On 31 January 2011 15:44, Anthony <o...@inbox.org> wrote: > On Mon, Jan 31, 2011 at 5:44 AM, M∡rtin Koppenhoefer > <dieterdre...@gmail.com> wrote: >> 2011/1/31 Steve Bennett <stevag...@gmail.com>: >>> I think I agree with your earlier point that mp's are better than >>> colinear ways, but colinear ways are still better than parallel ways >>> for areas that do actually touch. >> >> Yes, parallel ways are actually to be considered errors in the case >> that the polygons really do touch > > Agreed, although I'd like to point out that in a case where one of the > features is physical and one is virtual (for instance, a road and an > administrative boundary), I wouldn't classify that as features which > "touch", and I think parallel ways *are* a viable solution. > > To wit, I'd say parallel ways are the proper solution for TIGER > boundaries which coincide with TIGER lines. Especially when the way > is a dual carriageway. Fixing dual carriageways which share nodes > with TIGER boundaries sucks. Fortunately most TIGER boundaries > themselves suck, so a simple fix is to just delete the TIGER boundary. > > (Note that there's no problem with two *boundaries* sharing nodes or > (preferably) ways. I'm talking about a road sharing a way with a > boundary, which maybe is okay sometimes, but sometimes definitely is > not.)
The example that come to my mind is the case where an administrative boundary is _defined_ by a river or stream for example. In this case I'd say that the boundary and way should share nodes. However, I agree that this might get messy in the case of a dual-carriageway. -- Matt Williams http://milliams.com _______________________________________________ talk mailing list talk@openstreetmap.org http://lists.openstreetmap.org/listinfo/talk