On 22/09/2009, at 10:57 AM, Martin Koppenhoefer wrote: > 2009/9/22 Anthony <o...@inbox.org>: >> It is possible to represent different surfaces and different >> maxspeeds >> without using more than one way. "maxspeed:lane=130;110"; >> "surface:lane=asphalt;concrete". That's not necessarily the best >> solution, > > indeed, it won't be understood by none of the apps that are using our > data and it doesn't say, which lane has which value...
The first bit is always going to be a problem, however we decide to map lanes the tools will need changing. Special tags or relations? They'll need to parse those to be useful. A new database construct? They'll need to understand that. There is no way around needing to modify the tools if we want more than a pile of tags that somehow relate to ways. The second bit we can easily get around by picking either counting from the left or right. Maybe something like lane[n]:key=value where there is data, and on the node in a way where the lanes change or merge we have lane-mapping data that says lane A back along the way changes to lane B after the way. > but that's exactly what I propose (map lanes explicitly) and it's > against the separate-ways-only-when-physically-divided-paradigm > (because an ambulance could change from one way to another)... There's a fun question, what counts as "physically divided"? I'm seen many emergency vehicles cross "physically divided" roads. If we exclude those, then what counts as a vehicle for the purposes of dividing a way? Are bicycles includes? Four-wheel drives? > Usually there will be ~100mtres for this where you can > change at any time, but in OSM you have to decide on one merging > point. I haven't seen a proposed scheme yet that really deals with this properly, they pretty much all pretend that a lane starts at a point. Personally I don't think that figuring this out is necessary for having multiple lanes, as we could continue the fiction that they start at a point. It'd be nice, but not necessary. _______________________________________________ talk mailing list talk@openstreetmap.org http://lists.openstreetmap.org/listinfo/talk