To that, I’ve already added my recommendation of best benefit/effort ratio: just tag the whole way as highway=steps and call it a day, concentrating your saved time on things where your invested effort is likely to produce more significant improvements.
“Do people like eating oysters today?” Such general question is impossible to answer.
But yes, some do support it today[1]. Some even explicitly document that they do. To which extent each of them support barrier=*, and how much attention they pay to the actual value of barrier key, depends heavily on specific data consumer you have in mind[2] – and you would probably get more precise result by asking on their specific communication forum (or looking at the source code yourself).
But, not to leave you completely dry, here is one example: in this default Brouter gravel bike profile, barrier=step would get (maximum/default) penalty of 139 (whereas, if it were barrier=cycle_barrier instead, it would get penalty of mere 87).[3]
(Now, your wheelchair users probably would not be using gravel bike router profile, but some other profile, even better tailored to their specific needs, but it is not a worst profile to start from).
I’m relieved we agree there!
Yes. That is how OSM (and world in general) works. You have incomplete data, and you have to make assumptions, and sometimes assumptions work, and sometimes they don’t, and then you update assumptions so they might hopefully work better next time, and update the knowledge base (i.e. OSM data) with known verified data, so there needs to be less assumptions. Rinse, repeat.
If someone thinks that this process will ever end and all data will be perfectly mapped and we’d all live happily everafter: well, I admire their optimism, but cannot share it.[4]
And also, obviously different data consumers make different assumptions, so you’d pick one which works best for you.
And if you’re picky and willing to define exactly how picky you are, I’d recommend Brouter, with profile tailored exactly to your specification. Also e.g. OsmAnd knows how to use Brouter (if you’re looking for complete navigation solution and not just routing engine itself).
Once made, you can even share that profile to other users with similar needs!
There are whole another threads with hundreds of messages about that (and thus, I’d agree with you that we should prefer sticking to original point about those sidewalks/steps in Brazil here, lest this discussion becomes unreadable); so I’ll just quickly summarize that highway=steps is generally considered bad choice even for more offroad 4WD vehicles then regular sports car or sedan.
On highway=steps, only reasonable interpretation (IMHO) is that smoothness=* tag defines smoothness of each step[5] (alternative interpretation would be nonsensical, as there is no point tagging smoothness=impassable on all highway=steps ways in the world – if one subscribes to that POV, better not to tag smoothness at all in such cases).
and, conversely, some don’t – you’d should be contacting them if you see benefit in them implementing support for that ↩︎
Surely you’re not suggesting people should find all thousands of OSM data consumers in the world, analyze them, and serve you the nicely formatted result; right? ↩︎
Of course, you can easily add extra line
switch barrier=step 931if that strikes your fancy. ↩︎At best – or at worst? – at some point diminishing returns will make mappers bored and they’ll stop mapping. Otherwise, our OCD will continue driving us to ever asymptotically try to approach that ideal of “no assumptions needed, everything is clearly defined” ↩︎
i.e. exactly the thing you excluded in your quote ↩︎