The latest thread (not required reading - in fact, probably don’t read it) on separated pavements has become overly long and risks succumbing to “last person standing” - instead of coming to a consensus we seem to have landed up with a few people going over the same points, with most community members either disengaging or not entering the conversation in the first place.
This isn’t healthy.
This thread is an attempt to move the conversation towards mapping out where we agree the most, as well as getting a wider range of responses from the UK community. To aid it I have created a list of statements - please check the statements you agree with (the poll is multiple-choice) and then submit the poll.
Hopefully this will give us a basis for building consensus on how to map pavements in the UK
Note: I landed up needing to use two polls for this, I’m hoping this will work fine be sure to scroll down to the second one after submitting the first one!
“I currently live in the UK”
“I don’t live in the UK but I have in the past”
“I don’t live in the UK and never have (though I might have visited)”
“Separated pavements have made some of my journeys harder”
“Separated pavements have made some of my journeys easier”
“Separated pavements have generally made journey time estimates worse”
“Separated pavements have generally made journey time estimates better”
“OSM should have a separated pavement way for every pavement in the UK”
“OSM should only have a separated pavement to represent complex geometry or harder-to-cross roads”
“Separated pavements are a good idea for other countries but not for the UK”
“Separated pavements were a mistake and we should revert to tags on the highway everywhere”
“Separated pavements should be tagged with a name=”
“Separated pavements should be tagged with a street:name=”
“Separated pavements should be tagged with a is_sidepath:of:name=”
“Separated pavements should not be tagged with names”
“It is okay to partially map an area with separated pavements”
“Separated pavements should not be mapped unless a user/group commits to filling out most of an area consistently”
“Aerial mapping separated pavements everywhere is okay without on-the-ground surveys”
“Aerial mapping should only be used where the pavement characteristics are obvious and clear”
“Aerial mapping should not be performed unless it is accompanied by on-the-ground surveys”
0voters
“Local mapping approaches should be respected and followed over national or global approaches”
“The UK should have a consistent approach to separated pavements, but this can differ from other countries”
“The UK should use the same approach as separated ‘sidewalks’ in other parts of the world”
“A separated pavement should never be deleted if the pavement exists”
“A poorly separated pavement (e.g. without sufficient connections/tags) can be deleted“
“An ‘unnecessary’ separated pavement (e.g. on a simple, minor road) can be deleted”
“Pavements which are frequently blocked (e.g. with cars or bins) should not be mapped separately”
“Pavements which are narrower than can be used by a wheelchair should not be mapped separately”
“A line painted on the road to mark a pedestrian ‘lane’ should not be mapped separately”
“The sides of ‘shared spaces’ should not be mapped separately”
“Footpaths connecting to separated pavements should always have ‘crossings’ connecting out to the highway”
“Footpaths connecting to separated pavements should only connect to the highway if there is a crossing”
“Organised and/or paid-for additions of separated pavements have generally made the map worse”
“Organised and/or paid-for additions of separated pavements have generally made the map better”
“It is more important that we find consensus”
“It is more important that we pick the best approach”
0voters
Thank you for participating. Please keep the discourse civil
Thank you for trying to get some consensus going out of that ridiculous thread, but I feel like the methodology is wrong here. Not everyone is willing to read all 300+ posts.
Personally, some of these statements are quite loaded (like “Separated pavements were a mistake and we should revert to tags on the highway everywhere” and “A poorly separated pavement (e.g. without sufficient connections/tags) can be deleted”) - pavements being a “mistake” and “poorly separated” pavements are both subjective, there are differing levels of quality throughout the country and no two people map exactly the same.
The same goes for statements lacking context (“The UK should use the same approach as separated ‘sidewalks’ in other parts of the world”) - what is this supposed to mean for people who haven’t followed the thread in its entirety and/or aren’t clued up on what these approaches are?
perhaps these statements would’ve been better as yes/no questions but i’m not sure if the poll thing allows for it
The options for making journey times harder/easier/longet/wrong/wild are not helpful. Rather than see how people subjectively vote, if like to see a list of routers and if they make use of separated pavements (maybe with a count of reported issues, to indicate how well they do it).
Where is the option for “I do not believe a community poll is a good representation of OSM mappers, let alone a good representation of OSM users”?
I don’t think people need to read that other thread. In fact they shouldn’t. I’ll edit my post to say that
Yeh, I tried. Partly they are just quite subjective and controversial (you picked one I agree with and one I disagree with there, so at least balanced-loaded?), and partly it’s just - as you noted - insufficient polling support. It is meant to be the basis for consensus-building, mind. Hopefully off the back of this we could start building up more substantive positions which people could then coalesce around.
FWIW I have come to abhor that other thread (and that includes some of my own participation in it) but I do think we need to come together on an approach. If there is overwhelming consensus for something I don’t like then I’d prefer to implement that (as I do so many other things in OSM..).
The point is to capture what people’s subjective experience is, because I suspect that’s motivating a lot of opinions. Looking to myself, I engaged with this topic because it was making my experience of OSM worse.
The thing to take away from this would be “these opinions seem to be related to this experience”, which - yes - could be solved by better routing.
Another idea I’ve had was to start building up “verified routes” which could be used as sort-of a monitoring tool for identifying when edits have degraded routing. I’m sure this has been done before (in fact I think that’s exactly what’s done with public transport) but I haven’t seen it with pedestrian routes.
I think the point about a group committing to map a set of features to a particular method for an area prior to mapping to be deliberately stifling of growth of OSM.
Is that approach to be extended to all further mapping or just pavements?
OSM has always grown by iteration - someone adds the first version of something, and other people improve on that.
However, sometimes “the first version of something” is so poorly done that it is actually easier to delete and redraw than it is to enhance. It’s not just a “separate sidewalks” thing; a similar question was recently asked about poor quality buildings (the results of the poll there were slightly in favour of “It’s ok to delete poor quality AI buildings without replacing them immediately”, but it was a fairly even split).
An extra challenge with separate sidewalks is that (unlike with most OSM features) the usability of the data goes down when the job is only half done (put better here by @SK53 - “Unlike virtually every other #OpenStreetMap feature it is not resilient when incomplete.”).
Personally, I’d always try and extend rather than redraw, and preserve history rather than not, but sometimes it really just isn’t possible - what had been added was just so bad that it just needed to be deleted.
Not participating in the poll. Some observations: Somebody in my home-town started mapping separate sidewalks (in a fullt-time-job manner). Lots of times “upgrading” from sidewalks as mere attributes to a highway, which has been a long established method here. I questioned that. Got told: Sidewalk mapping here is abdominal. My mappings will aid OSM-based mutli-modal routers which are booming right now.
As a matter of fact, I am not aware of these routers. A question to name them went unanswered. All I know is, all the PT apps that people to my knowledge actually use not based on OSM (it is not needed at all) and the PT circum surroundings maps at stations (where OSM might prove useful) all try to abstract/generalise as much as possible.
Separate sidewalks are en-vogue. There are people with enough time to spare to map them. From my point of view, they might even be useful for some. I doubt they will make openstreetmap the next big thing, as what they are advertised. So welcome to a niche market.
PS: From what I learned with visually impaired people, separate pavements should mention the name of the street.
A general concern I have is that sidewalk mapping tends to be project driven - often a university project or some company wanting to show off some pedestrian-related software in a prime location. These projects have a limited lifetime; when they are at an end, people will lose interest and the remaining community is expected to maintain the sidewalk network they never asked for (and would never have added themselves). It’s the same problem as with imports - they boost the presence of something in OSM but it’s not sustainable. This concern does not apply to a hobby mapper adding sidewalks in the place where they live as they are likely to maintain that sidewalk network themselves.
I honestly don’t know how to vote in a lot of this for a couple of reasons:
Mapping coverage in the UK is far better than it is in the United States
Roads are very different in the United States than the UK
If I were to try to be as tag granular as seems to be needed in the UK I would never get anything mapped in the areas I am working on (which are all in the United States)
Well it is aimed at the UK, so pick them in the context of the UK. But so long as you tick the appropriate “living in the UK” statement we will be able to see how opinions align between the groups.
I understand the iterative nature of the development of OSM over time. Perhaps it is inbuilt to the iterative development that decisions made 15-20 years ago may turn out to be problematic when it comes to later iterative change. Add to that the common human nature to resist change and we have some FUD about explicitly mapping features that may be more generally implied.
I have seen various comments to the effect that sidewalk routing can be calculated from tags on the highways. Can these examples be demonstrated visually, showing which side of the road is walkable, crossings etc? I’m very much a visual user of maps, has anyone taken the highway based sidewalk tags and presented a map with those sidewalks drawn and appropriately connected with crossings etc? I realise such a presentation of implied sidewalks will have collisions with explicit sidewalks. It would be interesting to see how good well mapped implied sidewalks compare with well mapped explicit sidewalks. If anyone can define and show ‘well mapped’ for both.
I don’t think @SomeoneElse is spreading FUD. They acknowledged that this would be considered an exception and shared why they think so, referencing someone else making the same point.
I included the statement because I’ve seen it expressed (approximately), including in this SOTM Europe talk - which I don’t think comes down on the side of not mapping separated pavements, so I wouldn’t consider it a “partisan” statement either. The point seems to be “if you’re going to do it, have a high-level plan to complete it” rather than “don’t iterate”.
(That talk also explores adding separated pavements in post-processing based on tags, which is interesting, although - again - I don’t think the speaker is presenting it as a reason not to separate out pavements in the UK or anywhere else)
The results so far are interesting and definitely underline a “this is controversial” and “there is not even slight consensus” take on the subject.
For example, ten people agree with the statement that “OSM should have a separated pavement way for every pavement in the UK”, while twelve people agree with the statement that “OSM should only have a separated pavement to represent complex geometry or harder-to-cross roads”. Ten people participated but didn’t pick either option - one agreed with the statement “Separated pavements were a mistake and we should revert to tags on the highway everywhere” but the other 9 presumably felt there were nuances between these three options which weren’t represented in the survey.
I have so far presumed that the US community is far more united on this (since it’s the position of their Pedestrian Working Group) but I’ve been second-guessing my instincts on that and I wonder how this compares across the pond.
Anyway, I will let this run a bit longer because I suspect there are more people who will pop on here over the course of the weekend and find the thread. My feelings right now are that there must be something better than the approaches proposed so far (my own included) and that if someone can thread the needle of all (or at least most) of the issues/concerns/perspectives/etc then the way forward will become obvious and we can all just do that instead.
I’m really struggling to make up my mind on this. I have mapped separated pavements in the past, and found there’s lots of nuances that are hard to convey:
Places where a non-disabled person would definitely cross the road, but there is no formal crossing infrastructure (dropped kerbs, tactile paving). Mapping these requires making up crossings that goes against the “map what’s on the ground” principle.
Dropped kerbs that are meant for driveways, so in order for it to work properly, every driveway should be mapped too.
Long roads with no formal crossings, but quiet enough that a non-disabled person could cross anywhere they wanted.
I also think that armchair-mapping separate sidewalks is impossible and irresponsible, as leaving it half-done makes the data so much worse.
In this talk he talks about generating edges from sidewalk:left/right data and using highway=crossing nodes to allow crossing to the other edge of the road. It seems like the ideal solution and could be extended to wheelchair routing given the right tags.
I do believe there is a compromise which can reached on this. I do not believe that mapping all of the sidewalks separately is feasible, but I think there are some cases where separately mapping sidewalks is important.
My view at the moment is sidewalks should be separately mapped if the sidewalks cannot be accurately represented via sidewalk:left/right/both tags and extra information (sidewalk:left:surface, sidewalk:left:wheelchair). I think that adding separate sidewalks everywhere makes the data significantly worse, unless it’s done perfectly, in which case it’s just slightly better for most use cases.
Not necessarily, if you extend the rule to “map what’s observable/verifiable by a mapper on the ground”. If there’s a place where people would genuinely choose to cross the road on a certain journey, then that could (in theory) be observed by someone willing to wait around long enough. In practice, it’s often reasonably easy to spot the obvious routes that would be taken by people heading from one direction to another.
Not mentioned anywhere here, but important, is that any solution must degrade gracefully:
a) Anyone who wants to render a map without roadside pavements should be able to;
b) Anyone who wants to exclude roadside pavements from a routing graph should be able to.
b) is harder than it looks, because small sections of pavement often provide connectivity between non-pavement ways (for example, the junction between a cycleway and the carriageway). When these aren’t tagged correctly, routing breaks.
Can we please get an explanation for this viewpoint?? With good imagery, it’s super easy to add pavements and crossings! The only thing you can’t add is the height of kerbs, which need to get streetcompleted later (and can’t be streetcompleted unless they’ve been mapped from home first.
Unfortunately, I don’t think it’s possible to exclude all sidewalks from the data and then create a routing graph. There’s too many situations where a footpath joins a sidewalk only and does not connect to the road (for example because there is a guardrail there. The only possible fix would be to add virtual links passing through these barriers to connect to the main carriageway, but I hardly think that people would support such a solution.
And once again I must restate my opinion - It is not possible to create accurate and detailed wheelchair routing without separately mapped sidewalks, crossings and kerbs. (Well maybe you could but it would be horrendously complicated with loads of sidewalk tags on the roads)
I’d also like to see a follow-up question, for those that voted “OSM should only have a separated pavement to represent complex geometry or harder-to-cross roads”, what is their position on already mapped, detailed separated sidewalks - should we keep them as-is or delete them? Because many mappers have spent hundreds of hours mapping sidewalks with care, including adding details using streetcomplete, I doubt they’d look favourably at a consensus that advised against continuing that mapping or even outright recommended removing it.
Can we try and keep this thread to trying to understand perspectives and look for opportunies to build consensus. We can create another thread for debating if we want that again.
be sure to scroll down to the second one after submitting the first one!
