On Wed, Nov 13, 2019 at 06:54:52PM +0900, John Willis via Tagging wrote: > > On Nov 11, 2019, at 6:15 PM, Martin Koppenhoefer <dieterdre...@gmail.com> > > wrote: > > A relation seems easier to evaluate and explicit, while a spatial query > > heuristic will inevitably fail in some cases > > > I think there is a need for a basic relation, if I understand Martin > correctly, to simply associate the two lines, (for example, an =embankment > and an =embankment_base pair). When mapped, they are not joined. They are > merely adjacent. I am not sure of what “type” of relation to choose in iD, > but I assume someone will tell us which type to use. > > When mapping a simple cutting or embankment, you would have only one “base” > line adjacent - so there is little ambiguity, and the relationship can be > inferred (IIUC), but in complicated tagging, there could easily be a > situation where which base belongs to which line is unclear, and lead to > problems. > > Simply putting them into a relation says “these members are related” and the > renderer can know for certain that these two ways that don’t share nodes are > a pair, no inference needed.
perhaps as a last step to perfection we might need some relations. Otoh quite pragmatically - what is the use of associating/relating those two lines (base and top)? Do we map them to make it clear if you run there you fall down a cliff or earth bank or run into a cliff? No need for relations for that. Also I am not convinced there is always a one-one relation between a cliff base and cliff edge? If we really wanted to render the "slope/cliff area" in some special style we would probably have to map that as an area, not relation of two or more lines. But I think for small slopes, the top and base lines if rendered should be good enough and for high slopes/cliffs DEM derived countour lines would be better. Btw as we get into more details we might want to map ramps as well. > This again raises the question of levees - is the levee worthy of it’s own > levee relation? do you put all 4 embankment lines into relation with the > man_made=dyke line? this seems to be the only solution to: > > - properly group the embankments with the levee > - not have to use super=relations (putting the embankment relations into a > levee relation) > - providing the most flexibility to weird situations > - allowing for the extent of the top of the levee to be defined (large levees > have varying width tops with usable areas, as shown, in which a “way” is > insufficient ). We use man_made=bridge (area) to group ways on a bridge.. so I am wondering would man_made=levee to encompass the whole levee area work in an equivalent way? I think it is only a quirk of OSM history that dyke and embankment are linear features and we would do many things differently today - and maybe we should do it now. Also somewhat related, waterway=dam can be either linear (the crown of the dam) or area. I think we should have one tag for the crown of the dam and one for the area because it would be often useful to map both of them. Richard _______________________________________________ Tagging mailing list Tagging@openstreetmap.org https://lists.openstreetmap.org/listinfo/tagging