Re: [talk-au] Wither Sydney suburb boundaries?
On 28 April 2014 14:23, Michael Gratton m...@vee.net wrote: So you are saying the ABS suburb boundaries should be checked individually rather than imported en mass? How do you know that the quality of the GNB/Wikipedia/etc data is any better than that of the ABS dataset where they disagree? Yes. Firstly, the ABS data is several years out of date. The GNB is the authoritative source for whether a suburb exists or not. Comparing several sources - council, gazette, ABS, GNB - if they concur, then you've probably got something accurate on your hands. If they disagree, then further research is in order, and it is possible that there may not be an open and accurate data source for that suburb boundary. Take the opportunity to fix Wikipedia at the same time. Coastlines, waterway boundaries are also an issue. On a related note, what's the appropriate way to map suburb-sized areas that are partitions? A way for each suburb that share nodes along common borders, a way for each suburb that don't duplicate nodes along common borders, or using a single way for the border and using a relation? Yes. Hopefully we will have access to the accurate and official suburb boundaries in Sydney in an open format sometime in the future (like we have for other cities). Then this problem will go away. I literally just heard back from LPI about my enquiry: LPI is currently reviewing the licencing framework and will be in a better position to answer your query within the next 4 to 6 weeks, so maybe we will have the data sooner rather than later. I've asked if I can make a submission to whoever is doing the review, will post the details there if I get them so we can canvas for a good outcome. If we have the definitive suburb boundaries, then yes, they are definitive. We already have the definitive information for a number of Australian cities, and this is goodness. The ABS data has numerous problems, right around the country. From inaccuracies to completely fictitious regions. It's made for stats, and I'd be opposed to a blind import, and I think I can back up that position. Ian. ___ Talk-au mailing list Talk-au@openstreetmap.org https://lists.openstreetmap.org/listinfo/talk-au
Re: [talk-au] Wither Sydney suburb boundaries?
On Mon, 28 Apr, 2014 at 4:17 PM, Ian Sergeant inas66+...@gmail.com wrote: On 28 April 2014 14:23, Michael Gratton m...@vee.net wrote: So you are saying the ABS suburb boundaries should be checked individually rather than imported en mass? How do you know that the quality of the GNB/Wikipedia/etc data is any better than that of the ABS dataset where they disagree? Yes. Firstly, the ABS data is several years out of date. The GNB is the authoritative source for whether a suburb exists or not. Comparing several sources - council, gazette, ABS, GNB - if they concur, then you've probably got something accurate on your hands. So how accurate does it have to be? For example, I just downloaded Andrew's ABS OSM converted datafile (thanks Andrew!), loaded it into JOSM, and have been eyeballing the differences for the ABS version of Randwick with the LPI cadastre using in SIX Maps. It's confidence is rated very good and I can see that the ABS data matches quite well, but it's missing fine details such as where the suburb falls entirely one one side of road rather than the centerline, and some corners are a bit off. Does this matter? //Mike ___ Talk-au mailing list Talk-au@openstreetmap.org https://lists.openstreetmap.org/listinfo/talk-au
Re: [talk-au] Wither Sydney suburb boundaries?
On 28 Apr 2014, at 10:48 pm, Michael Gratton m...@vee.ne So how accurate does it have to be? For example, I just downloaded Andrew's ABS OSM converted datafile (thanks Andrew!), loaded it into JOSM, and have been eyeballing the differences for the ABS version of Randwick with the LPI cadastre using in SIX Maps. It's confidence is rated very good and I can see that the ABS data matches quite well, but it's missing fine details such as where the suburb falls entirely one one side of road rather than the centerline, and some corners are a bit off. When I'm doing a suburb boundary I think it's minimally important that each property is in the correct suburb. So sometimes that makes the road the boundary. At other times the property line not facing the road. I don't hesitate to modify the ABS data accordingly. What we are mapping is our very best estimate of where the suburb boundary lies. This is especially true for coastline. I think this is the best we can hope for with the data we have available to us today. Ian. ___ Talk-au mailing list Talk-au@openstreetmap.org https://lists.openstreetmap.org/listinfo/talk-au
Re: [talk-au] Wither Sydney suburb boundaries?
On 28 Apr 2014, at 1:53 pm, Michael Gratton m...@vee.net wrote: On a related note, what's the appropriate way to map suburb-sized areas that are partitions? A way for each suburb that share nodes along common borders, a way for each suburb that don't duplicate nodes along common borders, or using a single way for the border and using a relation? I might express and opinion about suburb mapping as I’ve done a fair bit of “mapping for the validator” which I suppose is not evil, unlike mapping for the renderer. I’d prefer relations that depend on single ways, this avoids JOSM complaining too much about duplicate ways and can also tie into the definition in words that might belong in Wikipedia. If appropriate ways do not exist, then create ways can be untagged or have a “ref=“ tag to indicate what they mean e.g. “Centreline of Smith Road” or “Southern side of Smith Road” etc that corresponds to their actual definition. Then build the relation (suburb) and super-relation (Postcode, LGA area) etc on top of these. As to the type of relation as “boundary” or “multipolygon” I’ve still not figured out which is best. Alex ___ Talk-au mailing list Talk-au@openstreetmap.org https://lists.openstreetmap.org/listinfo/talk-au
Re: [talk-au] Wither Sydney suburb boundaries?
On 29 April 2014 11:02, Alex Sims a...@softgrow.com wrote: I’d prefer relations that depend on single ways, this avoids JOSM complaining too much about duplicate ways and can also tie into the definition in words that might belong in Wikipedia. Yes. I general I do too. However, we should only use the way when it does represent the boundary, not because it happens to physically coincide with it. Usually this is apparent from the data. If appropriate ways do not exist, then create ways can be untagged or have a “ref=“ tag to indicate what they mean e.g. “Centreline of Smith Road” or “Southern side of Smith Road” etc that corresponds to their actual definition. Then build the relation (suburb) and super-relation (Postcode, LGA area) etc on top of these. I agree we can build the relations on these. Super-relations aren't well supported, and I see no need for them here. The LGA should be separate relation utilising the same ways. This is the only way I've seen it done, and it works well. Postcode relations? Well, if you are keen. However at least as far as Sydney goes, each suburb belongs to a single postcode, and I think it works well to just be a appropriate tag on the suburb relation. As to the type of relation as “boundary” or “multipolygon” I’ve still not figured out which is best. No winners here. Even the validators disagree. I've been known to use both. Ian. ___ Talk-au mailing list Talk-au@openstreetmap.org https://lists.openstreetmap.org/listinfo/talk-au
Re: [talk-au] Wither Sydney suburb boundaries?
I have intentions of following the British structure for QLD boundaries (no permission to use this dataset yet). Boundary is the chosen type there: http://www.openstreetmap.org/relation/1464290 multipolygon, though, is winning that race it seems: http://taginfo.openstreetmap.org/keys/type As to the type of relation as “boundary” or “multipolygon” I’ve still not figured out which is best. No winners here. Even the validators disagree. I've been known to use both. ___ Talk-au mailing list Talk-au@openstreetmap.org https://lists.openstreetmap.org/listinfo/talk-au
Re: [talk-au] Wither Sydney suburb boundaries?
On 29 April 2014 12:56, Jason Ward jasonjwa...@gmail.com wrote: I have intentions of following the British structure for QLD boundaries (no permission to use this dataset yet). Boundary is the chosen type there: http://www.openstreetmap.org/relation/1464290 multipolygon, though, is winning that race it seems: http://taginfo.openstreetmap.org/keys/type I don't think that indicates multipolygon type is more popular than a boundary type for a relation defining a boundary. Multipolygons have extensive usage independent of boundaries. I suspect the only reason that boundary is even in the race, is due to some large imports. The issue stems from overloading of the meaning of 'type', in the relation definition. Multipolygon uses 'type' as defining the geometry of a relation. This appears to be the original usage. And a boundary clearly has this geometry. So, if 'type' refers to geometry - as it did originally - then boundary should never have been used as a type, and it should just have been a multipolygon. This usage isn't apparent from the word 'type', and you can see why the person who came along and created boundaries thought they were a of a different type, even though they were of the same geometry. And if type was meant to define the geometry, then it would have been a idea to use a different tag originally. Now we have type being used with both meanings. Ian. ___ Talk-au mailing list Talk-au@openstreetmap.org https://lists.openstreetmap.org/listinfo/talk-au