I notice that the wiki page on multipolygon validity, under point 6, states that the “forest with lake with forest on island” falls foul of section 6.1.11.1 (e) of the OGC Simple Features spec, which states that “The interior of every Polygon is a connected point set”.
Thus far I agree, but is this really relevant? Shouldn’t the reference point for OSM multipolygon relations be the OGC MultiPolygon object?
OGC Polygons can have zero or more inner rings (but only one outer ring). In OSM, as soon as one wants an inner ring, it has to be represented as a multipolygon relation. Thus there are some OGM polygons that would have to be represented in OSM as a multipolygon relation. The example cited is a valid OGC multipolygon.
I take that to mean that the example should be removed from the wiki page, as the nested polygon is a valid OSM multipolygon relation (and a valid OGC Multipolygon, but not a valid OGC Polygon)?
In fact, the very same situation is shown as valid under the “Island within a hole” heading on the main multipolygon page.
Even if it is a valid OSM multipolygon relation, renderers cannot evaluate this shortcut construction of an “island within a hole” correctly. See the comment Beware though, that following this method, still a dual determination of the area mapped by way 3 will be created, as the hole and the island DO overlap!
So, the rendering only works if the hole has no tags - a special case, which almost never occurs.
I recently changed the German wiki accordingly to a formulation which is both a lot more intuitive and works in all cases.
Note “This page is only a proposition, it doesn’t represent a consenus of what is or isn’t valid, but only what could and couldn’t be considered valid” as header.
I guess that it should be moved to proposal namespace.
EDIT: I moved it to proposal namespace on 2024-04-24