[OSM-talk] [tagging] RFC :left/:right (asymmetrical roadside features)

Bernd Raichle bernd at dante.de
Tue Feb 17 12:49:14 GMT 2009

On Tuesday, 17 February 2009 10:36:16 +0000,
Andy Allan <gravitystorm at gmail.com> writes:
 > On Mon, Feb 16, 2009 at 7:11 PM, Norbert Hoffmann
 > <nhoffmann at spamfence.net> wrote:
 > > Andy Allan wrote:
 > >
 > >>And every time using :left and :right comes up, we all have a big
 > >>discussion about it and then nobody pays any attention and it comes up
 > >>again a few months later.
 > >
 > > Perhaps this is because the concept "left"<>"right" is so simple - and the
 > > aversion against editors, that are not totally key-ignorant is not so easy
 > > to understand.
 > And nobody pays attention. The main problem is that two-way roads have
 > no inherent, real-world, "direction" - neither side of the road is the
 > right or the left. Or rather, both sides of the road are the right or
 > the left, depending on which way you are facing.

Ok, if this is true, what is _your_ solution for the problem of
placing something on the "left" or "right" side of a road ... or a
direction-dependent tagging of a way attribute (e.g. oneway for a set
of vehicles, different speed limits for both directions etc.)?

Roads or ways have no ``inherent, real-world, "direction"'', but they
have a direction within the OSM model.  A way is represented by an
ordered set of nodes, thus the way has something which is called
"digitization direction".  If I want to tag attributes/properties
which are true only relative to this "digitization direction", I will
use a _simple_ means to specify this ... and after reading all the
current and past discussions about "left"/"right" or "in direction"/
"against direction" IMHO a direction relative tag _is_ a "simple" and
"normal" concept.

 >                                                   The only place that
 > right and left has any intrinsic sense is on one-way roads, which *do*
 > have an inherent direction (and signs to that effect).

One-way roads do _not_ have an inherent direction.  I am usually
allowed and I can walk _against_ a one-way road.  And I know a lot of
one-way roads where I can cycle against the direction, or where busses
are allowed to drive against the direction.  This property is

Additionally the current special handling of "oneway" and its special
cases (oneway=-1 for a oneway against the digitization direction!)
shows that there is a need for a concept for direction-dependent tags.
Why not use the inherent digitization direction, define and document a
simple tagging concept for direction dependent tags, and add support
for it to all editors and tools as it is already done for the oneway

 > Now the problem is that most people at the moment in OpenStreetMap are
 > tech-heads, and are so used to mental constructs and abstractions like
 > every road having a completely arbitrary intrinsic direction - but
 > that doesn't mean it's a great idea.

It is not the best idea.  On the other hand I have seen no other idea
which is simpler to understand.

 >                                      Editor support is less important
 > - and far easier to fix - than explaining to all the people who don't
 > even realise that all roads have a direction in openstreetmap - and
 > except for oneway roads, I have no idea which ways are pointing in
 > which directions, and it shouldn't be important unless it *has* to be
 > important.

If I want to add a direction- or side-dependent tag/object to the map,
the editors have to show the current (digitization) direction of the
road.  To edit a map we use a lot of mental constructs and
abstractions of the real world (a real world road is not a line with a
few pixel width, a real world intersection consists not only of two
lines connected by a simple node etc.).


More information about the talk mailing list