[talk-au] Roundabouts and routing

[Steve Bennett]

On Fri, Dec 18, 2009 at 7:59 AM, <morb.gis at beagle.com.au> wrote:

> In the first case I have edited the entry, exit and roundabout as meeting
> at
> exactly one node.  IMHO this represents reality and if the router can't
> handle
> it then the router should be upgraded to suit (or its OSM-to-router-format
> script suitably upgraded).
>

This single node approach looks wrong to me.  Clearly, I stand on the "map
topology, not asphalt" side of the divide, but I would point out that:
- there is a dashed line on the southern entrance, and presumably a
roundabout sign. You're entering the roundabout and leaving again.
- the way this is mapped is how I would map it if there were two lanes, one
which went directly to the other exit without entering the roundabout. This
isn't the case here, you have to enter the roundabout.
- clearly, to get from that south road to the west road, roundabout rules
would apply. (Not that I know enough about road rules to know the
difference).

I'm sorry, but this just doesn't even look close to me. I can imagine
ambiguous circumstances, but this isn't one. This is analogous to the
situation where two side streets join a street almost at the same point. You
map it as either two separate nodes or one, depending on whether there is
even a slight offset. Same thing here. If you can't get from the south road
to the west road without travelling, even for a short time, on the
roundabout, then it should be mapped as such: with a section of roundabout
between the two nodes.



>We shouldn't be "mapping for the router" to the extent it disagrees with
>reality, should we?

No, but:
- mapping reality breaks down when the scale is of a few metres
- there is not a trade-off here: mapping with two nodes correctly represents
the fact that you're entering the roundabout.
- the concept of "mapping for the [renderer/router]" is limited to mapping
in a particular way to address particular deficiencies in particular
software. You can't just chant this as justification for any point of view
that happens to work less well in existing software.

Steve


>
>
> Brendan
>
> Quoting Ben Kelley <ben.kelley at gmail.com>:
>
> > Heh. My sister-in-law lives near here. The directions on my eTrext Legend
> > did confuse me a little at that intersection. I assumed the intersection
> > hadn't been drawn quite right, but I never got around to going back and
> > fixing it.
> >
> >  - Ben.
> >
> > 2009/12/17 John Henderson <snowgum at gmx.com>
> >
> > > Here's an example:
> > >
> > > http://www.osm.org/?lat=-35.205609&lon=149.027603&zoom=18
> > >
> > > Heading north along Florey Drive, my route takes me right (towards the
> > > east) on Ginninderra Drive via the roundabout.  I should be taking the
> > > third exit, but the GPS with an OSM map says to take the second exit.
> > >
> > > And this is because of the way the roundabout has been drawn.  The
> first
> > > exit (west on Ginninderra Drive) isn't being counted as an exit because
> > > that exit point is also my entry point.  In other words, it shows that
> > > one need not actually enter the roundabout to take that first turn.
> > >
> > > I believe that this roundabout does not accurately represent the
> reality
> > > of the situation.
> > >
> > >
> >
>
>
>
>
>
> _______________________________________________
> Talk-au mailing list
> Talk-au at openstreetmap.org
> http://lists.openstreetmap.org/listinfo/talk-au
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.openstreetmap.org/pipermail/talk-au/attachments/20091218/0db28081/attachment.html>