On Mon, 2009-01-12 at 09:03 +1100, Ben Kelley wrote: > Does anyone have any thoughts on how to mark suburb boundaries (in > areas that have them)? > > The closest thing I can find is boundary=administrative at > http://wiki.openstreetmap.org/wiki/Key:boundary but I haven't seen > this used anywhere. London uses this to mark boroughs (equivalent to > council areas) with left:district=name and right:district=name to > indicate the names on either side of the way. > > The above page seems to indicate that admin_level=10 shows a suburb > border in Australia. Has anyone used this tag? How do you show the > suburb names? Are there any examples of how this renders?
There seem to be three approaches in common use. The first (I think) approach was to draw an area around the suburb's entire boundary tagged with place=suburb and place_name=<the suburb's name> (and possible also name=<the suburb's name> - the Wiki is ambiguous on this) and then also to put a node at the suburb's town centre, tagged with place=suburb and name=<the suburb's name>. It's also helpful to add a couple of other tags to the central node: is_in to describe the state & country the suburb belongs to & postal_code to describe the postcode(s) in that suburb - not sure whether or not it's helpful to add those to the area as well. See http://wiki.openstreetmap.org/wiki/Key:place and http://wiki.openstreetmap.org/wiki/Tag:place%3Dsuburb That approach has the advantages that the tagging of town/suburb boundaries and their town centres is consistent, and that the boundary is defined by a single area (the minimal data required to define it). And also if you know where the town centre is but don't have authoritative boundary data, you can still place & tag the town centre node, and hope that someone will come along later who does have authoritative boundary data and put the associated area in place. It's also the approach that requires the least amount of post-processing for search. There's a German example on the Tag:place=suburb page on the Wiki. For a local example, see Glandore in Adelaide's inner southwestern suburbs. Personally I think that is still the best approach (the only downside I can see with it would be if a suburb was not defined by a closed area - although I'd imagine that would be quite rare). However, you'll find plenty of others that prefer one of the other two approaches. The most recent approach centres around using relations and having each way that constitutes a part of the suburb's boundary a member of a relation that defines that boundary. The name is then a property of the relation rather than of its constituent ways. That works quite well when the boundaries are roads, riverbanks, coastlines or other physical things that would require individual ways on the may anyway and the relation just groups them together to provide a logical boundary. But when the boundaries (or more often, parts of them) are just imaginary lines, creating multiple ways just for a boundary, then grouping them together as a relation seems like an awful lot of double handling (both for the mapper putting them in the map and for any automated process trying to reassemble them for any useful purpose). Darrin's mapped most of Adelaide's nothern suburbs using this method, and that's probably the best Australian example of using relations for suburb boundaries (as well as postcode & local government boundaries). The middle approach is the one you identified, tagging ways with boundary=administrative, admin_level=, left:suburb= & right:suburb= (etc. etc. for other boundary types). Of the three, that approach seems to make for the best looking paper maps, but seems the least useful for search (i.e. the greatest amount of post-processing needs to be done for search when this method is used). I haven't seen it in broad use in Australia [everyone else: please show a good Aussie example for Ben if I'm wrong on that], but in a handful of places where there are ways without other tags [that define physical characteristics] used in the relation-based approach, I've added this approach in parallel, to make it easier to identify what the ways are there for (i.e. without having to look up what relations they belong to) without breaking the relation-based approach. As Darrin pointed out, that method is dependent on the direction of the way. However, all recent releases of JOSM will offer the choice to swap left:/right: tags whenever you reverse a way that has them. Methods 2 & 3 still require a node at the town centre (just like method 1), tagged with place=suburb & name=whatever, since the town centre is often not at the geometric centre of a town/suburb's boundaries. At this point in time, I guess it's really a matter of just picking whichever approach (or combination of approaches) you like best. If others mapping your area don't like the one you've chosen, they'll just add one of the others in parallel - as far as I can tell none of the three approaches break any of the others. On a related topic (which Franc also mentioned), finding good non-copyright-encumbered data sources for mapping towns/suburbs can be interesting too: Placing the node for the town centre is fairly easy - a quick trip there (or recent memory of one) should give you a very good idea of where the town centre is. In some places, the council even helps with some signage (I've always been amused that in Canberra, there's a little sign that reads "City Centre", in case you didn't notice that's where you were!). Worst case, you can always put it at the geometric centre of the town/suburb and hope that someone with better local knowledge will come along and move it to the real town centre. Ascertaining boundaries where they're defined by physical features (road, river, etc.) is also fairly straightforward - just know someone who lives (or visit a business & get a business card or receipt with an address on it) close enough to each side of the estimated boundary to get sufficient confirmation. But surveying those "imaginary line" parts of boundaries, particularly in areas where there are no houses or businesses close enough to the estimated boundary to be authoritative is a bit more problematic - I haven't come up with a good method yet; perhaps someone else on the list can suggest one? (the Government - including Aussie Post - published data all appears to be encumbered). Regards, Jack Burton <j...@saosce.com.au> _______________________________________________ Talk-au mailing list Talk-au@openstreetmap.org http://lists.openstreetmap.org/listinfo/talk-au
