[OSM-talk] Validator Plugin (landuse=forrest+highway=xy)

David Dean ddean at ieee.org
Fri Nov 23 16:01:08 GMT 2007


I agree that parks/forests/etc should probably re-use the nodes of any
linear features on their boundaries.

However, I have not been doing this in practice because a
park/forest/etc completely bordered by other ways cannot be selected
for editing (in JOSM at least).

If we are going to re-use nodes for multiple overlapping ways, we
really needs some sort of layering in the editors so the right way
from a group of overlapping ways can be selected.

Of course it is possible I am missing something here, so let me know
if this can already be done in some manner.

- David

On 24/11/2007, bvh <bvh-osm at irule.be> wrote:
> On Fri, Nov 23, 2007 at 01:26:32PM +0100, Joerg Ostertag (OSM Munich/Germany) wrote:
> > I thought we want to reflect the reality as good as possible. And the forest
>
> Obviously, that is not the yardstick because then we would be tagging
> roads as an area too. I can't speak for everyone, but I still
> think mapping is the main purpose for openstreetmap. And a map is
> by definition a stylized version of reality. (idem dito for routing)
>
> > simply is not starting in the middle of the road. Only this way we increase
> > the chance of later being able to use the data for other stuff than
> > map-rendering.
>
> I also don't think this is a valid argument. When such an application
> arises it is a mechanical step to decouple the street from the forest
> and offset the latter inwards for a certain amount.
>
> > So if a renderer wants to really draw it really correct in low zoom, it would
> > have to guess that if there is a railway he has to erase 2 meters of wrong
> > drawn forest wood, if there is a highway primary he'll have to remove 6
> > meters or maybe more ... and so on.  That sounds really wrong! So I would
>
> Yes that is really wrong, they just draw the street on top of the forest.
>
> > really encourage everyone to enter Data as close to reality as possible and
> > not only draw for our current rendering engines which we know still have some
> > way to go.
>
> I think the effort is counterproductive. Compare with a fysics experiment
> where you use a handstopped watch and you know the your reaction time is
> 1 second. In that case when you stop your clock and it reads 17.45s
> you write down in your report : 17s. Adding the extra digits is
> wrong because you can't guarantee the accuracy of them: the 4 could just
> as well be 3.
>
> I feel the same way with this kind of information in openstreetmap :
> at a certain point adding physical detail to the boundaries of
> objects does not add any information (because the information is
> inaccurate since your gps is generally at least a few meters of in a
> forest, because the detail changes too much like next spring when
> a new shoot roots just outside your border, or maybe because what
> exactly is the border of a forest. Is a tree 5 meters out still part
> of a forest, 3meters maybe, 1meter then?).
>
> So instead of going for uninformational geographical detail, go for
> valuable logical structure : ie explicitly show that the forest
> is bordered by the street. And you can do this by reusing the nodes...
>
> cu bart
>
> _______________________________________________
> talk mailing list
> talk at openstreetmap.org
> http://lists.openstreetmap.org/cgi-bin/mailman/listinfo/talk
>


-- 
David Dean
PhD Research Student, RP-SAIVT, QUT
(me) http://www.davidbdean.com
(saivt) http://www.bee.qut.edu.au/projects/saivt/
(post) Room S1102, GPO Box 2434, Brisbane, Australia 4001
(p) +61 7 3138 1414 (m) 0407 151 912 (f) +61 7 3138 1516
(CRICOS) 00213J




More information about the talk mailing list