[Talk-us] routing tags used by actual routing applications

[Tod Fitch]

The western US is also full of unpaved roads. When I've mapped them my criteria for deciding between highway=track and highway=unclassified or highway=residential has been twofold: 1) Could a family sedan reasonably be expected to be able to use the road? And 2) Is it wide enough for two vehicles to pass one another? If it passes both those tests then then I use either unclassified or residential depending on the number of houses served on it. If it fails either test, then I use track.

In routing I suspect that the vast majority of real world users will be expecting a fast and smooth route to take in the family car so routing software should probably default to avoiding tracks. But there are other significant use cases: Mountain bikes would probably love to be routed over tracks, road bikes would probably prefer paved bike lanes, trucking concerns will be interested in having the routing software take account of HGV restrictions, etc.

In compiling the best practices tags used for routing, it may be a good idea to break the page into those different usages and for each usage list the tags used by the various routers.

On Jul 2, 2014, at 7:07 AM, Kevin Broderick wrote:

> IMO (based on both the wiki and what I've seen on the map), highway=track implies something that is not reasonably drivable by normal passenger cars at a normal rate of travel. In Vermont, we have a whole lot of unpaved roads that are perfectly fine at 35-45 MPH (well, except for mud season); those seem to fit highway=unclassified or highway=residential better than highway=track.
> 
> The routing discussion does get into a bit of sticky area that applies at least to Vermont, New Hampshire and Massachusetts—all three have old roadways that remain legal right-of-ways but that are not town-maintained. Some are privately maintained to normal road standards, others are maintained to some lesser standard (e.g. "So I can get my pickup up to camp"), others have very little to no ongoing maintenance but get traveled by 4x4s and dual-sport motorcycles (whose operators are likely to clear deadfall but not to replace washed-out culverts, for example), and others have reverted to forest or may even have been "paper roads" that were plotted with inadequate knowledge of topography (up cliffs, etc.). According to the wiki, those should be tagged motor_vehicle=yes because a road-legal vehicle is *legally* allowed to travel them ("Access values are used to describe the legal access for highway=*s"), but I sure as heck don't want to get routed down some of them when driving my Taurus. I may very well want to get routed down them while riding my dual-sport motorcycle or if I was out in a 4x4 truck.
> 
> As I've been updating data in Vermont, I've been relabeling highway=unclassified or highway=residential to highway=track if it would seem to be an unpleasant surprise while operating said Taurus, and I've been using the somewhat debated smoothness= tag to add further data where possible. I've also been adding in missing sections of those unmaintained right-of-ways, usually as highway=track, that were not on the TIGER imports.
> 
> IMO, the default expected behavior of a routing system should be to avoid highway=track unless specifically encouraged to do so by user input (whether by selecting a particular activity or by the user putting a waypoint on a track), and we should encourage renderers to clearly distinguish tracks from roads.
> 
> Also IMO, any track that is at all visible on the ground and congruent with a legal, public right-of-way, ought to be on the map. However, we do need to tag them appropriately so that routing and rendering systems can distinguish those ways that are legal for motor vehicles but physically impassible for most from those that are legal and readily passable (and, where possible, also distinguish the converse—those that are passable but not legal ROW).
> 
> 
> On Wed, Jul 2, 2014 at 9:11 AM, Paul Johnson <baloo at ursamundi.org> wrote:
> I would expect tracks to be in play except when explicitly excluding unpaved roads, barring surface tags to the contrary, otherwise as a last resort.  Much of the US doesn't pave county roads, yet they're often packed and graded to the point someone with a low slung sedan can safely do 45 on them.
> 
> On Jul 1, 2014 3:37 PM, "Martin Koppenhoefer" <dieterdreist at gmail.com> wrote:
> 
> 
> > Il giorno 01/lug/2014, alle ore 23:15, Jason Remillard <remillard.jason at gmail.com> ha scritto:
> >
> > For example, scout does
> > not route over highway=tracks, unless you are in pedestrian mode. It
> > seems like a reasonable decision, perhaps all of the routers do this,
> 
> 
> no, some routers do use tracks for car routing (I'd expect a router to use tracks for cars, but only as a last resort when there are no alternatives)
> 
> In your original post you mentioned path and cycleway, those should indeed not route cars
> 
> 
> > but the wiki documentation says nothing of the sort, and it surprised
> > me.
> 
> 
> I'd file a bug at scout and see what they respond, you should definitely not adapt osm data based on one router
> 
> 
> cheers,
> Martin

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.openstreetmap.org/pipermail/talk-us/attachments/20140702/f6e5e5a5/attachment-0001.html>