I’m sorry to revive an old topic, but I came across a situation that the current wiki page does not seem to cover, and I’d like your opinion on it so hopefully it can be clarified on the wiki:
(It is similar to Case 3 listed by @willkmis in this post of the original discussion, but I couldn’t notice a consensus in the discussion in response to that particular case.)
In cases like the above, where the sidewalk continues seamlessly into the driveway, with the same surface, it seems that neither crossing:continuous=yes nor crossing:continuous=no applies — or, in other words, neither the sidewalk, nor the driveway is more prominent in the crossing. In fact, it appears as if there isn’t a crossing at all, but instead more of a coexistence zone?
It should be mapped as an untagged node that is part of the driveway and part of the sidewalk. This communicates that the driveway and the sidewalk intersect. There are no crossing markings or crossing signals, so no need to tag the node as a crossing.
Thanks all for the recommendations! So if I understand correctly, the three possible situations are as follows:
road (e.g. driveway) is continuous, path (e.g. sidewalk) is interrupted: → highway=crossing + crossing:continuous=no
road is interrupted, path is continuous → highway=crossing + crossing:continuous=yes
both road and path are continuous, i.e. the same surface extends seamlessly along both ways → bare node (no highway=crossing and no crossing:continuous=* tag)
Does this accurately represent your recommendations?
Yes, with the caveat that I wouldn’t split hairs about the surface of the driveway. Many driveways have a driveway apron connecting it to the street, whether or not there’s a sidewalk. So when there is a sidewalk, it may very well meet the driveway at the point where the driveway’s asphalt or dirt surface transitions to the concrete surface of a driveway apron. A sidewalk meeting a driveway is already so insignificant that I don’t bother mapping a crossing unless there’s something exceptional like a curb cut or painted stop line.
I’m not sure I follow what you’re trying to tell me (or rather, why you’re telling me that). In what sense would the situation you describe interfere with the three categories I listed? Perhaps my description of them as “the three possible situations” made it seem like I was aiming for an exhaustive listing, and in retrospect it was indeed a poor choice of words.
Would make sense to include the third situation I mentioned (with continuous surface along both ways) as merely one of several cases where a crossing node is unwarranted?
A crossing (highway=crossing) is a place where pedestrians (or cyclists) can cross the road. When a sidewalk meets a driveway, that’s not a pedestrian crossing because there isn’t really a road there that they need to cross - it’s just a driveway. Same for the entrances to car parks, etc.
When there’s a crossing, it can be further described with tags like crossing:continuous and crossing:markings, but these tags don’t apply when there isn’t a crossing.
The more interesting case in my view is what do we tag when a sidewalk is interrupted by something as minor as a driveway, e.g. there are hard kerbs on each side. Does that make it a crossing?
Yes, depending on the expected traffic or perhaps a history of incidents, a driveway could have some the same characteristics as a bone fide street crossing: a surface change, curb cuts, markings, a stop line, or even a sign. It’s a gray area, but I’d err on the side of tagging a crossing if there’s something exceptional about it.
(This query is pretty handy for finding these cases in areas where crossings are reliably mapped as ways. However, it also turns up lots of false positives with inaccurate service=driveway tags and unnecessary crossing tags.)
Yes, this is very common here in Portugal, especially in older constructions. And even when the sidewalk continues over a driveway, it’s common for the sidewalk to be lowered to the level of the street, the rather than for the driveway to be elevated to the sidewalk’s level. That’s why for us a deliberately continuous sidewalk is worth mapping, even when it is crossing a mere driveway.
Yes, depending on the expected traffic or perhaps a history of incidents, a driveway could have some the same characteristics as a bone fide street crossing: a surface change, curb cuts, markings, a stop line, or even a sign.
Also cycleways can be interrupted by crossing driveways, even with (officially set up but nonconforming) stop signs for the cyclists. The situation doesn’t feel like crossing a road on the cycleway, more like the motorists are crossing the cycleway, but the signs are there and one could maybe opt for tagging a crossing
I strongly believe that sidewalk-driveway intersections should be mapped as crossings.
Especially in the case of business “driveways” that are often very high-traffic roads, pedestrians and motorists need to be aware of the fact that their paths are crossing at that location.
Seattle’s DOT has a markings pattern just for cycle path-service road crossings (very prominent painted green rectangle along entire crossing width + cyclist pictograms) which absolutely qualifies as “mappable” infrastructure.
Even for the most minor intersections of sidewalks and service roads, the continuous sidewalk crossings of single-house residential driveways, any user whether pedestrian, cyclist, or motorist needs to be more aware of the potential for dangerous intersections with others which are much more likely there than while traveling along the sidewalk or road.
It’s fine for any given mapper to decide they don’t want to map minor crossings, but these crossings do exist and are locations that can benefit from being tagged in detail if one wants to.
As a pedestrian, I do want routers to be able to take into consideration that it’s safer and faster to walk along a sidewalk with no driveways than one saturated with many driveway crossings.
So if crossing:continuous alone isn’t an appropriate tag to indicate this, what is?
If there’s a road marking or any other infrastructure specific to the crossing, I don’t think anyone would oppose mapping a crossing there. It doesn’t even need to be as prominent as the ones you’re describing.
To elaborate on the comments I made in this wiki talk page discussion: Indiscriminately tagging highway=crossing on driveway–sidewalk connections would dilute the meaning of the tag. If the routing profile intends to penalize a sidewalk that cuts across many driveways, it can already penalize every connection to a highway=* way, regardless of any tagging.
Some applications such as OsmAnd automatically inform the user as they’re about to cross a highway=crossing. But many users don’t want or need to be alerted to every possible point of conflict with vehicular traffic. Instead, they may require an affirmative instruction to “cross the street”, but only at a more substantial crossing (what traffic engineers would call a “designated crossing”). How would a router serve these users if there’s no distinction in the database? crossing:continuous=* doesn’t quite capture this distinction, as it has more to do with prominence or visual appearance than right of way, at least superficially.
Feel free to add highway=crossing to your list of “skunked tags” then because “Don’t tag nodes at intersections of sidewalks and driveways as crossings” is already an entirely lost battle.
Try looking around with Overpass and a query like this:
[bbox:{{bbox}}];
way["highway"="footway"]->.footwayWays;
node(w.footwayWays)->.footwayNodes;
way["service"="driveway"]->.drivewayWays;
node(w.drivewayWays)->.drivewayNodes;
node["highway"="crossing"].footwayNodes.drivewayNodes;
out skel;
(Or this which seems to perform better on larger search areas)
That’s >50k examples in just a part of the US alone and there are many such cases in basically every part of the world that I checked.
So, keeping in mind that these sidewalk-driveway intersections are tagged as crossings, if you want to avoid the original (?) meaning of crossing:continuous=yes being “diluted” then what tag do you suggest?
If anything, iteratively refining the tags of these crossings (whether with crossing:continuous=yes or another tag) would serve to do the opposite of diluting the meaning of a highway=crossing, since a specific tag communicating the fact of the matter of “this is a relatively minor crossing” would help data consumers to filter these out.
The article on skunked tags makes the point that sometimes a tag’s original meaning can be salvaged. I guess I’m more optimistic about that than you are.
What you’re observing is the effect of iD giving mappers no way to connect a sidewalk to a driveway without tagging the intersection node with highway=crossing. We fixed that recently, but to my knowledge, there hasn’t been a concerted effort to clean up the iD-generated crossings. (railway=level_crossing has the same problem in cities that have streetcar lines.)
This is a job for a MapRoulette challenge. We have the technology and the knowhow. We don’t need to commandeer a second tag to serve as a troll tag that says highway=crossing shouldn’t have been added in the first place.
By the way, I’d filter out the crossings that have crossing=*, crossing:markings=*, or crossing:signals=* on them. This more likely indicates a crossing that should be tagged as a crossing. I’ve apparently mapped many such crossings myself.
For context, there are 375,234 driveway–sidewalk connections in the United States. Of those, 122,302 are tagged highway=crossing; 106,592 have only highway=crossing without some other crossing=* or crossing:*=* tag that would indicate something out of the ordinary. I’d say 28% is significant but not insurmountable. I just reviewed about a hundred in one sitting. Moreover, this misapplication of highway=crossing has likely subsided since last July, when the bug fix in iD was released.
I think calling this a troll tag is hash and not really improving the discussion. They have a broader definition of crossing than you and are adding a tag to help clarify between their definition and yours. That’s not a troll tag. It is a tag made in good faith with a reasoned explanation for it’s proper use. You don’t agree with that, that’s fine but calling it a troll tag isn’t really appropriate.
