Florian Lohoff <[email protected]> writes: > On Sat, Apr 04, 2020 at 08:51:45PM -0400, Greg Troxel wrote: >> All that said, you are basically right in your complaint. As I see it, >> routers should allow a single transition from access=yes to >> access=private (but not back) for roads that are near the destination, >> basically assuming that if you ask for a route, you have permission. > > This is wrong - access=private means no router can determin whether > you as user may AT ALL use that road. So all routers drop roads > with access=private from their graph. No way to use it.
OsmAnd does not do this, from what I can tell from discussion, from experience, and reading routing.xml. If you turn on enable private access, it just allows private roads. But it gives them a high cost, so it typically will not use them as shortcuts. This is wrong, but usually gets an ok answer. I am not defending it on a theoretical basis. > This is IMHO the correct way of treating *=private and in accordance > with the wiki. The wiki is quite clear that =destination is only to be used for places where one has a right to go, but the road can only be used if going there. It specifically says: Note that "access only for residents" is private (for example vehicle=private when applied only to vehicles). private says "Only with individual permission" and indeed, my understanding of the roads in question is that you may use them if: 1) you own property in the trailer park, thereby having permission, or 2) you have received an invitation from someone in the park, and therefore have permission (or are delivering a package, sort of the same). To be a bit pedantic, let's assume I am engaging in visiting houses to talk to them about religion, or to ask people to sign a petition to get a candidate on the ballot. (In most of the US, these activities are exeempted from various rules prohibiting or regulating soliciting. I am merely taking that off the table.) I still cannot go into the trailer park to knock on doors without an invitation, because it says "No trespassing", not "No through traffic". If the way were access=destination, and I decided to go visit a particular house, then I could properly use the way. So according to the wiki, access=private really is correct for this situation. The problem here is that we have two tags and three concepts. They are: 1) you may not use this way unless you have permission, and you are basically not going to get it 2) you may use this way only with permission, and that permission is routinely granted to people who have permission to visit places reachable by the way 3) you way use this way only if you are going someplace that is reachable by the way -- but you do not need permission to decide to visit that place Case 1 and 2 are both access=private. This is what I was talking about with the main tourist road into Blenheim Palace, that you are suppposed to use if you are going there -- even though it's private and there is no right of access, vs the roads for staff only in the back that unless you are employed as a groundskeeper you will not get permission for. Case 3 is destination. You are arguing to use destination for point 2. > Even that is broken with most routers as they only increase > cost per distance on those ways which is wrong as it will still > use that roads as through roads if the alternative cost is high > enough. And there are even more artefacts by solving *=destination > by cost per distance. Agreed. > I fight against excessive use of *=private in Germany for a long time > but people are confused and mix up ownership, destination and > are pretty quick with issuing access=private on roads. I agree this is tricky, and a road being privately owned technically leads to private in most cases (in the English view), but usually that is silly. For example there is a shopping center in my town. it's private property, but the public is welcome to come shop. Technically you maybe sort of need permission (they could tell you to leave), but in practice everyone treats it like a public road. I have left the ways without accesss tags. If I wanted to be pedantic, I would use permissive, because there is no "no trespassing except for customers" sign, and I have never heard of anyone behaving reasonably being hassled for being there. (In England, signs prohibiting all sorts of things on private property seem much more common, mostly prohibiting parking, as from my travels England (and I mean England not UK) seems to have slightly more than one car per parking space in total and parking is almost always painful. I began to understand OSM origins better from this experience.) But the road we are discussing is not like the supermarket example. It has (we believe) a sign that says "residents only - no trepassing". Here, that sign makes it is a crime to enter without permission. > I am analysing data for parts of Germany for problems with that > by using the nearest api call with OSRM and listing OSM Addresses > with more than e.g. 100m to the next legal road. Sometimes the nearest > road is on the other side of the railway, river or whatever. Sure - I do realize how this breaks. > Using access=private means -> broken - Same is with excessive use > of track. > > The problem here is NOT the routers who drop access=private. I disagree with that conclusion. The problem is that we need access=with_destination_permission (bad name) for case 2 above. And, routers need to be able to say "use many segemnts of public access, and then the end of the route may use 0 or more segments with access=destination, followed by 0 or more segments with access=with_destination_permission". That keeps them from routing through. (Yes, I realize what I said is nonlinear and doesn't fit the simple sum cost model. THat makes it hard to compute, not the wrong answer.) -- You received this message because you are subscribed to the Google Groups "OsmAnd" group. To unsubscribe from this group and stop receiving emails from it, send an email to [email protected]. To view this discussion on the web visit https://groups.google.com/d/msgid/osmand/rmizhbountl.fsf%40s1.lexort.com.
