On Thu, Sep 25, 2008 at 11:16 AM, Frederik Ramm <[EMAIL PROTECTED]> wrote: > Hi, > > On 25.09.2008, at 12:03, Andy Allan wrote: >> And it means that areas preserve their actual dimensions >> instead of growing outwards beyond their actual boundary - an >> increasingly proportionately large problem as we add ever smaller >> details to OSM. > > But is that not a problem for roads as well? What use is it to have > the car park exactly where it is in reality when the *road* next to > it will grow outwards beyond its actual boundary because people > thought in unnecessary to actually map the extent of the road surface? >
It is a problem for roads, but that's just an argument for mapping road areas. My main problem with mapping areas to the centre line (I'm not really counting landuse here, its possibly an exception) is the farcical situations to which it leads. The buildings example is the best: it's not uncommon at all to have a road abutted on both sides by buildings. If we map them to the centreline then we possibly increase their size by 50%, and happily record that they share a wall. Obviously they don't, and if you somehow figure out there's also a road in there, you might be able to reconstruct the real situation given a road width (which very few people actually record). So we say not to do it for buildings, but only for parks and car parks (I've seen this advice given). So now we have these sticking out into the road compared to a building right next to them. And there's lots of roads with grass verges; how big do they get before the object doesn't abut the road, and where is this width accounted for? The situation is confused greatly as soon as you try to map the park's fence or wall (if it exists): where's the road, inside or outside? Or we have a post box on the side of the road on the pavement: these are almost universally mapped as nodes where they are, rather than as a way node -- if we mapped to the centre line these post boxes are now inside the park? Or if the park diverges from the road at some point, do we maintain the offset and get the angles right, break the angles, or introduce an arbitrary curve? I get that the roads are being mapped as abstractions, and I get that it makes a lot of sense to stay abstracted, but on the whole I'd say the number of questions raised by this approach makes it a lot easier in the long run to just give up. We already break the abstraction for so many features precisely because coming up with tagging schemes to cope is just too hard, and because users find it so much easier to just plonk something on a satellite photo. Dave _______________________________________________ talk mailing list talk@openstreetmap.org http://lists.openstreetmap.org/listinfo/talk
