googlesat&lon=6.08529&lat=49.58708&zoom=18 France: http://tools.geofabrik.de/mc/?mt0=mapnik&mt1=googlesat&lon=2.47817&lat=48.88683&zoom=18 UK: http://tools.geofabrik.de/mc/?mt0=mapnik&mt1=googlesat&lon=0.15021&lat=51.30202&zoom=18 Italy: http://tools.geofabrik.de/mc/?mt0=mapnik&mt1=googlesat&lon=12.38766&lat=41.82854&zoom=18 Austria: http://tools.geofabrik.de/mc/?mt0=mapnik&mt1=googlesat&lon=16.44579&lat=48.21632&zoom=18 Switzerland: http://tools.geofabrik.de/mc/?mt0=mapnik&mt1=googlesat&lon=6.13588&lat=46.25578&zoom=18

and the same applies to motorway_links (for exits). What these examples have in common is that the restricted zone (see: http://www.auguszt.de/english/VZ/zeichen4.htm) is not used. Imagics example a means the restricted zone can be used. Because it's not a physical seperator. But, as I said, please surprise me by showing a *motorway *example being used in OSM which shows example a.

2. An OSM'er started a discussion on a German page: http://wiki.openstreetmap.org/wiki/DE_talk:Tag:highway%3Dmotorway_link. It seems that the current wiki on motorway_junctions requires the start of the split to be at the beginning of the deceleration lane. "In der Regel ist das der Punkt, an dem die Ausfädelspur beginnt oder die Auffahrt von der kreuzenden Straße abzweigt." Thus, the current osm tagging standards imply that Imagic's option a is wrong (option b too by the way [?]).

3. You could have guessed, but i'm in favour of the way OSM is being used at this moment, so option b. My arguments are about the same as Colin's. And because I like the current tagging as shown in above examples.

Simone Saviolo > I think there's some confusion here. Imagic's question was on a motorway example. Where did you get this from? Sure, he referred to a picture with the model of a motorway, but he esplicitly said "consider only section 5". We're not talking about that section 5 *on a motorway*.

Also, your distinction motorway / non-motorway is meaningless. Motorways are usually mapped with two separate ways because they, uh, are physically separated carriageways. There's a guard-rail, or a new jersey barrier, sometimes even a ditch or a gap between bridges. There used to be (at least in Italy) motorways with a single carriageway, where lanes were separated by a painted line; this was not a physical separation and would be tagged as a single way.

> 2. An OSM'er started a discussion on a German page: http://wiki.openstreetmap.org/wiki/DE_talk:Tag:highway%3Dmotorway_link. It seems that the current wiki on motorway_junctions requires the start of the split to be at the beginning of the deceleration lane. "In der Regel ist das der Punkt, an dem die Ausfädelspur beginnt oder die Auffahrt von der kreuzenden Straße abzweigt." Thus, the current osm tagging standards imply that Imagic's option a is wrong (option b too by the way [?]).

As I said above, for short sections of legal separation (white paint) immediately followed by a physical separation (the guard-rail) I am in favour of anticipating the split, just like it has been suggested in the discussion you talk about. What I'm modeling this way is that there is in fact a physical separation, and I'm just being imprecise about where the separation starts. I advocate this imprecision because it has advantages on the topology of the split (Y-shaped rather than T-shaped), and an early indication is better than a late one.


> 3. You could have guessed, but i'm in favour of the way OSM is being usedat this moment, so option b. My arguments are about the same as Colin's. And because I like the current tagging as shown in above examples.

You may be confused about this. The "current tagging" is in fact option A: where the carriageways are physically separated (or at least a single continuous strip of asphalt is divided by a physical barrier), draw two ways; otherwise, a single one. Your motorway examples are examples of option A. The fact the we split ways entering a large roundabout is my bending of option A: the way is split where the white paint begins (instead of where the kerb with grass begins) for better topology. I fail to see where a Main Street with no barrier in the middle is drawn as two ways (don't bring me a couple of examples, name a country or a community that does this).

Eckhart Wörner Am Montag, 15. Oktober 2012, 18:26:02 schrieb Svavar Kjarrval: > I think most laws require that even emergency vehicles observe restrictions like oneway streets. If there are any restrictions which can be broken in case of emergency vehicles, I think they'd program their routing software to them. at least in Germany, emergency vehicles can ignore all rules of the roads if necessary (they are not allowed to endanger other traffic though).

John F. Eldredge The emergency vehicles in the USA are also allowed to disobey traffic laws in an emergency. They have to have the flashing lights and siren on, and take precautions to watch out for regular traffic. For example, a police car, fire truck, or ambulance on an emergency call can go through an intersection even if the traffic signals have ordinary traffic stopped, but they slow down enough to check for any ordinary traffic approaching the intersection on the cross-street, since some drivers may have not figured out which direction the siren sound is coming from

Paul Johnson Ambulances are only allowed 10 MPH over the posted speed limit, and then nly when operating lights and sirens. I didn't know this until I got this question wrong on the Oregon driver's exam. I reused the correct answer from Oregon on my Oklahoma exam and got it right. So, at least ambulances in two states still have to kinda-sorta obey speed rules.

Tobias Knerr On 15.10.2012 20:08, Colin Smale wrote: > I don't understand why emergency vehicles are so important in this discussion.

They are often allowed to ignore legal restrictions, but cannot generally ignore physical restrictions, so it's an obvious example where this distinction matters.

But before this discussion is completely transformed into a discussion about what emergency vehicles are or are not allowed to do, let me offer another example: realistic rendering, particularly 3D "virtual reality". Physical separation and road markings look quite different from each other and should therefore also look different in that kind of rendering - whether it is in, say, a virtual globe or maybe a road/lane assistant type of application.

Tobias Knerr

>> realistic rendering, particularly 3D "virtual reality". Physical separation and road markings look quite different from each other and should therefore also look different in that kind of rendering - whether it is in, say, a virtual globe or maybe a road/lane assistant type of application. > As long as you have width information on the ways, I don't see the problem. The amount of physical separation, whether it be an inch between two lines, or a meter, should be easily calculated from the data.

There are actually several problems:

  • Most importantly, neither width tags nor the distance between two parallel ways are mapped with one-inch accuracy.
  • If you represent the two lines as a separation between ways, then you don't know that the two lines are there - so you cannot render them.
  • The "physical separation" between two road marking lines is part of the same paved surface as the rest of the highway. If you do not know that this is the case, you cannot render this as a paved surface either.

Tobias Knerr > On Tue, Oct 16, 2012 at 9:30 AM, Tobias Knerr <osm at tobias-knerr.de> wrote: >> On 16.10.2012 01:07, Anthony wrote: >>> As long as you have width information on the ways, I don't see the >>> problem. The amount of physical separation, whether it be an inch >>> between two lines, or a meter, should be easily calculated from the >>> data. >> >> There are actually several problems: >> >> * Most importantly, neither width tags nor the distance between two >> parallel ways are mapped with one-inch accuracy. > > If this is the most important, then you are seriously stretching it. > What's the need for one-inch accuracy?

You suggested that "the amount of physical separation, whether it be an inch between two lines, or a meter, should be easily calculated from the data."

I'm saying a separation of an inch between two lines can _not_ be calculated from the data because it's not remotely precise enough, and shouldn't be expected to be.

>> * If you represent the two lines as a separation between ways, then you >> don't know that the two lines are there - so you cannot render them. > > I'm not sure what this means. The ways don't represent lines, they > represent paths of travel.

In my statement above, "the two lines" refers to the road markings.

> There are lots of paved surfaces which are not mapped. [...] > Furthermore, what do you suggest we do with section 6

If someone cares about the surface of an area which is not part of any highway, they can map it as an area with surface=*.

This does not seem like an appropriate solution for the tiny "area" between two white lines of paint, though.

Simone Saviolo 2012/10/15 Colin Smale <colin.smale at xs4all.nl>

> I don't understand why emergency vehicles are so important in this > discussion.


Because OSM publicly advertises the fact that its maps are being used in the Gaza's strip by emergency vehicles that would otherwise have no map? Just to name one. Also because emergency vehicles may happen to operate in unknown territory. Don't limit yourself to the ambulance that runs around its home town.


> In the first place they have wide-ranging exemptions from traffic rules, > which (let's be honest) we are never going to tag in OSM.


This is meaningless. We will map all restrictions; consumers will have an "emergency vehicle" flag that will route ignoring those restrictions. Also, while an ambulance is allowed to go the wrong way in a one-way street (when its siren is on, of course), it is usually advised against doing so, as in a narrow road it may find regular traffic, which would be dangerous or at least slow down the ambulance. So, restrictions may be ignored, but drivers should be informed about them.


> Secondly they are never going to be relying on OSM data (or indeed any > normal sat-nav) for lane-precise routing. They are trained to use their > eyes and brains to make split-second decisions on what is safe and an > acceptable risk under the circumstances of that moment.


Sure. Let's make an example. There's a long primary road between towns, with solid double lines all the way from town A to town B. Let's say it runs North-South from town A to town B. A farmyard east of the road is burning. It can be accessed by a road that reaches the large road; under normal circumstances, someone coming from town A couldn't reach the crossroad and go to the farmyard, but would be legally forced to go to town B, turn back, reach the crossroad and go the farm.

The firemen turn on their GPS navigator, because they're being called in from a far away city and they've never heard the name of that farmyard. The router lawfully sends them to town A, then to town B. They go past the crossroad and can't notice the farmyard, because it's far away and there's a wood in between that covers it. Also they can't see on the screen that ten kilometres ahead they would have to turn back and go back to the crossroad they're approaching. The firemen get there fifteen minutes late and the farmyard is a bunch of ashes. Great work dividing the way on a legal restriction! :-)


> Thirdly, they will be about 0.0000000001% of the potential users of OSM > data - why should we compromise "service" to the vast majority of real > users for the hypothetical benefit of the very few.


I know who makes this consideration: commercial map providers. It's just not worth the cost, right?

Colin Smale Are you seriously suggesting that emergency services will trust a satnav in preference to their own eyes and brains? Especially a satnav driven by data with no proactive quality control and no-one you can sue/complain to? And seriously incomplete data? I think you are looking at a multi-year project to get all this information into OSM (read: "review and correct the tagging and topology of every road in the database") while all the time thousands of people are adding new "bad" data? A popular dutch saying "mopping up with the tap still running" comes to mind.

Let's not have a purely hypothetical debate, let's keep it practical. Assuming that emergency services currently use satnavs (consumer or special-purpose) based on commercial data, what would they say if we asked them "what would you need from OSM data to make it a better choice than your current supplier?" I can imagine that the time delay between changes on the ground and their availability in a map update might be one concern; inappropriate road classifications might be another. I would like to hear it from them, though. Then we can look at the requirements and assess whether it is a viable project.

Simone Saviolo > Are you seriously suggesting that emergency services will trust a satnav > in preference to their own eyes and brains? >

I hope not, and I hope this is true for everyone and not only for emergency vehicle drivers. However, not all places are within eye distance. An ambulance doesn't need a router to see that the target is on the other side of the road, but crossing that solid line may be a part of a much longer route. It may be what makes the difference between a long, convoluted route and a very short one.


> Especially a satnav driven by data with no proactive quality control and > no-one you can sue/complain to? And seriously incomplete data? I think you > are looking at a multi-year project to get all this information into OSM > (read: "review and correct the tagging and topology of every road in the > database") while all the time thousands of people are adding new "bad" > data? A popular dutch saying "mopping up with the tap still running" comes > to mind. >

So let's shut down any publishing service of OSM data. They're incomplete! How can you trust incomplete or inconsistent data to tell you where you should go to go to that restaurant?

OSM's data is not "seriously incomplete". And I understand the difference between day-to-day map needs and emergency needs. But if I were an ambulance driver heading to an unknown place, maybe one in the countryside, I'd much rather use OSM, where tracks are marked as tracks, than Google Maps or Tom Tom, where tracks are usually marked the same way as a primary or secondary road.

Besides, if an emergency service were interested in using OSM, it would be easy for them to update the map as real-world changes happen. Not the same with a commercial map. But I hope you know all the advantages by now.


> Let's not have a purely hypothetical debate, let's keep it practical. > Assuming that emergency services currently use satnavs (consumer or > special-purpose) based on commercial data, what would they say if we asked > them "what would you need from OSM data to make it a better choice than > your current supplier?" I can imagine that the time delay between changes > on the ground and their availability in a map update might be one concern; > inappropriate road classifications might be another. I would like to hear > it from them, though. Then we can look at the requirements and assess > whether it is a viable project. >

It's not a purely hypotetical debate. Nobody here is saying that we should use some who-knows-what tag because possibly maybe I-think-that it could be a useful information in some absurd situation. We are just saying that mapping a road with two non-communicating ways where the two ways would actually have to be communicating is not only wrong from a conceptual point of view, but it also creates errors and hard-to-understand representations of a simple reality for data consumers. Seriously, I can't see why we're complicating this!

Markus Lindholm To be able to do proper routing for emergency vehicles perhaps it would be a good idea to introduce something like landuse=highway that would denote an area suitable for motor vehicles and that is free of physical obstacles.

Martin Koppenhöfer There is a relation proposal (area) that suggests a solution to this without explicitly drawn closed areas

Simone Saviolo Martin, could we please try to revive the proposal and make it go forward? I am one of those who would like to draw roads as an area, as you know :-)

Janko Mihelić I posted this picture the last time this came up. It shows that dividing roads is silly in some situations, for example countryside roads:

http://i.imgur.com/p5Oto.png

You have to divide the road each time there is not a full line on the road, ad you should put a restriction where those roads meet that restricts U-turns. What is the answer to that?

Colin Smale There's maybe a difference between the case of two lanes in the same direction, and two lanes in opposite directions.

Simone Saviolo > There's maybe a difference between the case of two lanes in the same > direction, and two lanes in opposite directions. >

There's none. If a solid line is painted between lanes going in opposite directions, it's legally impossible to cross the line. If a solid line is painted between lanes going in the same direction, it's legally impossible to cross the line. The directions of the lanes makes no difference at all.

Chris Hill > If a solid line is painted between lanes going in opposite directions, > it's legally impossible to cross the line. > If a solid line is painted between lanes going in the same direction, > it's legally impossible to cross the line. > The directions of the lanes makes no difference at all.

Perhaps in your part of the world, but not everywhere. Crossing solid lines, as centrelines or lane separators have exceptions for ordinary vehicles (not just emergency vehicles) here. Yet another example of how local influences must be applied to documentation. The page showing how highway types are interpreted in different countries may be copied for such definitions.

On the other hand, maybe we should stop kidding our selves that OSM, as a wiki, can ever reflect the legal situation, since anyone can change it at any time. Maybe we should try to map the physical characteristics and leave the legal interpretation to the drivers who are required to interpret what they see on the ground before them and cannot rely on any map or database for legal guidance. Every pair of hands on a steering wheel comes with a free brain.

Pieren > Maybe we should try to map the physical characteristics and leave > the legal interpretation to the drivers who are required to interpret what > they see on the ground before them and cannot rely on any map or database > for legal guidance. It's always much harder for a data consumer. I guess a tag "oneway=yes" on a way is easier to interpret than a floating node tagged with "traffic_sign_color=blue" + "symbol=white_vertical_arrow"... We always translate what we see into something that can be interpreted by applications (means not only renderers but routers, etc). Another example ? The tag address. What you really see physically is always just a number. But nobody just adds the "addr:housenumber" tag because instinctively they know that data consumers may fail into linking the number and its street name, especially at intersections. So a tag "do not cross" is more valuable in OSM than "a single or double painted solid line" tag.

> Every pair of hands on a steering wheel comes with a > free brain. hmmm. not sure about that. We all heard stories about drivers blindly following GPS instructions ...

Martin Koppenhoefer > Perhaps in your part of the world, but not everywhere. Crossing solid lines, > as centrelines or lane separators have exceptions for ordinary vehicles (not > just emergency vehicles) here. Yet another example of how local influences > must be applied to documentation. The page showing how highway types are > interpreted in different countries may be copied for such definitions.


yes, traffic rules are different in different areas, and someone offering routing should make himself familiar with local exceptions. This is even more an argument to keep straight definitions. If you map that there is no physical division but only a specific legal divider, this information will be available to those who interpret the data according to their rules, but if you treat all kind of divisions the same (separating the ways in OSM) routers will asume for everybody (pedestrians, emergency vehicles, bicycles) that there is no connection (unless you create these connections, either explicitly (connecting (foot?)ways) or conceptionally by creating a relation between the two).

When you choose to split the ways and create a relation between the two, there will still remain some questions open. Actually you shouldn't tag them as highways then, but as lanes, so the question is how to deal with this.

Would you make a) no highway at all and dataconsumers would need to take all lanes as well in account (or maybe get single highways from preprocessing) or they would have holes in their data. b) assign the highway-tag only to one of the ways (actually not working without introducing further "helper ways") c) assign a highway-tag anway to all these lanes and additionally state that they are lanes. d) do nothing of this and renounce from the details.


Frankly I think that none of these alternatives with the relation is really nice. b) would probably be the easiest to introduce, because dataconsumers could in theory ignore the lanes and rely only on the highways, but practically you would have to continuously fix newly introduced holes in the graph, because of higher complexity.

The divider tag (almost not in use) could solve the problem we have in countries where there is no exception for driveways on the opposite side, and in general it could make mapping easier because far less turn restrictions would be needed (if routers adopted this mapping style).

Martin Koppenhoefer > I posted this picture the last time this came up. It shows that dividing > roads is silly in some situations, for example countryside roads: > > http://i.imgur.com/p5Oto.png > > You have to divide the road each time there is not a full line on the road, > ad you should put a restriction where those roads meet that restricts > U-turns. What is the answer to that?


I find it strange that we are still discussion whether roads that are only legally divided should/could be split into 2 parallel ways, a solution that we did - since ever - reserve for physically divided roads.

The answer to your situation above could be the divider-tag, applied on a single way, where it would solve many of these situations. For more complex situations (several parallel lanes, some of them divided legally, others not) we still would to have something at lane-level, as the simple divider approach does not tell you in these cases, where the divisions apply.

The interruption of the continuous line on ~points (=very short ways not much longer than the width of a lane) could be modeled at node-level (to avoid excessive way splitting) and for longer pieces it would be on the way (which would then have to be split).

Erik Johansson <dieterdreist at gmail.com> wrote: > 2012/10/16 Janko Mihelić <janjko at gmail.com>: >> You have to divide the road each time there is not a full line on the road, >> ad you should put a restriction where those roads meet that restricts >> U-turns. What is the answer to that? > > > I find it strange that we are still discussion whether roads that are > only legally divided should/could be split into 2 parallel ways, a > solution that we did - since ever - reserve for physically divided > roads.

It's basically "Tagging for the renderer". As long as the divider tag is not supported by routers, people will continue to use this solution, and these discussion will only make the situation worse for the divider tag.

Alberto Ferraris Please read carefully: one thing is a legal barrier (e.g. continuous line), one thing is a real barrier (e.g. guard rail, Jersey barrier). A legal barrier can be crossed by a vehicle if necessary, a real barrier not. We don't need to tag any exception for emergency vehicles. The routing software (now or in future) can be set to ignore legal barriers in emergency navigation. But if we don't tag in different ways legal and real barriers, there is no way to distinguish them. Think to this case: a motorway with continuous line and another with guardrail in the middle, and your destination is behind you. If routing software can't distinguish them, it will in both cases suggest to take the first exit 5 km ahead, and it may calculate a tortuous route along secondary ways. In your GPS you see "5 km ahead, go right" and a part of the route. If you don't know that your destination is just behind you, how can you imagine that the suggested route is misleading and you can simply do a U turn? In emergency situations you have many things to pay attention to and maybe you don't notice that GPS is wrong. I know these situations, I'm a fireman. Ok, usually you know your operating zone and you can ignore GPS, but why should we deliberately put incorrect data in the map?

By the way, in all cases of motorway junctions probably option b) is fine too, there is no much difference. But it is important to understand that using option b) for longer ways implies turning a legal barrier into a physical barrier, and that can produce unwanted and unexpected consequences.

>Having said that there are certain restrictions which they would do well >to follow, such as maxheight and maxwidth. If it don't fit, it don't >fit, with or without blue lights! Exactly, and a guard rail is one of those restrictions that you can't pass with or without blue lights, we should distinguish it from a line drawn on asphalt.

> why should we compromise "service" to the vast majority of real users for the hypothetical benefit of the very few. I can't see how the service can be "compromised". We can anyway refine and use divider proposal: http://wiki.openstreetmap.org/wiki/Proposed_features/Divider

Paul Johnson I'd go with option b. Despite being a single way, you're committed to taking the ramp by that point (due to the double-white solid lines), making it functionally an extension of the ramp.

Simone Saviolo > I'd go with option b. Despite being a single way, you're committed to > taking the ramp by that point (due to the double-white solid lines), making > it functionally an extension of the ramp.


The OP explicitly asks you to focus on section 5 alone, NOT on section 5 as the introduction to section 6.

Kytömaa Lauri >Only part 5 is relevant. Having just returned from my (mapping) trip, and having finally browsed through all these messages on this subjet, I don't think anybody mentioned it explicitly: You can't consider only part 5. At part 6, the ways are physically separated, so IMO there should be two separate nodes on the 5 - 6 border. With two nodes there, the transition from one way (part 4) two two ways happens during part 5. Often, but not always, before the physical separation begins, the solid line is of convenient length for said transition.

>Assume there is no physical separation just a >double line between the upper and lower two lanes. How would you tag >this: >a) One way with lanes=4 >b) Two separate ways with lanes=2 each >c) Tell me!

Depending on the length of part 5, fully b, or if it's a really long part, first as one way, and as two ways from the point where one or both of those ways' angles match the angles of the ways in part 6. Where I'd draft the definition of "match" as "equals, or makes a bit shallower or a bit steeper angle with the orientation of the part 5 carriageway"; this is to say "no abrupt right-left-curves, but don't do separate ways at the start of a long decelaration lane either".

Simone Saviolo > >Only part 5 is relevant. Having just returned from my (mapping) trip, and having finally browsed through all these messages on this subjet, I don't think anybody mentioned it explicitly: You can't consider only part 5.


I think he means that, for lack of a better picture, he reused this one and said, ok guys, let's pretende for a moment that we have a road that is completely made like section 5 here. If we were to consider section 5 in the context of that picture, however, I'd agree with you, as I said in previous messages. I think the whole debate has been created by this misunderstanding.