[OSM-talk-be] some thoughts and remarks about border mapping

[Ben Laenen]

On Wednesday 24 September 2008, Maarten Deen wrote:
> Ben Laenen wrote:
> > On Tuesday 23 September 2008, Lennard wrote:
> >> The Dutch border system as it is now is comprised of overlapping
> >> ways using shared nodes.
> >
> > ugly :-)
>
> It's not. Certainly not with the merkaartor option of "split node"
> where you can copy a node and attach a different way to each node.

I doubt you can call it a beautiful solution either. Why add so many 
duplicate information? If the methods are there to re-use the same way 
many times, there's no need to duplicate the same way over and over. 
Data duplication is the first thing you should avoid in a database. If 
we take the current border between Antwerp en Woensdrecht for example, 
I count no less than 10 boundary types for it with current admin_levels 
as said on the osm wiki:
* The Netherlands
* Noord-Holland
* Woensdrecht
* Belgium
* Flemish Community
* Flemish Region
* Antwerp province
* Antwerp arrondissement
* Antwerp municipality
* Berendrecht-Zandvliet-Lillo district
(you can add an 11th one if you count neighbourhoods (wijken) which is 
also a further division made by Antwerp as well: Polder)
So you think that the same way 10 times on top of each other isn't ugly?

Secondly, it gives renderers a hard time. They have to somehow detect 
that there are other boundaries that overlap at certain places, which 
results in the kind of ugly overlapping dashed borders you see right 
now in the Netherlands.

> And I certainly would hope that nobody is going to change the dutch
> border soon. It is properly mapped and does not need changing.

Don't worry, I'll keep away from every boundary inside the Netherlands.

But you never know what'll happen with the Belgian-Dutch border if we 
start tagging things.

> I have a big problem with this "left" and "right" notion. You don't
> have left and right with lakes or islands or any other area. IMHO a
> administrative boundary also defines an area. An area is closed (be
> it by way of closed way or by relation) and therefore does not need
> indication what is inside and what is outside.

Well, I'm not sure it's needed yet. Just the fact that a situation like 
Baarle can be so complex that it might need it.

Ben