In changeset #155281608, a number of separately mapped sidewalks were deleted on the spurious ground that they’re “fictitious”.
There are a lot of pointless and largely decorative separate sidewalks along residential streets, particularly long Victorian terraces, which have no defined crossing points and are unhelpful for pedestrian routing.
On the other hand, separate sidewalks on main roads (highway=tertiary and above) which do have defined crossing points, are generally useful for pedestrian routing. In addition to this, they allow accessibility information to be captured at crossing points with more detail than simply tagging the crossing node allows.
Could someone point me toward the discussion where separate sidewalks were deprecated?
Mapping sidewalks as separate ways unfortunately seems to provoke a lot of hostility from some people. I have encountered it myself. I would not be surprised if someone unilaterally decided to delete them.
Mapping sidewalks as separate ways is a type of mapping which adds detail and information for accessibility, routing and other purposes. And it is not at all deprecated. If the footways were actually sidewalks and correctly tagged, that deletion in my opinion should be definetely reverted.
This mass-deletion should definitely be reverted. The sidewalks were correctly mapped as highway=footway + footway=sidewalk and the main road with sidewalk:both=separate.
As mentioned by others, the separately-mapped sidewalks are entirely valid, verifiable, and can provide great value to pedestrians due to the specificity of the information that can be provided, e.g. existence of lowered curbs and crossings. No idea why this user deleted them.
One change I would consider to the sidewalks is the removal of the name tag on them. I am unfamiliar with the UK’s local mapping guidelines, but typically in the US the sidewalks themselves do not have or inherit the names from the roads they border. I imagine it is similar across the pond, but of course feel free to ignore me if that isn’t the case
I don’t think there’s a consensus on that here in the UK, or more specifically London. I’ve tended to add them because it’s useful for routing software to be able to tell you which road’s sidewalk to use. It would be better if the name was not rendered when footway=sidewalk or cycleway=sidewalk is present, except perhaps at very high levels of zoom.
It depends very much “where in the world you are”. If you’re living somewhere where the cities were planned on a grid pattern by relatively recent invaders, there are nice wide roads, and there are traffic rules that restrict pedestrians from doing basic things like crossing the road, then separate sidewalks make perfect sense. If however the invaders’ town planning was upwards of 1500 years ago, then mapping of separate sidewalks would grossly misrepresent the level and type of pedestrian provision, even if done well. Also, too often separate sidewalk provision is mapped poorly, e.g. forgetting to add sidewalk=separatehere**.
This is one of those mapping decisions that really needs people who actually live there making the decisions. Here (in the same town as my example above, but created by more recent town planners) is an example where it might actually make sense to separate the footway from the cycleway (which is already separated from the road).
If they’re mapped properly with footway=sidewalk then renderers can remove that if they want (OSM Carto won’t, but that’s another issue). If they really do have a separate name (and I’m not convinced that most “roadside sidewalks” in the UK do) then it would make sense to record it.
** the last editor of the separate sidewalk there was me, so I have to accept some responsibility for that, but my excuse was that I was just fixing the bus stops round the corner
When I map separate sidewalks, the question I ask is “Is adding this useful for pedestrian routing and accessibility? If I use it to plan a run, will it help?” That usually requires defined crossing points (with at least something like a lowered kerb) close enough that routers won’t use an excessively circuitous route.
We have acquired a lot of what I refer to as “decorative sidewalks”. There are no or infrequent crossings, they are poorly connected to their parent streets and the sidewalk tags on the parent street are not updated. If these are to be micromapped, adding something like area:highway=sidewalk (undocumented) might be better. Frimley has a lot of these, but it’s too far from home for me to do anything about.
The sidewalks removed in #155281608 were mostly added by one user in 2022-2023, misusing a tasking manager. I spent several weeks fixing the useful sidewalks they added and removing many of the useless ones. I didn’t finish cleaning up in the area and would probably have deleted some of those on residential streets, but not on the secondary and tertiary highways. Given the amount of work involved (see OSMCha: #waymap-project-SB), I am more than a little peeved by this arbitrary deletion.
Just for info - here is a visualisation of the changeset that prompted this thread. I believe that the mapper of that lives very locally. One of the mappers who added those roadside sidewalks has been asked** to discuss edits with the community but has not replied to questions. The other is a Facebook editor from the US (with no indication that they’ve even visited London let alone being familiar with this part of it). It can be seen that this particular bit of mapping was of poor quality - it suggested that if you follow the sidewalk down Dagmar Gardens you can’t cross the road at the end, which fails a basic plausibility test.
** That’s a message sent by the DWG that the mapper has to read before continuing to edit. They’re then free to continue editing. For completeness, I’m a member of the DWG and although I’ve been involved in “sidewalk” issues before, I don’t think I was involved in this one.
In some areas around the world, such patches of sidewalk decorations stem from some sort of time-boxed project - often a university thesis or somesuch - where someone hacked together a “pedestrian router” that would absolutely depend on pedestrian-exclusive ways, instead of sensibly assuming that pedestrians can usually walk along minor roads even in the absence of an explicitly mapped sidewalk. After the project has ended, the data is left there to rot.
We also get sidewalks that just stop halfway down the street, because people make poor use of task managers and have a square clearly labelled with “Do not map outside this square” in MapWithAI. That might make sense if someone is tracing buildings, but for a routable object it should ask mappers to continue to the next intersection with another highway.
@SomeoneElse I was in discussion with @Fizzie41 on the DWG side over those edits - you should find it under Ticket#2023011710000075
Most of the issues with the sidewalks added by @alisonlung (of which there were a considerable number over an area of West and North West London). There are some which I missed and there are certainly sidewalks on residential streets which I should have deleted. If the changeset is reverted, most of those will go subsequently, but the ones on main roads will not. The edits by that user were particularly problematic because they had a habit of “accidentally” breaking cycle routing.
I’m not going to delete @VLD316’s additions without the basic courtesy of discussing it with them first. I might do something with the tagging and geometry of crossings where they’re connecting sidewalks along main roads.
I’m also a big fan of mapping sidewalks separately, doesn’t matter if it’s along big or small roads, as long as the sidewalks really exists (e.g. if there is a curb, even 1 cm high, separating the sidewalk from the roadbed). There are many clear cases e.g. on some living streets, where there really is no separation between where vehicular traffic and pedestrian traffic goes, where you should not map “fictitious sidewalks”.
It’s obviously a form of micro-mapping, and we can debate about which tags need to be added (I try to add footway=sidewalk/crossing, crossing=*, as well as sidewalk:both=separate and foot=use_sidepath on the adjacent road) . But if even if they are just tagged with highway=footway, that doesn’t mean anyone can just delete separately-mapped footways because they don’t like it.
It’s clear that it makes pedestrian routing more accurate, and I think they genuinely add useful context on the rendered map as well, such as showing on which side a single sidewalk is, where sidewalks sometimes end along a certain road and how/where they connect with crossings and other foot/cycle paths.
It could potentially make pedestrian routing better if it was done very carefully and thoroughly. At the very least, proper sidewalk way mapping requires drawing every single crossing possibility – not just the officially marked ones.
Of course, even if that best practice was followed, there would still be unsolved problems (no “follow the left sidewalk of Brühlstraße” instructions with separately mapped sidewalks, for instance) which stem from the fact that there is no machine-readable relationship between a sidewalk way and the road it belongs to. But that can only be resolved by establishing new tagging standards, not by individual mappers.
One of the most frequent pedestrian routing instructions is "Somewhere
in the next 200m cross the road when it is safe to do so. If there is
any traffic I walk a bit further.
Assuming the pavements are correctly tagged with footway=sidewalk, the problem in @Tordanik’s example is arguably with the router rather than the data.
In the UK at least, it should probably be assumed that you can cross the road, or transition from the pavement to the road or the road to the pavement on the other side, at any point unless there’s a mapped barrier in the way or the road is a motorway/motorroad or has an explicitly tagged prohibition for pedestrians.
Admittedly, it would be quite resource intensive for a router to deal with this properly. Clearly making sure crossing points are fully mapped is a good thing. Perhaps there should be a tagging standard to help routers associate sidewalks with their corresponding streets, to make crossing decisions easier.
