> I'm seeking advice as to best practice in the following type of situation: > > As an increasingly common example, now that people are getting around to > mapping areas such as leisure=, natural= and landuse= ... > > Consider the case of landuse=farm on one side of a highway (say a > secondary road) and leisure=golf_course on the other side of the highway. The > easiest way to map this - and the one usually adopted it seems - is to make > the > boundaries of the farm and the golf course both coterminous with the > highway so that the three lines are superimposed in the editors (not quite > sure > how the various renderers handle this) and the representation of the highway > has zero width. > > There are, however, potential problems with this (quite apart from the > slightly clumsy editing when several ways are superimposed) where detailed > mapping would ideally show that in real life the golf course and the farm do > not in fact have a common boundary but both are, for example, separated by > hedges (which may or may not be mapped) from the road. > > It is clearly possible to map the farm and the golf course as separated > areas with the road mapped as a line drawn between them - i.e. the mapping > has three separate parallel lines. This assists with mapping more clearly > features such as junctions of paths with the road (and stiles on paths at such > junctions). But is this unduly messy or does it create rendering issues > (e.g. if the lines are not absolutely parallel and just far enough apart to > render with random gaps between, say, the golf course and the road. > > The situation is even trickier where, say, a farm has been mapped as a > single area (same land use) with, say, a road crossing it - whereas in > practice, this is two separate farms - one on each side of the road - that > may at > some stage need to be named separately. Then we have to go back and split > the area, etc. > > This seems to be a quite a generic issue and I am wondering how people see > the pros and cons of (a) the simple approach with coterminous lines giving > a notional zero width to the highway, vs. (b) the more precise approach of > mapping the areas either side of the highway as areas that are separate > both from each other and from the highway. > > In practice, almost all mapping seems to use approach (a) - but would > approach (b) be easier for subsequent editing and addition of detail, and > rather clearer as it avoids superimposed ways and potential editing errors? > > Views?
IMO (a) is the correct way to do this. We are trying to represent reality in our database. In order to achieve this, certain abstractions are necessary. For a road, we can either choose to map it as a linear object (this is the common case), or we can map its geometry more exactly by using an area. In both cases, however, the object in our database represents the entire road (i.e. not only the middle line). Because in reality, there is no gap between the road and the areas next to it, there shouldn't be one in the database either. In other words, we should keep the topology intact, even if we choose to simplify the geometry. Regards, Marc -- Neu: GMX Doppel-FLAT mit Internet-Flatrate + Telefon-Flatrate für nur 19,99 Euro/mtl.!* http://portal.gmx.net/de/go/dsl02 _______________________________________________ talk mailing list talk@openstreetmap.org http://lists.openstreetmap.org/listinfo/talk
