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

Andrew Chadwick (email lists) andrewc-email-lists at piffle.org
Tue Feb 17 13:25:28 GMT 2009

Mike Harris wrote:
> While sympathetic to the underlying need being discussed in this thread, I
> suspect there is a further problem. Although a way has an intrinsic sense in
> OSM, this is fairly volatile! All it needs is someone to reverse a way - and
> this can happen rather easily, say, when combining two ways with the same
> tags but different senses (yes - there is a warning but it's all too easy to
> click through). Reversing the ways then, of course, reverses the 'left' and
> 'right' descriptors with their differing tags!

Yes, it'd be nice and simple to implement. Just swap any :left and
:right -suffixed tags. That's why
discusses it in some detail.

If people whine about it enough, I'll make osm2go do something like
this, as a sort of reference implementation :) We're probably being lazy
about oneway, and I'm entirely in favour of building simple smarts like
this into an editor that's as simple and Joe-Consumer-focused as osm2go:
I want to make it difficult for Joe Consumer to do the wrong thing
accidentally without getting in his/her face (no warning dialogs, just
do the right thing given that there's a single, obvious, 99%-of-the-time
right thing to do here)

> This leads me to wonder
> whether an absolute sense (north, south, etc.) is still better even though
> it might require that a way is divided a bit. Most ways do have a 'general'
> compass direction for long segments even if this is often more human-obvious
> than machine-obvious. The main exceptions are likely to be short residential
> streets on housing developments etc. - 'circles' etc. - but these are less
> likely to require unilateral tagging.

Hmm. An argument from plausible estate design; innovative! :)

For the reasons you stated, :north, :south etc would break down for some
ways. Also, if an editor rotates the way, it'd also have to "rotate"
some of the tags. Some of the time. It's really quite a bit more
complicated than a :left/:right scheme for coding, and it doesn't buy
you much more expressiveness.

For points, it makes more sense. But pub:heading=<degrees> and
pub:distance=<metres> would be finer-grained for those XD

> Btw, I have encountered the same problem with canals. Some mappers describe
> towpaths as being 'left' or 'right'. Personally, I prefer to map the towpath
> as a separate way alongside the canal - with the added advantage that this
> allows me to tag the towpath, e.g. with access rights, surface condition,
> barriers, reference numbers, route relations, etc.

For towpaths, I'd agree. They're 'sufficiently segregated' from the
waterway in my head (you have to change mode fairly significantly to go
from one to the other!)

> Perhaps this would also be a better approach for e.g. cycleways alongside
> motor roads? Although, I have to admit that it doesn’t solve the problem of
> unilateral naming.

Sometimes, sometimes not. Cycleways where you can/sometimes have to
rejoin the ordinary carriageway should not be, :left and :right (or
cycleway=opposite_*, broken though it is) do that more plausibly because
then you're dealing with the same road. Cycleways that are long,
continuous, separated by a grass verge, or go along the sides of 70MPH
roads get the separate treatment when I tag them (often they need things
like lit=yes/no too, those ones).

Andrew Chadwick

More information about the talk mailing list