[Tagging] Carriageway divider

Martin Koppenhoefer dieterdreist at gmail.com
Sun Aug 26 22:30:42 BST 2012

2012/8/26 Markus Lindholm <markus.lindholm at gmail.com>:
> On 25 August 2012 01:25, Martin Koppenhoefer <dieterdreist at gmail.com> wrote:
>> "purposeful" in this case translates to "mapping for the router" *1 in
>> OSM-speak.
> We're not supposed to map for the renderer nor the router. Exactly for
> whom are we to map?

I guess this is a misconception. With "mapping for the router" I
didn't mean that the data should not be used also for routing. Rather
this is read to mean: map something in a way that it does represent
something else but in a certain application under certain conditions
(e.g. router that does not do routing for emergency vehicles or
pedestrians) it still works as if was mapped correctly.

>> There is a convention in OSM that two highways represent
>> two carriageways, so when a single carriageway with a legal divider is
>> mapped like this, it is simply "wrong" according to our conventions.
> Sounds like you're the official spokesperson for OSM, are you?

No, I am not. I am simply telling you what I remember from former
discussions about this topic. This is not the first time someone
thinks that it doesn't matter to distinguish between physical
(impossible) and legal (forbidden but possible) separation.

> The convention you're referring to simply states
> "(physically) Divided highways should be drawn as separate ways."
> It doesn't say anything about legally divided highways,

yes, it's how we do documentation. We do (almost) never state what
something is not to mean or when it is not to be used, instead we say
when it _is_ to be used. Otherwise the wiki gets really hard to read,
because there is usually far more things to which a certain tag does
not apply.

> I'm definitely not the only one to map them as two ways.

yes, hence the public comment, otherwise I might have written in private to you

> Also, no one has offered any other solution to the routing issue.

there is a proposal (divider tag) and there are turn restrictions. If
you separate highways which are not physically divided you create
problems for some other use cases like emergency vehicles,
pedestrians, bank robbers and so on.

> The
> divider tag has been proposed, but I think it has been demonstrated
> not to work, as routing decision are made on the node and not on the
> line.

I can't remember that there was a demonstration that this approach
doesn't work. Routing engines usually get data that is postprocessed,
so the important thing is that the correct information is contained in
the data and can be derived. My guess is that nobody has seriously
tried to evaluate the divider tag for routing so far.


More information about the Tagging mailing list