On 26/11/2009, at 11.13, Jean-Marc Liotier wrote: > Morten Kjeldgaard wrote: >> On 25/11/2009, at 14.11, Jean-Marc Liotier wrote: >> >> The "map-drawing approach" is valuable in OSM because it allows us to >> indicate residential areas parks, etc. However, in addition, OSM has >> a graph-based approach for a description of the network of roads >> which makes it *uniquely* valuable. Graphs prefectly represents the >> road map and can be used for many applications, routing is an example >> that many people use daily. > > "Graph" is the word I was looking for... Thanks for introducing it to > the debate. Indeed the graph (nodes+edges) is the simplest way > to model a network. In modeling, the simplest way is likely to be > the best. > > But if we map everything as an area, do we lose the ability to perform > graph calculations ? Can't an area be considered as a set of edges > connecting all nodes inside it ? > >> In my view it would make much more sense to work on a more expressive >> (perhaps BNF based?) tagging scheme. This would enable a gradual >> enhancement of the map, where the new tagging syntax could live >> along-side the old. > > From my OpenStreetMap novice point of view, modeling ordered sub-ways > inside a highway, each with its own set of tags would go a long way > toward removing the need to model ways as areas. Special cases would > remain, but if for a given way I can define the order of sidewalk, > bicycle lane, bus lane, car lanes, separators and whatever else, each > with speed limit, width and various other tags, I barely see the need > for area mapping of ways. > > Is there any problem with this approach ? It would introduce hierarchy > and ordering, but it would reuse all the existing tags and remain > compatible with the existing scheme. Notice that introducing hierarchy > and ordering fits the existing OpenStreetMap XML schema quite > naturally: > all that would be needed is to nest a <way> inside a <way> - except > that > the nested <way> would have no <nd> but only tags.
I'm no big XML expert, but I do know that it is an intrinsically hierachial language. However, it seems to me that the original designers of the OSM XML schema went through quite some effort to flatten it by using attributes and make it consist of "lists of things". That has the great merit of making the parsers much simpler. Thinking about it, I think it's probably not realistic to change this in a fundamental way, so in that sense I have to rescind my former comments about a BNF formulated language :-) However, it is possible to achieve an hierachical data-structure using lists of lists, and so in that way I think there's more promise in the discussion about the "multiplex objects" in another sub-thread of this thread. You yourself actually suggested something similar AFAICS. Cheers, Morten _______________________________________________ talk mailing list talk@openstreetmap.org http://lists.openstreetmap.org/listinfo/talk
