On Apr 27, 2012 12:43 AM, "Steve Bennett" <[email protected]> wrote:
> Verifiability and objectivity are the principles here - so let's call > them that. "Map what's on the ground" fails in plenty of cases: > - walking/bike routes (what's "on the ground" is a few signposts, nothing more) > - ferry routes (nothing) > - many names, like business names, parks, mountains, etc > - administrative boundaries > - ski runs (minimal signage, and very hard to distinguish without > supporting evidence) > - abandoned railways > - and, if we're being pedantic, tunnels, bridges, overhead cables, etc etc. > > It's a marginally useful rule of thumb that quickly breaks down in any > serious discussion. Don't overuse it. > On the contrary. It is our touchstone whenever we encounter confusion or ambiguity. I'd argue it is useful in all the cases you have mentioned. I think you are interpreting "on the ground" too literally. It means when there is any controversy or ambiguity over what to map, we look to what is physically present at a site right now to help us resolve it. It means what is physically present overrides an image, documentation, import, or any other secondary document. It means we rely on primary research of actually going there and finding out what the situation is to be our final arbiter. Cycle and walking routes get relocated, torn up, fences put across them. Ferry routes are where the ferry actually goes, not where the timetable says it does (although on an open space, some degree of interpolation is always required). Park names? The name that is on the park prevails - quite often different to other sources. Abandoned railways should have evidence of having been there. Tunnels, bridges, overhead cables can all be verified by what is physically present, which overrides any other source. On the ground doesn't mean we can't use other sources. It just means that when sources clash, we defer to what is on the ground. This is in contrast to Wikipedia, for instance, which will use a secondary source to determine what to include. OSM will always choose what is physically present, on the ground. > > Cycle routes are tricky, and we haven't got there yet. Ask three > > different road routing algorithms the for a best route, and expect > > similar answers. With cycle routes, that won't be the case, and many > > different factors need to go into the weighting, and isolating the > > routing factors and their weighting is developing. > > You're talking about something different. There's a difference between > a pre-defined "cycle route", chosen by humans and backed up by some > kind of signage, website, publication or whatever, and a dynamic, > computed route between two places. > I know. I'm simply saying that where no such human defined route backed up by cycle facilities on the ground exists, we should include the features present, not invent a route. > To help people visualise a complete path from A to B to C to D, even > if the "B to C" bit is indistinguishable from any other road "on the > ground". > > For example, see this diversion from the Great Southern Rail Trail > onto the South Gippsland Highway here: > http://osm.org/go/uGumFQcy-?layers=C > > Let's say for the sake of argument that there is no bike lane, no > signage, and the road is a busy one. It's still clearly correct to > indicate the route along that road, because that's where the published > route ( http://railtrails.org.au/index.php?option=com_railtrails&view=trail&id=144&Itemid=15 ) > goes. No doubt it is in the map providers interest to make the route appear connected, even when in reality they may not have made a single change to the facilities on the ground. However, the reality may be in some case that there are actually gaps in the route. I see this very much as an edge case, and I can see arguments for including the connecting segments in an official named route like this one. However, in the case at hand, we're talking about separate cycle facilities that are officially unconnected. When navigating around cities, people may be joining for a section, and maybe utilising three or four different routes in a trip, so finding a connecting path between cycle routes is a common occurrence. > Well, as we all know, everyone maps in different ways. I'd suggest > that any and all of these things are potentially worth mapping: > - physical infrastructure (bike lanes, bike paths, wide shoulders...) > - signed routes > - published official routes > - published official planned routes > - published unofficial routes (in moderation) > > It depends on the local circumstances, the likely audience, the > inclinations of the mapper (and their time availability) and lots of > other things. > > Steve Adding physical infrastructure is always useful. It never does any harm. If accurate, it can either inform route choices, or be ignored. Every bad route we add severely diminishes the value of the good ones, and we see this happening already in Sydney. There is a signed cycle route heading down Parramatta Rd at Croydon. This is a 3 lane, very narrow laned road with heavy vehicles and high traffic volumes. If we mark that in OSM in a cycle route, we may as well give the game away. I've no idea how you allow unofficial routes to be used in moderation. How do you decide which to keep? How do you decide who gets to decide? Would OSM really be a better project if we were to do a data import from bikely? Let's be careful with cycle routes in Australia. The cycle facilities are sparse compared to the bicycle signs and council routes. There is no central coordinating or certifying authority. I'd argue against adding personal routes, add the physical information for a router instead. This information is far more valuable, and as a project it plays to our strengths. I'd also argue against adding other routes where no cycle facilities exist, or the route is dangerous. Particularly where the information is source from council maps that aren't recently updated, or from bicycle signs that point off a main road or cycleway. Ian.
_______________________________________________ Talk-au mailing list [email protected] http://lists.openstreetmap.org/listinfo/talk-au
