On Fri, 8 Mar 2019 at 18:25, Martin Koppenhoefer <dieterdre...@gmail.com> wrote:
Right, there's this "incompatibility" of the highway graph with the rest of > our data. There are efforts to map roads also as areas though, and sooner > or later this kind of mapping will be established (in built up areas and > particularly where the shape is not regular, so that it actually makes > sense to do it, e.g. historic town centres). > "Sooner or later" is likely to be a lot later for much of the map. For instance, I've mapped several lay-bys (rest areas) from aerial imagery alone. They tend to be on very long roads (otherwise you wouldn't need to pull in for a break). Mapping the full extent of a road with a lay-by as an area would be very tedious, and there are far too many other things as yet unmapped of more importance. > > Your way, we'd have a gap between the parking area and the road when it >> renders. >> > > no, just in the editor. In the rendering the parking area will probably be > on the road ;-) > If you tweak it enough, and check at all zoom levels. Which there isn't because in reality they're conjoined and contiguous. > Your way, the parking area > wouldn't be routeable. > of course you could route to the parking area. Every router does this all > the time. We don't add housenumbers to the road, do we? > Theoretically (perhaps even it's already feasible in some apps) you could route from the house end of a driveway (selected by mouse click). You can't do that with a disconnected parking area unless the router makes guesses. But there could be roads adjacent to two sides of a parking space, only one of which actually connects to it and the router could guess wrong. > generally our roads are thicker than they are in reality, only in the > highest zoom levels this might change to the opposite. It is impossible to > get this right in all zoom levels. > Indeed. But if the parking area connects to the road, it does get it right at all zoom levels. Well, right-ish. One of the reasons the renderer puts things at different z indexes is so that things like this end up looking right (ish) so that the parking space isn't shown as extending to the middle of the road. > routing is not an issue, for the rendering results _might_ look more > "clean" if you extend the parking up to the middle of the road (not that > the representation would be more accurate though), but at the cost that the > parking area will become much bigger than it actually is. > Just about everything in OSM could be done better. Usually with associated costs, like mapping roads as areas, and living with the fact that we don't have the time to do the entire extent of every road that way, so we'd get jarring width changes everywhere. We have to go with the best solution we currently have and live with the fact that the map is a stylized representation. I see extending the parking area to the highway as the best trade-off we have, you disagree. -- Paul
_______________________________________________ Tagging mailing list Tagging@openstreetmap.org https://lists.openstreetmap.org/listinfo/tagging
