admin relation of relations

Is it good practice to build a higher level border relation out of lower level border relations? I’m fixing up all admin_level=4 relations in South America, and I had some trouble with this one: http://www.openstreetmap.org/relation/301542
From what I understand, a complete border relation should form a closed circle without self-intersections. So if you join several bordering relations into one relation, there are going te be self intersections by default.
Also, I’ve got the impression that no software ‘gets’ meta-relations. The level 6’s out of which is built seem just fine. But my own spacialite database built on the Geofabrik continent .osm doesn’t seem to get it. And http://layers.openstreetmap.fr/ gets the layer 6’s, but doesn’t get the layer 4.
Third, building a relation out of other relations makes them even more fragile to clumsiness then they already are.

I found nothing on the wiki to clarify this. Also, I don’t know a simple way to rebuild this specific relation out of ways. And there are a huge amount of those in this case.

I think in that case you should only use the outer ways so you get a closed circle as a level 4 relation. But I am fairly new to relations myself so maybe an expert can shed more light onto this issue.

Maybe you can ahve a closer look at all the admin relations in Germany … becaise there have been efforts to make them one of the best admin data structure in the whole OSM data.

Try via http://osm.wno-edv-service.de:8080/boundaries/

And AFAIK it is no good to create relations that contain other relations themselves.

But a goog way is to have a border way that is member of different unique relations with different admin-level.

As writen in the wiki, the role “subarea” is “Optional, disputed and redundant (references to sub levels may also be found with spatial queries). Also referencing other relations makes editing more complicated in some cases.”

http://wiki.openstreetmap.org/wiki/Relation:boundary

Modeling an “admin boundary” by the sum of sub-areas is questionable. At least, the relation shouldn’t contain only subareas. Most if not all data consumers are expecting the normal multipolygons modeling with outers/inners ways.

OK, so I know what to do (I thought the talk about “subarea” was from a different point of view). Now can anyone imagine a workflow to recreate the metarelation without having to collect the thousands of bits and pieces the subrelations are made of?

I’m curious why you want to recreate the relations? I would have assumed these relations already existed, and you were just fixing them. If there are broken sections, just zoom into those areas and find the missing ways.

…or are there no admin_level=4 relations currently?

There are valid level 6 relations. There is a somewhat valid level 4 raltionship, which has as its only members those valid level sixes. Most renders and databases do not understand a realtionship containing only relationships. So a better solution would be to remove those level 6 members from the level 4 relation. Then replace them with the ways making the level sixes. Then removing the ways you added twice because of the internal frontiers which don’t exist at the level 4. I think this should be reasonably easy to do in JOSM.