But admittedly I would get rid of this primary_link completely here, it is just road starting/ending here. But it seems to me that it would be incorrect. Maybe?
As mentioned _link is on of the most confusing ways for me in OSM tagging, and usually I just ignore it. But here current tagging seems to be really weird and I keep seeing it.
I heard “if it is named then it surely is not highway=track or highway=service” before which maybe works well as heuristic in some areas but not here. We have official road names for not-yet-existing roads, for unpaved forest tracks, motorways, for roads which existed but will never exist again, service roads, driveways and more.
I suspect that the same works poorly for links. Though I would love to have reason to purge them.
But you’re expecting a universal rule as to whether _link is supposed to be used in such weird way?
A rule for all of OSM will necessarily be very very general and have lots of exceptions. I gave you one. Feel free to come up with a better one if you want…
From my American perspective, the whole eastbound roadway south of Bora-Komorowskiego Street resembles a freeway or expressway off-ramp, from this unsigned exit all the way to the entrance back onto Bora-Komorowskiego Street. I often encounter such ramps in a partially built interchange or as a relic from an earlier era. But Bora-Komorowskiego Street isn’t tagged expressway=yes, so I don’t know if this exit design is more typical of a conventional road.
This is a good rule of thumb, but by the same token the way could remain highway=primary_link and the name=* could be replaced by destination=* on that basis.
There are a few things at play here and I think the main ones are access rules and road classification.
It is possible that pedestrians and certain categories of vehicle are not permitted on the main road (or that category of main road), except at designated places like the bus stops and traffic lights. However, those classes are allowed use the bridge to come and go from the aviation museum.
Situations like this seem to occur in locations constrained by geography or economics that prevent the ideal solution being constructed.
Here: we change from highway=secondary to highway=trunk_link without a junction, so as to allow access to the adjacent private land. Node: 270205274 | OpenStreetMap
Construction of this highway=trunk_link needed some houses to be demolished, so the residents were allowed build new houses on the highway=trunk_link: Way: 923747267 | OpenStreetMap
looking at primary_link, 9% have a name: highway=primary_link | Tags | OpenStreetMap Taginfo
compared to 74,5% for primary.
For secondary it’s the same, and tertiary is similar but slightly more balanced (10% and 59%). So yes, there is a trend, but it is not sufficient to use it as a simple rule because there are still too many to think of them as outliers.
Admittedly the no-name rule holds up a lot better in North America, where we lack a true concept of link roads and mostly only use highway=*_link for ramps and slip lanes.[1] As I understand it, some European countries have a more formal concept of a link road, even with a different style of route marker when it’s part of a route, so applying the same tag for anything resembling a ramp might be dicier.
name=* is on only 3% of highway=primary_link versus 90% of highway=primary in the U.S., 4% versus 97% in Canada. Most of it is probably because of a controversial practice by a commercial mapping team that we don’t need to get into here. ↩︎
2 Likes
Mammi71
(One feature, Six mappers and still More ways to map it)
10
Thank you for the premature praise. I have no particular expertise outside Germany. That’s why I can only give a very general answer.
In urban mapping, the distinction between _link and non-_link is very often very difficult. Planners never think about the future or about us mappers.
Some structural features, e.g. the ramps, suggest primary_link. But then the whole of Stefana Jurczaka should be a primary_link.
However, the surrounding area, neighbouring junctions, and the entrances and exits to the petrol station and shopping centre tend to suggest that this should not be treated as a classic exit/entrance to a primary.
Given that it hasn’t been a primary_link for many years and the user is new, I am cautiously inclined to revert these changes.
I do consider these rules of thumb to be a reasonable first approach. Of course, this isn’t an absolute rule, nor does it apply everywhere and always. But it can be one of many criteria for an initial classification.
2 Likes
Mammi71
(One feature, Six mappers and still More ways to map it)
11
If you look at the data and available street views:
Many of them are short connections between two carriageways with no street name signs—people simply assume they share the name of the carriageways.
Many *_link are simply right-turn lanes, and the “name” tag is being misused; it should actually be “destination.”
around here, the typical situation is that these oneway links near intersections (or linking in and out of roundabouts) are seen as part of the street that they are the continuation of, and the street signs (names) are typically put at their start.
A better example is the RIRO junctions along expressways in the San Francisco Bay Area. These glorified slip lanes are signposted just like freeway exit ramps, replete with advisory exit speed limits. Based on the signage and geometry, we tag them as unnamed link ways. The name and non-link tagging continues just after the one-ways meet, where the ordinary speed limit and street parking regulations begin.
Sometimes these RIROs lead to something probably resembling a European link road, which we therefore have no idea how to tag. Ryder Street has toggled back and forth between link and non-link classifications five times, even though everyone agrees that it has a name. You might recognize this junction design from an even worse conundrum a couple months back.
There’s really no escaping the visual awkwardness of a thin link way leading to a much fatter street in OSM Carto.
From what I’ve seen, the flares around a roundabout normally aren’t tagged as link ways (other than when the roundabout exit leads to a highway on-ramp).
In my opinion every grade-separated junction should be tagged as link to the first junction/driveway/gate etc. In sometimes looks weird in cities, but it’s probably the easiest way to do it.
Sometimes accesses are huge help with links, as often they have similiar accesses to the main road they link to, but in this particular situation it wouldn’t work.
Mammi71
(One feature, Six mappers and still More ways to map it)
15
The problem is that, particularly in urban areas, it is sometimes difficult to tell whether it is a grade-separated junction or just a standard junction where a road joins a carriageway, like any ordinary T-junction, perhaps coincidentally situated near a bridge …
I therefore prefer not to have a rigid rule, but to use common sense in each individual case.
By the way, is it actually standard practice in Poland to always use the higher classification of the connected roads for *_link, as per the table shown there? So, does this also apply to every separate right-turn lane (Triangle)? Is/was this also the case on conventional printed road maps? Pl:Highway link – OpenStreetMap Wiki
It is no coincidence that this table was removed from the English-language main page regarding highway=*_link or replaced by this one in the background: Link roads between different highways types – OpenStreetMap Wiki
That was always my least liked thing about _link tagging and that is why I tend to pretend they do not exist. Luckily, I almost never need to map new larger roads.
For actually standard practice in Poland for road classes - it is a mess as we have some prolific accounts, often “new” accounts, that keep mass changing road classifications. Often according to rules they personally invented.
Some even managed to invent new weird road classifications rules that go against their own weird road classifications rules from few months ago.
_link appears to be recent favourite of some accounts, that is why I asked on wider forum
Yeah, _link road tagging in OSM can definitely be confusing, especially in cases like this where it’s not a clear motorway/primary interchange.
From what I understand, a _link tag is usually meant for short connector roads between main classified roads (like motorway_link, trunk_link, etc.), mainly to show entry/exit ramps or slip roads that serve that function.
In your example, if the road is just acting as a local connector or standard access road rather than a true ramp between higher-class roads, then tagging the whole segment as _link might not be appropriate. It would make more sense to only use link tags where there is a real directional ramp function, and keep the rest as regular road classification.
So your confusion is valid this looks more like inconsistent tagging than a strict link use case.
Well, I should have written “most of” instead of “every”. That one is my fault
But wouldn’t most of grade-separated junction near bridges still be links?
The problem with T-junctions is, I think, where link changes into “normal” road. Whole discussing on Poland, I did notice that people have different opinion on that.
Polish Wiki page is, well, problematic. It’s mostly written by one user - I’m not sure on what basis this was done, but we noticed recently that a lot of people disagree with it.
Weirdly enough, English Wiki uses examples from Poland to show how to use different link classifications…
Mammi71
(One feature, Six mappers and still More ways to map it)
19
Well, the table currently linked on the Polish Wiki was added at the same time it was added (without discussion) to the English Wiki. And this table was the subject of controversy for a long time, until it was removed from the Wiki again because this recommendation was incorrect for many, and probably most, regions. Perhaps the Polish Wiki is simply no longer up to date.
The new table shows both sides, and in my opinion, the left-hand side – classification according to the lower category – seems to be the better choice in most cases. Important: please note the exceptions at the end of the table.
I’m staying out of the discussion about road classifications in Poland as a whole. I stopped following that at some point. I would simply prefer that no one dictates to mappers in the neighbouring country which classification to use, simply because two different classifications meet at the national border.
The original question, however, was whether such types of junctions with higher-level roads should always be *_link. Well, I cannot give a completely clear-cut answer to that. Personally, I would recommend using them sparingly.
