OK, I'll bite about the cycleway. Go a bit south and you will see something called Macharisdreef and something called Baan nr. 90. Imagine you want to cycle from Macharisdreef to Baan nr. 90 along that N47 and use only the map data as a routing algorithm would do. * First attempt: well it is obvious, you just take the cycleway at the east of the N47 to cycle from the one to the other. Shortest, lots of cycleway, it looks perfect. Reality: this is illegal, you only can follow that cycleway north, otherwise you cycle on the wrong side of the road. So we have to correct the data. Let's say we put a oneway=yes on the eastern cycleway. * Second attempt: well obvious again, you simply take the N47 instead of that cycleway. Only slightly longer, a bit less attractive, but not too bad. Reality: this is illegal. There are cycleways, you have to follow those. So we have to correct the data. Somehow we have to make explicit that you are not supposed to cycle on the N47 but have to follow the parallel cycleways. So let's put some tag explaining that on the N47. I don't really care which one. It could be boldly bicycle=no or the proposed bicycle=use_sidepath or whatever. I'm just not supposed to cycle over the highway=primary. * Third attempt: let's see. When leaving Macharisdreef, I have only two options. Either I take the N47, but that is illegal, or I take the cycleway east of N47. Nothing else connects to Marcharisdreef. So my only option is to follow the eastern cycleway to the north. Hmm, I then can cross the N47 right before that terribly complicated junction and start following the western cycleway. Until I find a place to cross the N47 again. The first one that I find is right at the roundabout about 500m to the south. There, I can traverse to the eastern cycleway and follow that north to Baan nr. 90. OK, I have a solution. Not exactly short (about 1km instead of the 50m in my previous attempts), but at least it is legal. Reality: not entirely legal (at least at the roundabout you should go around it instead of traversing before the roundabout), but close enough for me. With the slightest bit of common sense I wouldn't follow that anyhow. What I would do is traverse N47, follow the western cycleway and traverse again when I am at Baan nr. 90. Why didn't the algorithm propose that? Simple: the road to the east of N47 are not connected to the western cycleway and vice verse. So we have to modify the data again. At every point where there is a sideroad from only one side we need to add a short cycleway to connect the cycleway on the other side with the junction.
Once those junctions are made, I will get the obvious, correct route. Conclusion: lots of work and near impossible to maintain. Now, let's see what happens if I tag the cycle paths on the highway=primary in stead of drawing them separately. It is in any case a lot less work. No need to draw the separate cycleways and no need to add all the technical tags on both highway=primary and highway=cycleway that I described previously to get correct results. I just add cycleway=lane or something similar to the highway=primary. What does the algorithm say? I will just say: "At the end of Macharisdreef turn left on N47. I know there are are cycle lanes so you should follow those instead of cycling in the middle of the road. And after 50m you turn left in Baan nr. 90.". Simple, clear, robust. That is why I only will draw separate cycleways if there really is no other option. Even if it is not wrong to draw to the cycleway separately, it is just a lot of work, impossible to maintain and a huge source of errors waiting to happen. Bonus question 1: what happens with routing for pedestrians in both situations? Bonus question 2: in how many ways is it possible to make mistakes when mapping cycling routes? Especially the case of a route that can be followed in both directions is enlightening. Bonus question 3: which situation has the least tags (=lowest database size) and the least junctions and ways (=greater efficiency for routing algorithms like Dijkstra)? wouter On Tue, May 13, 2014 at 9:16 AM, Jo <[email protected]> wrote: > We have more recent AGIV imagery now. All I see wrong is that the cycleway > is connected to the underground waterway. But the main road is too. Probably > to silence validator warnings in a totally inappropriate way... > > I still think it's correct to draw the cycleways separate cases like this. > > Marc, you opened a can of worms there :-) But it's good that the subject of > using separate ways to represent lanes is brought up on the list. > > Jo > > > 2014-05-13 8:15 GMT+02:00 Wouter Hamelinck <[email protected]>: > >> Wow, based on the Bing images a simple T-crossing and a bypass is all >> there is in reality. >> Also, note the nice examples of about everything that can go wrong >> when drawing parallel cycleways along the N47. >> >> wouter >> >> 2014-05-13 6:35 GMT+02:00 Jo <[email protected]>: >> > Zeker wel, het klopt niet om een aparte weg te tekenen voor elk rijvak. >> > >> > Absolutely, using a separate way to represent traffic lanes is not how >> > it's >> > supposed to be done. >> > >> > Jo >> > >> > >> > 2014-05-13 5:48 GMT+02:00 Marc Gemis <[email protected]>: >> >> >> >> Hallo, >> >> >> >> Ik vraag me af of het OK is het volgende kruispunt te vereenvoudigen >> >> via >> >> turn:lanes : http://osm.org/go/0Ejo5_fqb--?way=51738440 >> >> >> >> I wonder whether it's ok to simplify the following crossing with >> >> turn:lanes tagging: http://osm.org/go/0Ejo5_fqb--?way=51738440 >> >> >> >> met vriendelijke groeten >> >> regards >> >> >> >> m >> >> >> >> _______________________________________________ >> >> Talk-be mailing list >> >> [email protected] >> >> https://lists.openstreetmap.org/listinfo/talk-be >> >> >> > >> > >> > _______________________________________________ >> > Talk-be mailing list >> > [email protected] >> > https://lists.openstreetmap.org/listinfo/talk-be >> > >> >> >> >> -- >> "Den som ikke tror på seg selv kommer ingen vei." >> - Thor Heyerdahl > > -- "Den som ikke tror på seg selv kommer ingen vei." - Thor Heyerdahl _______________________________________________ Talk-be mailing list [email protected] https://lists.openstreetmap.org/listinfo/talk-be
