On Monday 13 October 2008 22:41:36 you wrote: > Matias, > > > Is there anything inherently wrong with this system? > > No, it is just another way to do it. If you feel it suits you better > than the existing schemes, use it. > Excellent!
> > I wrote it thinking about ease of editing and efficiency both in storage > > and for use by software, most schemes I have seen are bad in two or more > > of those. > > You will not find a scheme that has no problems. One big problem with > yours is that if a way is split in two, then things become veery > difficult; you basically need an editor plugin that then splits the > relation into two as well, with the nodes from one half of the way in > one relations and the others in the other. Splitting ways is a > relatively common operation I think. > Hopefully it's relatively fixable. It's easily fixable by a bot if need be (by removing from the relation the nodes not belonging to the way, for example). But yes, splitting would be an issue with this system. Since in my experience numbering is the last thing one does on a street, it shouldn't be much of a problem. > Also, you are re-using the existing nodes in a street. This means that > if someone moves these around to improve the shape, the house numbers > move with it which might not be desirable. - What happens if you have an > extra node beside the way (e.g. a pharmacy) for which you want to add a > house number? > These are street numbers, numbers belong to a street here in Argentina, and nodes in the relation should belong to the way (usually only nodes in an intersection would get a number). > With the current state of editors, I cannot see the "ease of editing" > you talk about. > Well, unlike what seems to be the case in Germany (at least I assume so given the Karlsrue schema), here the numbers are associated to the street, and this relation would keep that association. Clicking on a way would display the relation (this is already done in JOSM and Merkaartor, I believe), so you can jump to it and change/add numbers. > As for the efficiency in storage, I suggest you take the long-term view: > I am 100% sure that at some point in the future, OSM will have at least > one node for every house, more likely a building outline for every > house. Look at this if you don't believe me: I fully believe you, I don't think it makes any sense in residential areas, though, and in my country it will take years before that's the case (only a handful of cities have /any/ streets). I would say efficient storage is always worth considering, if it won't hurt other priorities. In any case, I designed this mainly because of ease of editing, I tried a method using one relation per node and it was basically a mess. > At this point it will be trivial (easy to edit, easy to handle, and > requiring little extra storage) to simply add a house number tag to > every one of these buildings. Any sort of complex relations for roads > with interpolation rules for house numbers will then simply be > unnecessary. Your scheme is not suitable for giving a number to every > single house on a street (of course it *could* be done but would become > very cumbersome), so you will have to phase it out slowly and replace it > by direct tagging. > Yes, indeed, it could be replaced by direct tagging and I don't think they interfere at all. Of course my schema can be used to add numbers to an entire city in one day. In my country it's very common to say "<street name> at 2100", ie, a 100 meter long stretch of road. Without street numbers that means OSM can't be used for finding a place, one has to ask between which street and which street, which is often not known. > The Karlsruhe schema which I use surely has its flaws as well, but the > "long term view" is already built in there - you can tag individual > houses now *or* you can do interpolation, and there's a (IMHO) rather > seamless transition. The schema is widely used already (325k house > numbers tagged world-wide), and we'll soon also have a cool debug tool > for it that helps you spot problems (or omissions) in numbering. > I have read how the Karlsruhe schema works, but it doesn't apply well to our case, I think. It might be useful as a complement one day. > But of course, you can try out your own scheme as well. With a little > programming effort, every one of them can be converted into the other > anyway. Good, I just didn't want to break anything or step on someone else's toes :-) _______________________________________________ talk mailing list talk@openstreetmap.org http://lists.openstreetmap.org/listinfo/talk
