It’s the latter. For info, the original mention was here and the bit that addressed your concern was a bit of an aside in a follow-up post in that thread. This overpass query is an example of what I’d miss (and where routing would break).
However, where remote “separately sidewalk mappers” have been active the biggest errors seem to be:
not mapping places where you can get from the road to the sidewalk (e.g. driveways)
not updating the road to say that there is a separate sidewalk.
It’s the second of those that has those biggest negative consequences if you omit separate sidewalks from a map if you assume (as I do) that I don’t want to walk along a trunk road without a sidewalk or a verge.
…Looking at the imagery I can see lots of driveways. In order to do a good job of “separate sidewalk” mapping you must map places where people can cross between the pavement and the road…
VlD318 mentions that the private driveways did not fit the use case for a traditional ‘crossing’ in their opinion.
Concerning Brownley Road, I acknowledge that there are several residential driveways in this area that I did not include as crossings…My understanding was that private driveways like these typically should not be tagged as crossings,
For clarity and consensus, what is the best way to map the sidewalk along Brownley Road and why? Should it not be mapped separately at all and just add the sidewalk presence tags to the road? Thank you!
I don’t have strong opinions on whether the sidewalks in that area should be mapped separately or not, but since continous crossings have come up, just a small clarification from me:
What is being discussed in this thread is that if you are drawing the sidewalk separately, then you should also map the places where there’s a dropped kerb, so that pedestrian and wheelchair routers knows where you can get from the road to the sidewalk. If you draw the driveway, then the router can make use of the little segment of driveway betwen the sidewalk and road.
You can do this without mapping the place where the sidewalk crosses over the driveway as a highway=crossing. Just leave the shared node untagged.
(I’m the author of the Wiki page on continuous crossings so I felt that I should clarify.)
In respect of other edits by VLD318, while I was queuing to get into a gig in Camden last night, I opened StreetComplete to pass the time. I noticed quite a few useless sidewalk stubs and missing connections. These were added at the start of September. If these are not going to be fixed in the very near future, can you give any reason why I shouldn’t make them remotely useful by truncating them to the nearest properly mapped junctions?
Even where crossings have been mapped, as at Greenland Road’s junction with Bayham Street (n13120352940), the lowered kerbs and tactile paving had been mapped in 2017. The newly added crossing way does not intersect with these, ignoring the accessibility information and presumably Rabid’s warnings about crossing barriers.
On the general point of “if I’m physically in an area is it OK if I tidy up the mapping there” I’d say that the answer was a resounding yes - it’s quite possible that if the mapping is wrong the person who added it has never been there, and is just guessing from imagery.
Perhaps we could do with a list of circumstances in which separate sidewalk segments can and should be deleted, rather than attempting to fix the mess?
Was it added by a remote user using a task manager?
Was the edit made more than a month ago?
Does the separate sidewalk ignore connections via crossings which are clearly visible in aerial imagery?
Does the separate sidewalk cross kerb ways without intersection nodes?
Does the separate sidewalk pass through buildings or barriers?
Do crossing ways end at the kerbs, or form perpendicular junctions with sidewalks (making it harder to capture accessibility features later on with tools like StreetComplete)?
Are there useless sidewalk stubs without a fixme to indicate at least the intent of continuation?
Are the sidewalk:$side tags on the parent highway incorrect?
We can use the ‘fixme’ going forward if there are any sidewalk stubs. Otherwise, sidewalks should be extended to the nearest properly mapped junction as you mentioned.
Hi @osmuser63783 thank you for your feedback. Please let me know if I’m interpreting this correctly. Here is a driveway that intersects a sidewalk. I’ve mapped the driveway with tags ‘highway’ = ‘service’ and ‘service’= ‘driveway.’ This is then connected to the main road. Where the sidewalk intersects the driveway, I’ve added a blank node. (This is not published on OSM. I just did a mockup to make sure I’m understanding this clearly.).
The second part to this is about imagery clarity and local knowledge (as @SomeoneElse has mentioned). If a mapper cannot map these features (driveway and sidewalk intersection) due to imagery or other forms of verification, then it should not be mapped separately. Is this what is being said? Thank you so much.
Following up with an example from @SomeoneElse:
Here is a sidewalk that intersects a driveway. The driveway is mapped and then a node is added. The mockup example I did is similar to this. Node for reference: Node History: 13271261383 | OpenStreetMap
For this footway, why does it not extend to the other side of the sidewalk? Also, why is it considered a footway vs a crossing? This will be helpful insight for me. Way History: 349337051 | OpenStreetMap
Because I haven’t had a look at that side of the road yet in detail, and its raining at the moment!
As I write this, this way was only added 19 hours ago. The mapping there is far from complete - the road is still mapped sidewalk=both, which it isn’t really. I’ve added some of the kerbs to the north, including those on “obvious access paths”, but there’s more to do. Pop back a week to see the “before” state and you’ll get an idea where this area is iterating from. I’m still not convinced that I’ve got the eastern side of that road correct yet, either. I’ve been past there a few times and what it looks like on the ground (what you see as your options about where to go as a pedestrian) are very different from how it looks on aerial imagery.
(back to how to get across the road by the school)
The node that you’ve indicated above isn’t part of a crossing, but is a way of getting to and from the road without going over a kerb. The one to the north is a crossing with actual crossing infrastructure (tactile_paving etc.), and the one to the north of that is again just a way of getting to and from the road without going over a kerb. With those three that close together obviously if a router is out by 3m or so no-one’s going to mind, but further south it would make a real routing difference (and no, the mapping there isn’t the finished article either - there are no access tags yet, and those double driveways are a bit “topologically rather than geometrically correct”).
More widely, the use of crossing tags on ways vs crossing tags on just the intersection nodes is a bit confusing. There are a wide variety of ways of representing pedestrian crossing infrastructure (both “actual crossings” and “just somewhere you can cross the road”). With regard to crossings, I’m still trying to work out “how people map them” before trying to make statements about “what is correct”.
Taking a step back, sections of the Highway Code say things like “Pedestrians may use any part of the road and use cycle tracks as well as the pavement, unless there are signs prohibiting pedestrians”, which may be a bit of a surprise to those from pedestrian-hostile jurisdictions. We need to try and ap all of it.
@SomeoneElse If someone just tags the road with sidewalk presence, how is this different from mapping a sidewalk separately without mapping the driveways that intersect them?
That’s an example from an unrelated thread about Birmingham, but it illustrates the problem nicely.
I tried to give a summary of all sides of the argument at the top of this thread. The problem with the mapping in that screenshot isn’t that those sidewalks do not exist, it’s that the part that they play in the network of available routes that someone can take isn’t represented - there’s no connectivity, which results in the silly routing instructions you see above. The problem isn’t “separate sidewalk mapping” itself, the problem is doing it badly.
