[Tagging] Very large multipolygons

[Christoph Hormann]

On Tuesday 02 May 2017, Frederik Ramm wrote:
> [...]
>
> I realize that what I've done here is close to "mapping for the data
> extractor" and I'm offering these decisions for discussion here. I
> definitely think none of the three areas qualifies as a proper
> "multipolygon" and *if* these huge areas need mapping, then some
> solution based on "boundary" is better.

I agree there is no point in mapping this kind of feature as 
multipolygons in OSM but i disagree that changing type=multipolygon to 
type=boundary makes it any better.

Mapping them as polygons does not work well because:

* the frequent lack of local verifiability of the mapped geometry, 
especially with things like mountain ranges.
* the redundant nature of much of the mapped geometry, in particular 
with coastline bound features.  The coastline is already mapped and 
mapping a sea or peninsula by adding a subset of the coastline data to 
a relation is pretty pointless because a sea or peninsula is implicitly 
bound by the coastline.  Collecting the relevant parts of the coastline 
in a relation would simply mean having the mapper spare the programmer 
and computer some work in data analysis.

I thought about how this kind of feature could be represented in OSM in 
a practically useful and manageable way in the past and came to the 
conclusion that

* it would be useful to have a a robust option to sparsely map this kind 
of feature, i.e. map them in a way that avoids the need for the mappers 
to add redundant and non-verifiable information but allows them to add 
relevant and verifiable information as far as it exists and still have 
a formally valid and consistent object.
* it will be very hard to get acceptance for such a mapping concept 
because of the widespread paradigm that everything in the world can and 
should be represented as points, linestrings or polygons (or nodes, 
ways and multipolygons in OSM).

Some further discussion on the matter can be found here by the way:

http://wiki.openstreetmap.org/wiki/Talk:Tag:natural%3Dvalley#Valley_as_relation.3F

-- 
Christoph Hormann
http://www.imagico.de/