Let's not get carried away with what Cloudmade does and does not do. Nick Black has mentioned on talk less than a year ago exactly what their two largest routing bugs were. Other than that they are a service provider, who doesn't publish anything (no testing framework, not even devices with firmware that can be disassembled). So tagging for them is certainly not on.
Using a big penalty like Appolinaris mentioned may at some point break, possibly in a rural location. Different routers will end up with different penalties so there may even be a few locations where some break and others do not. I like Alan's idea, but only as a temporary measure until the editors (esp. Potlatch) makes it very easy to get things right. Note that in Allan's case there are two sharp left hand turns that are illegal and dangerous. One onto the trunk and one onto the trunk_link, So if you use an only_ restriction, then you need only one restriction. http://nroets.dev.openstreetmap.org/demo/?lat=45.42132&lon=-122.32352&zoom=15&markers=!45.41879,-122.32406!45.41824,-122.31106&v=motorcar&fast=1&layers=B000FTFTT I've posed the same question on help, so now we can vote on it: http://help.openstreetmap.org/questions/275/how-do-i-connect-a-high-speed-bidirectional-trunk-road-to-the-network Regards, Nic On Sat, Jul 17, 2010 at 12:43 AM, Tristram Gräbener <tristr...@gmail.com> wrote: > They say on their webpage that cloudmade uses Contraction Hierachies. > If they used the plain algorithm, then there is no such stuff as > penalities and so on (of course I have no idea if they tweaked it). It > wouldn't suprise me that they chose high performance routing (1ms per > calculation) over high-precision routing with tons off human-made > tweaking and setting to give a more natural feeling. > > My personal opinion is that the algorithm should be simple and the > effort should be put on the data. Other people prefer to work on > medium quality data and adapt their algorithms (as perfect data is > impossible to get). > > > On Sat, Jul 17, 2010 at 12:35 AM, Apollinaris Schoell > <ascho...@gmail.com> wrote: >> >> On Fri, Jul 16, 2010 at 2:21 PM, Alan Millar <a...@bolis.com> wrote: >>> >>> Hi- I'm a user, not developer, trying to clean up the data in my area to >>> make it more usable for routing and directions. >>> >>> Several times now I have come across a situation that seems wrong, and has >>> not improved by itself in the last year or two as far as I can tell. That >>> is where routers want you to make a U-turn on a highway link. >>> >>> My most recent example is here: >>> >>> >>> http://maps.cloudmade.com/?lat=45.418666&lng=-122.321563&zoom=16&directions=45.41881686582581,-122.32349395751953,45.4324308667588,-122.36778259277344&travel=car&styleId=1 >>> >> >> to make it absolutely correct you will need a no-left turn restriction but >> this looks like a wrong cost function for the router. It's not exactly a >> U-turn just a very bad angle to turn onto the trunk in this direction. >> Garmin has a very high cost for such turns because they require to slow >> down dramatically and default algorithm is fastest. Here it's a clear sign >> for a bad router algorithm or cloudmade has an incomplete data. I still get >> wrong tiles in some zoom levels. So maybe the correct ramp is not yet seen >> for routing. >> >> >>> >> >> _______________________________________________ >> Routing mailing list >> Routing@openstreetmap.org >> http://lists.openstreetmap.org/listinfo/routing >> >> > > _______________________________________________ > Routing mailing list > Routing@openstreetmap.org > http://lists.openstreetmap.org/listinfo/routing > _______________________________________________ Routing mailing list Routing@openstreetmap.org http://lists.openstreetmap.org/listinfo/routing
