Talk:Lane assist/Examples/Motorway Deceleration Lane at Exit

Placement of nodes J

I, and many others, would place nodes J closer to H and I; given this image, even at the location of H and I (at the middle of the respective carriageways): the sharp triangle ("outside" the solid line) between the motorway and the motorway_link is in a sense part of the shoulder, and not a part of the carriageway, so the separation starts there. In a way, anyway. In the example the "real" physical separation (grass) even starts some 15 meters more to the east. If nodes J are ~20 meters more to the west, node G is also, but just a bit; my though is, that I'd draw a virtual line to continue the link road west from node J at the same angle (or at an even shallower angle) to see where it meets the main carriageway. The data model can not totally avoid short bits where

• a continuous surface of asphalt is in part presented by several ways (the first "segment" of a *_link way)
• angles of the first "segment" way don't correspond to the angles of the painted lines.

The best way to fine tune the nodes can even vary between simple junctions and places with complex link layouts very close to each other. Alv 23:03, 7 December 2012 (UTC)

• Depending on the course of the roads it might make more sense to put the nodes closer. I placed them rather far away from G to emphasize the prevention of steep angles, although some might suggest that there should be rather sharp angles because the OSM way should be in the middle of the carriageway. I decided to go for the less sharp angles for the reasons mentioned in the note. But all mapping variants have the same common problem: in case of splits or merges of roads we usually can not place the OSM way in the middle of the carriage way without creating some "zig-zag". So either we draw this zig-zag-way and risk misleading rendering results and routing instructions or we "smooth" the way a little bit. This is of course "mapping for the XXX" but unless we find a good way to specify what part of the OSM way is not in the middle of the carriage way (I'm preparing a proposal for this right now) I think it is a good compromise.
• Regarding the placement of G at the beginning of the legal separation: this is mentioned in the note. I would also prefer to put it at the beginning of the physical separation, but only if we have an accepted way to tag the legal separation. For this I proposed the key change=* and hope that it will be adopted.
• Regarding putting node J more to the west: don't forget our goal is a lane assist. It needs the exact beginning and end of a lane.
• To put all together: a lane assist needs very accurate mapping - micro mapping par excellence. Right now we have the keys lanes=*, turn=* and destination=* (together with its proposed sub-keys) as well as the proposed key change=*. What we are still missing is a solution for the transition problem, i.e. if the lane count changes the OSM way is not in the middle of a carriage way. If we find a solution for this we are done... well, besides tagging the whole world with the new scheme ;-) Suggestions for a solution to the transition problem are very welcome. --Imagic 08:33, 10 December 2012 (UTC)