[Tagging] complete tagging of all 'right of way'-cases

Ruben Kelevra ruben at vfn-nrw.de
Fri May 11 01:07:59 UTC 2018

10 May 2018 23:57:00 Martin Koppenhoefer <dieterdreist at gmail.com>:

> generally, a relation is capable of explicitly modelling the
> situation, at the cost of complicated/time consuming actions required
> from the mapper (and nowadays, with a lot of mapping on mobile phones
> going on, average editor relation support is even worse). Personally,
> I am mapping in a context with a lot of oneway streets, so it often
> is not an issue to just use a node close to the intersection, I
> believe if you can, you should avoid relations because it makes the
> map more complex for others to understand and modify (maybe with the
> exception of well supported multipolygon relations).
I understand you're concerned, but we already have very complex
solutions for bus routes, turn restrictions and hiking routes. A
typical street inside a city usually contains easily dozend relations.

In the early days, this was an issue with certain editors, but this time
is long gone. Sure, newbies tend to break stuff, but we have tools to
find those broken relations and fix them.

I don't see an issue with this kind of complexity.

I've added some of those relations to test out different usecases and
it's not harder or take longer than to add a node and move it around. I
might be trained working with relations, but I don't think it's a huge
disadvantage in this case.

> The particular proposal seems thought through, but might eventually be
> overengineered.
Yeah, might look like first, but our data users always envolve and we
had a long time no useable data on this topic. Noone is using those
highway=stop/give_way. The reason is simple, they are often broken, not
easy to parse and this might lead to unexpected results. I think this
might change in the future - like it did with the turn_restrictions.

In the future, we're able to even tag those cases :P


> In the simplest representation you would only need a
> via node (at the stop line) and a from way (ending at the stop line)
> and be done.
No, you misunderstood the proposal. It's basically exactly the same as a
turn restriction. You don't split the way between the stop line and the
intersection, but just add the intersecting node with a via-role.

If you have two ways crossing: X and Y with intersecting node A.

You select A as via-node and X as from-way and Y as to-way.

If you want to map the exact location of the sign, you can add an
additionally node at the place and add this node as traffic_sign and
map it with traffic_sign=DE:205.

Now a renderer might draw the sign exactly at the location you've

If you want just the old behavior the proposal explains it, too:

role location_hint - "a hint to a renderer as to where might be a good
place to position a symbol indicating the give way. [...] Note that
this member has the same meaning as an obsoleted highway=give_way node."

> I am not sure how several via ways should work together with several
> from and to ways, and I guess even if it works, it will be
> complicated to evaluate (for other mappers) and several very simple
> relations (only from way and via node) would probably be much easier
> to understand (at the cost of having to split the ways at the
> stop/via).
One relation is just for one stop or give_way sign. So if there are all
ways exactly the same, a simpler solution without any relation is much
better. I want to create an additional proposal for this.

This would look like this:


A bit more complex case, which was previously not tagable at all can be
found a bit south:

https://www.openstreetmap.org/node/102694679 (intersecting node)

The street has just a curb, no signs to the right and the left. Since
we were just able to tag signs, previously no information about this was
available at all in our database.

PS: I hope my explanations were understandable, I'm a bit tired and
that's the last mail today. If not, just let me know.

Best regards


More information about the Tagging mailing list