2009/9/10 <lulu-...@gmx.de>: > Currently the problem with routing is, that routing applications can not > check for millions of nodes, but only for thousands of ways without > performace problems. > > That results in the problem, that for example barrier bollards in the middle > of the road can not be considered. > > I recommend to add a route_nodes=yes or router:check_nodes=yes to such ways, > so that the routing software can check only the ways in first run and then > check the nodes on ways only where this is set. > > Comments?
I hate to say the same thing everyone else already said, but yeah, software doesn't work this way. There's something called complexity ([0]) that increases and decreases when the problem changes, but sometimes in unobvious ways. A change that lowers the complexity of the routing problem would be worth considering, but the change you propose does not affect the problem's complexity, where the operation with highest cost is still the path finding. Ways with a barrier can be completely removed from the database during population, or be given higher cost when the barrier is one that only slows down traffic such as crossings with traffic lights. The software still has to process all crossings with other roads because a level-crossing with a primary road is costlier than a crossing with a tertiary road on your path. [0] http://en.wikipedia.org/wiki/Computational_complexity_theory Cheers _______________________________________________ talk mailing list talk@openstreetmap.org http://lists.openstreetmap.org/listinfo/talk