[Tagging] Traffic sign's relevant direction: direction=* vs. relation [Was: traffic_signals:direction=* vs. direction=*]

Jean-Marc Liotier jm at liotier.org
Thu Mar 23 16:23:48 UTC 2017

On Thu, 23 Mar 2017 16:31:47 +0100
Martin Koppenhoefer <dieterdreist at gmail.com> wrote:

> 2017-03-23 15:47 GMT+01:00 Jean-Marc Liotier <jm at liotier.org>:
>> I'm not operating a routing system so maybe someone who is
>>   might offer his opinion on that point
> in another thread a few days ago there was a comment by Daniel
> Hofmann who works at mapbox on OSRM:
> https://lists.openstreetmap.org/pipermail/tagging/2017-March/031727.html

Wow, thanks - how did I miss this thread... Well, the good news is
that the question of traffic sign direction modeling is in the air !

So, OSRM doesn't grok relations and doesn't process direction=*
either, nor any of its cousin tags - and the whole thread does not
mention anything else that does, with the exception of Kenzi3D which
does  direction=* ... Very well then: since there is very little
consumption to impact with the choice of a modeling method, I would say
that we are rather free to propose an Openstreetmapically correct way
and data consumers shall follow.

After reading the thread, it seems that our ideas here are not
unlike what has been discussed there.

So I stand by my current position, supporting either or both:

- relation (type: to be defined) including the highway=stop node and
  the incoming way with a from:forward or from:backward role...
  Universal but some consumers such as OSRM balk at processing relations

- highway=stop+direction=forward node on the incoming way... Only
  covers the simple case but covers it simply

I believe that boiling down the issue to just those two (and maybe live
with them both for an indeterminate duration) would significantly
improve over the current state.

More information about the Tagging mailing list