From OpenStreetMap Wiki
Jump to navigation Jump to search

Unposted overtaking restrictions

Seems that the wording I wrote years ago omitted a crucial point: overtaking bans are not always visible in the form of solid lines or traffic signs. In most countries overtaking is also forbidden when the visible stretch of road isn't long enough, because of a hill summit or because of curve, or other similar (road layout) reason limiting visibility. On smaller roads these aren't marked in any way, but they can not be deduced from any data we have, nor could they be from any dataset we could reasonably expect to become available in the next decade or so (terrain models etc.). (If a vehicle has broken down in such a "blind spot", getting around it with extra caution is AFAIK not consider "overtaking", but who-knows-what-it's-called in English mandatory special manouver.)

Tagging overtaking=both in a tight bend where nobody can legally overtake a moving vehicle would seem utterly stupid. Even if the height of the driver's seat has a minor effect in some of the hill summit cases, the exact length of the implied restriction is less interesting than the fact that it exists - and hgv drivers sitting really high can use overtaking:hgv. As to the effect of the slower vehicle's speed (you need less room to overtake farm machinery than a car going 10 under the limit), the length and curve of the section can help somewhat?) To make it clear in these cases that there are no official indicators of the restriction, one could tag (for example a hill summit) as below. On one side, where the osm way(s) first go "uphill":

  • overtaking=backward up to the highest point (meaning: The drivers going downhill see far ahead, but those going uphill don't see enough for a legal overtake)
  • overtaking:forward:reason=visibility (or just omit the ":forward" in the middle) (The reason why you can't overtake, even if there are no signs, is that you don't see far enough.)

and then after the hill summit, where the osm way(s) go downhill:

Naturally, the "highest point" above can be an intermediate in an uphill slope, the point where the incline changes so much that the visibility is reduced even if both sections are uphill in the same direction.

The only other option would be to omit overtaking=* completely for both those stretches, and to use for example overtaking:forward:visibility=no + overtaking:backward=yes. Which makes it more error prone for consumers, as there's a risk of conflicting data and a higher risk of mappers mixing up their intentions and the data. When the actual restriction is in overtaking=*, it's always either/or. Alv (talk) 07:08, 3 October 2013 (UTC)

Turn_left allowed?

I think that this tag implies turn_left forbidden. then any turn to a way at left is forbidden. This way many, many restriction relations can be avoided. Pedrobv (talk) 04:53, 4 September 2015‎ (UTC)

While it's possible on that on major roads the solid middle line in that direction is continuous through the sideroad intersections, effectively banning left turns, quite a few intersections introduce a gap in the banning line just to allow left turns. Also, I don't remember seeing any of these overtaking bans extending through intersections in urban areas. Additionally, if we were to assume the left turns were banned, we would still lack a way to indicate which intersections did allow the left turns when within the overtaking restriction. Alv (talk) 05:52, 4 September 2015 (UTC)
Does exists a Tag that banns left turns for major roads? Those that you have to drive to the next roundabout to turn left. Pedrobv (talk) 17:19, 4 September 2015 (UTC)
Not really, at least such tags are not supported by any current software. (P.S. It's considered good to sign your comments to wiki talk pages, you can do that by typing four tilde signs ("~") at the end.) Alv (talk) 18:18, 4 September 2015 (UTC)

Definition over "overtaking"

A definition of "overtaking" would be helpful. In the US we use the term "passing" whether you are using a lane going in the same direction as you to pass someone or a lane going in the opposite direction. If I apply overtaking=no to a four-lane road, I want to know what that implies or if it is wrong/assumed. Germyb (talk) 21:01, 27 August 2017 (UTC)

My interpretation is that the key is about overtaking/passing, regardless which lane is being used for that. So on a four-lane road, it would still ban overtaking, even if you don't need to use the opposite direction's lanes for that. --Tordanik 18:02, 29 August 2017 (UTC)
So is there a simple way to record that you can't use lanes going in the opposite direction for passing? Should I use change=*? It seems like more to maintain if the number of lanes changes frequently because of, say, turning lanes. Germyb (talk) 22:38, 30 August 2017 (UTC)
At the moment, change=* indeed seems to be the best available solution for this. --Tordanik 15:28, 31 August 2017 (UTC)

The descriptions of line markings are not suitable for overtaking

Overtaking and divider lines are different properties / issues. You may not cross certain divider lines, but it has no implications on overtaking as long as there is sufficient space and you do not have to cross the divider line. And when there are overtaking restrictions you may not overtake, regardless of the road markings. --Dieterdreist (talk) 10:51, 10 October 2019 (UTC)

That may be a regional distinction. In North America 4-wheeled vehicles are only allowed to overtake another motor vehicle if it is possible to change lanes, and only some States allow motorcycles or bicycles to share or split lanes for overtaking. In these locations the pavement markings signify whether or not passing (overtaking) is permitted. --Jeisenbe (talk) 11:50, 10 October 2019 (UTC)
quite possible, if this is about country specific regulations it should be marked as such, right now it suggests that certain road markings are generally the same as no-overtaking restrictions. --Dieterdreist (talk) 12:06, 10 October 2019 (UTC)

Default value

Since this page was first created, it has always stated, "Default value on a two lane road should be assumed to be yes." In the United States, this would be quite an unfortunate default value. overtaking=* is still pretty rare in the U.S., but virtually all striped roads in urban and suburban areas prohibit overtaking in either direction, while the relatively fewer striped roads in rural areas often allow overtaking, as long as there are no more than three lanes total. I don't think it would be prudent for a router to assume overtaking is allowed or disallowed in the absence of lanes=* or overtaking=*. It isn't as though routers currently need to account for overtaking being allowed anyways. As for renderers, users in the U.S. are more familiar with double yellow lines, so it would make for a better default representation of a road centerline than a dashed yellow line. – Minh Nguyễn 💬 20:14, 8 April 2021 (UTC)

I agree that the default should allow overtaking, at least in the areas which I know of, current tagging is asuming this, and overtaking=no is by large the majority: Just add overtaking=no to all roads where it is not allowed, and if this is the more common situation, I suggest you also add overtaking=yes where it is allowed. --Dieterdreist (talk) 21:56, 8 April 2021 (UTC)
To the contrary, in the U.S., your suggestion would result in the vast majority of roads needing overtaking=no, yet the vast majority of roads currently have no overtaking=* tag. It is true that overtaking=no is currently tagged on more ways, even in the U.S., but that's only a tiny fraction of the road ways in the database. It reflects more on where mappers have bothered to use this tag than on the actual frequency of overtaking restrictions in the real world. – Minh Nguyễn 💬 14:54, 9 April 2021 (UTC)
Forget about defaults and add overtaking=no. Either the information is present or you will have to guess. The overtaking default is "yes", it has always been documented like this. By the way, taginfo US confirms that US mappers agree with the default for "yes", they have added very few "yes" values, just a tenth of the number they have added "no": Seriously, if you believe the default should be the other way round, it is very much suggested to add both values, according to the situation. Defaults only work when everybody agrees. --Dieterdreist (talk) 16:04, 9 April 2021 (UTC)
Not only has the tag always been documented as defaulting to yes, but it has also always been documented with an image from the U.S., where this default value would make much less sense. In any case, I'm not arguing that the default should change to no but rather that routers shouldn't assume a global default. Why does there need to be a global default for something that a router wouldn't normally incorporate into routing decisions? – Minh Nguyễn 💬 16:19, 9 April 2021 (UTC)
I see. Routing software providers will have to make a decision when data is missing. Here we only document how the data is structured. IMHO if you are providing routing software, you should encourage people to be explicit with the information you need. Defaults are not working well, in general, because they are inherently unreliable: you do not know whether the mapper did not care about the property or deliberately omitted it because someone has written in the wiki that it was the default.--Dieterdreist (talk) 16:29, 9 April 2021 (UTC)
Exactly. It seems like the user who added that sentence about the default was projecting their own country's norms onto tagging guidance (while unwittingly contradicting it with an image). But in order to make it true, tens of millions more ways in the U.S. would need to be tagged with overtaking=no. Ideally, all of them would have some overtaking=* tag, but it's going to be a long time before we get there (absent a very aggressive automated mass edit). In the meantime, I don't think it would disadvantage routers if we remove any mention about a default from the article. – Minh Nguyễn 💬 16:47, 9 April 2021 (UTC)
I support removing the default, or maybe we write explicitly that the likeliness of yes or no in absence of the tag will depend on the area. --Dieterdreist (talk) 23:17, 9 April 2021 (UTC)
Done in Special:Diff/2139395, thanks. – Minh Nguyễn 💬 01:16, 10 April 2021 (UTC)