[Tagging] Forward/backward routes

Kevin Kenny kevin.b.kenny at gmail.com
Mon Aug 19 13:18:59 UTC 2019

On Mon, Aug 19, 2019 at 6:34 AM Peter Elderson <pelderson at gmail.com> wrote:
> Keeping linear main route and alteratives separate is actually quite straightforward and much less work than creating and maintaining routes with roles. Especially when the forward/backward roles do not indicate the direction of travel, as is the case with bicycle routes.

The 'forward' and 'backward' roles are well defined. At least two
editors that I've used understand them. JOSM appears to sort them
correctly. (It does stumble a bit in the special case where a
route=road ends at junction among dual carriageways, because it cannot
find a single endpoint.)  We seem to have this problem mostly solved.

I disagree with the 'much less work' when you're maintaining a route
that's hundreds of km long, follows dedicated trails or rural roads
most of the way, but once in a while enters a city and splits because
of one-way streets. Maintaining separate 'forward' and 'backward'
relations for the whole thing would be quite annoying.

> The point of route relations is that they are already routed, i.e. they prescribe exacty what ways to travel. They are not suggestions for routing. Of course, if there are problems, you can try to make up for that using routng techniques, but you should not turn that around and say hey, I am routing anyway, so forget the data issues. The idea is to record exactly how to travel, so people can travel along the path without rerouting. Just take a way, then the next, etc until the end. These are not suggestions, it's a prescribed route to follow, as seen on the road because it's signposted/waymarked.

Right. Structuring the route in that form also allows software to
produce the skeleton of the route book.  I started with the output of
a PostgreSQL query when I wrote the initial draft of
(I found several gross errors in the 'official' route book that way:
there was one segment that was shown as 1700 m shorter than its actual
length, for instance).

73 de ke9tv/2, Kevin

More information about the Tagging mailing list