Emergency lanes are generally a bit less wide than normal traffic lanes. I think whether a car fits is a good general criterion to distinguish between “full” and “narrow”. (In the Netherlands, the end of “full” is typically marked with a traffic sign.)
According to Wikipedia, minimum lane widths are generally 2.5m to 3.25m. (Maximum truck width is 2.6m.)
For comparison, the lane width of a well-built motorway is ~3.75m.
=narrow is a more “subjective” adjective, that depends on the perspective. I would at least use =substandard. As mentioned, it is better to treat =yes as a unspecified positive, and use =full or =standard. I should point there is advantage that some compatilbity can be maintained by directly using shoulder:*=, leaving shoulder= =left / =right / =both / =no for applications that already use it.
That being said, another term that should be considered is “marginal strip”. I see this official terminology as superior over =suibstandard. Still, question of verge= remains.
1 Like
The issue with this what to do with a 2.5m shoulder when the jurisdiction has 3~3.3m as the standard.
shoulder:width=2.5
2 Likes
The standard of what? In most areas most streets don’t have a shoulder. The most common shoulder width of secondary roads will generally be very different to that of trunks or motorways.
shoulder:width exists and gives a very clear information about the actual size of the shoulder. There’s no need for very vaguely defined tags everybody understands in a different way.
3 Likes
Alhough shoulder:width=* is also problematic, as it has no wiki documentation, and it is unknown whether it applies only to hard shoulder part (i.e. paved part of the shoulder), or also at soft shoulder part (i.e. unpaved part of the shoulder).
Which happens to be exactly the problem the shoulder=yes also has - as it is unknown if:
- width of the paved part is enough so car can stop safely there (and/or emergency services to ride on)
- width of the paved part is not enough, but width of paved+unpaved parts combined is enough so car can safely stop there
- even the combined width of paved+unpaved parts is not enough so car can safely stop there (but e.g. bicycles can easily fit on that hard part).
2 Likes
Do you have a better suggestion for the categorization then? Of course having the exact dimension is the best, but not everyone is going to measure it every time. Most will only see what vehicle size it fits. Requiring them to either add a measurement, or leave it as =yes doesn’t encourage improvement.
The same question remains after you add eg *:width=1.5. Should it be =yes, or something else? How do you factor in the verge=? shoulder:*=yes + narrow *:width= of =1 or =0.5 might be quite misleading, if not borderline trolling.
Besides, I was only responding in the context of the original thought. If it is up to me, I might suggest eg shoulder:*:size= =motorcycle, =motorcar, =hgv, etc.
1 Like
“yes” and “narrow” would be much better than the “standard” and “substandard” I commented on.
I might suggest eg
shoulder:*:size==motorcycle,=motorcar,=hgv, etc.
Then we need clear definitions of how wide these types of vehicles are, which seems close to impossible given that e.g. ‘motorcar’ includes everything between a 1.2m Kei car and a 2.5m SUV.
As others said, we shouldn’t redefine what ‘yes’ means (too many cases to check / retag), so how about:
- yes: there is any kind of shoulder
- full: there is a shoulder that can fit a full car (e.g. at least 2m wide)
- narrow: there is a shoulder that can be used by narrow vehicles or as extension of the main lane by slow vehicles
4 Likes
IMHO, surely not: would you use it as an emergency lane? Surely not.
100% inline with @mueschel
1 Like
I’d say there are three things in this image, if the hardly visible, solid white line is supposed to be there and hasn’t been officially removed.
There’s a ~ 1m hard shoulder with asphalt, a 1.5m soft shoulder with gravel and another 1m verge with grass. This is really one of the complicated examples that are hard to map with a few simple tags.
I’d personally count the gravel area as verge, but others will for sure disagree.
In OSM, that’s about here. Currently the last OSM mapper there thinks “shoulder=no”, which I can understand.
I’d go with “verge=both” for that, personally. I’m not sure what I’d call the gap between the white line at the left of the main driving lane and the (initially gravel) verge. It’s useful for cycling on**, but not a usable hard shoulder for cars, so shoulder=no (as currently tagged) makes sense I think.
** but if it’s anything like the roads near me the white line may be used as the “official edge of the road” and anything to the left not maintained, meaning a cyclist would have to occasionally hop onto the “real” road if using it; not great for safety reasons.
I don’t see much difference in the possibility of self-explanation and misinterpretation. I would think =full means fitting all vehicles including truck and bus. In turns, =narrow is not clear either. Narrower than what? Narrower than the high standard for heavy vehicles, thus fitting the widest car in the market, and conflicting with your =full ? Or narrower than a car, ie your =narrow for a critical difference in usability? But you include yet another variable in “narrow vehicles or as extension of the main lane by slow vehicles”. So is it the width of a narrow city car, k-car, tractor, ATV, quadricycle, tricycle, or motorcycle? While I’m not familiar with agricultural and forestry vehicles. I have the impression a tractor can be quite wide.
For comparison, the notion of parking_space=normal is widely used. That’s still not ideal.
I already said in the beginning =yes should be treated as an unspecified positive. For the variation in class, =motorcar could equally be defined as the widest car.
I think legislation doesn’t matter in this case. OSM can have its own definition of what constitutes a full-width shoulder. It is information on the ground, after all.
To all who write “why not just use shoulder:width”: This would mean you always need a measuring tape, i.e. tag it very precisely, just to record whether something is a “proper” shoulder or just a narrow strip.
It’s really a similar reason why maxheight=default exists.
1 Like
FYI I now removed the AddShoulder quest from StreetComplete until such time that shoulders, soft shoulders and verges can be tagged easily with a clearer definition.
4 Likes
That reads a bit like a stereotypically German answer, and I’m going to reply with a stereotypically British one:
“No, just like with road widths, some of us are perfectly happy to guess to the nearest meter”.
( 
, in case not obvious)
1 Like
ObXKCD !
(
Oh. Good thing I hadn’t already started writing rebuttal about https://wiki.openstreetmap.org/wiki/Key:width vs. https://wiki.openstreetmap.org/wiki/Key:est_width before I saw that disclaimer 
1 Like
Technically you do have to consider whether shoulder:*:est_width= will be used. It has been, on some cycleway:*= and some less sidewalk:*=.
(Oh now it’s mentioned)
It’s more than standards. A wider shoulder allows you to walk around if needed (though you shouldn’t stay on it), with some more comfort away from traffic. Also means it can be converted to a travel lane easily. Maybe something *=lane to approximate.
Actually I don’t, because I’ve checked and it isn’t (at least not anywhere that I’m worried about). Unfortunately I did spend time wondering about how to handle some of he more esoteric concoctions in this thread in lua before doing that.
(a bit of a diversion onto width)
Let’s pick a mapillary picture of somewhere I’m familiar with, roughly here in OSM.
I didn’t add width to that when I was last there, but it’s mostly about 3m wide with occasional wider sections. How accurate is it worth measuring it as? One bit might be 3.0m, the next 3.1m - is it really helpful to “accurately measure” each bit? I’d end up splitting that section of road into dozens, possibly hundreds, or pieces.