[OSM-talk] Layer transitions

Lambert Carsten lhc.osm at solcon.nl
Thu Aug 13 13:20:04 BST 2009


On Thursday 13 August 2009 00:31:23 Lauri Kytömaa wrote:
> Lambert Carsten wrote:
> >sense. Even though the smaller road ends at the edge of the larger road
> >not the middle of the road.
>
> Inside the crossing area the roads overlap, neither ends there - you're
> on both roads. But you're not on the bridge that starts only several
> meters away - or inches away if you're already moving towards the
> bridge.
>
> Taking the canal bridges mentioned previously: if you draw the
> riverbanks/canal edges, it has been the recommendation (I'd have to dig
> the talk list archives for that) also that the bridge=yes starts and
> ends at the points where one can get under the bridge - at the water's
> edge, meters away from the ways marking the roads running parallel to
> the canal.
It isn't about inches or even meters. It is about representing reality in such 
a way that renderers, routeplanners whatever can do interesting things with 
the data.
Starting the bridge at the edge of the water is not a good recommendation 
because it has to be made clear to whatever program parsing the data that it 
is ok to have a road going 'through' water because we have a bridge. Just for 
practical editing reasons the bridge node should not be too close to the 
water edge.
>
> One thing about the same layer check occurs sometimes: a motorway link
> road joins the motorway right at the point where the bridge starts: most
> notably the case where the road markings indicate three lanes on the
> bridge and only two (+1 approaching but still separate) up to the exact
> point where the bridge starts. We can make the connected node something
> other than the point where the lanes come into contact (the acceleration
> lane's hundreds of meters long anyway, but that would need some kind of a
> note to be consistent - at least once more people start to get to that
> level of detail.
If I understand this example correctly it is another example that the 'rule' 
as rule is wrong.

On Thursday 13 August 2009 11:51:36 Jochen Topf wrote:
> On Wed, Aug 12, 2009 at 10:58:36PM +0200, Lambert Carsten wrote:
> > > That often leads to inconsistencies, but
> > > inconsistent is not necessarily bad.
> >
> > Exactly, no need to 'force' junctions to have connecting roads on the
> > same layer. If you really believe in the 'middle of the junction theory'
> > let us be inconsistent with that one and help mappers clean up real
> > issues rather than adding titsy bits of road between junctions and
> > bridges.
>
> I didn't say anything about "titsy bits". If you only add this to work
> around the Keepright message, that doesn't make any sense. It only makes
> sense to add a non-bridge part of the way, if it actually is large enough
> to make a difference when rendering.

Well sadly while I have been discussing this here that is exactly what has 
been done. (No one's fault, it's the way things go.)
Most of the bridges in Amsterdam now have 2-3m connection bits (less than half 
the width of most roads) between the bridge way(s) and the junction (example: 
http://www.openstreetmap.org/?lat=52.36607&lon=4.88279&zoom=17). 

Hopefully this will not cause (too many) other issues in the future. Often the 
renderers try to insert the name of those bits separately. I dread that next 
people will start 'cleaning up' that by removing the name there after which 
Keepright will (correctly) mark it as a street with no name....  :(

I hope I am wrong, I really do.

>
> > > And if we get better software or more
> > > detailed data or want to support new uses, we'll change things around.
> >
> > Can we agree on getting rid of the "Right" and especially "Wrong" texts
> > that deal with this issue in the Map Features and give some advice like:
> > "Often bridges/tunnels will not connect directly to a junction in which
> > case you should/can (?) add a piece of road connecting the two." ?
>
> The page could definitely have a more thorough description of the problems
> with either way of mapping and give some advice, when each makes sense.
> Feel free to add this. Your sentence above sounds like a good start to me.
Well thanks for that! Consider it done.

btw I noticed there are some translations regarding this issue. Is there a way 
to inform them or is it their job to keep an eye on the English version?
>
> Jochen


Lambert Carsten




More information about the talk mailing list