On Sat, 21 Mar 2009, Richard Weait wrote:
> > My understanding of Karlsruhe schema is imperfect, but I believe that > interpolated addresses require a way rather than two nodes? If so, and > if we address a "curvy" block, then the address-interpolation-way may > cross the block twice or more. I don't know if that is fatal to > addressing, but it would look suboptimal. Given this situation, is it > possible to add the interpolated way by: > > duplicating the road / block > clip four meters from each end > offset from the road by two meters > apply address interpolation data as appropriate. Having the interpolation way follow the shape of the road seems to have precedent. See http://www.openstreetmap.org/?lat=49.028355&lon=8.361898&zoom=18&layers=B000FTTT > This does triple the db size (node and way count) for each road with > address info. But it's just Canadian road data, it's not like we're > tripling the size of the US data.... Tripling the number of nodes+ways in the DB isn't something we want to do lightly. The Canadian DB will be a lot larger once all of the NRN roads are imported, tripling that number might mean a difference of hundreds of megabytes. It also isn't just Canada, recent TIGER updates have included address information (I suspect block face based). If we know the coordinates of a house/building I don't have a problem with creating a separate node for it. If all we are doing is offsetting the road geometry by 4 meters I'm not convinced of the value of storing all those fake nodes the renders can apply a 4 meter offset just as easily as the import scripts can and we don't triple the database size. Another thing to consider is what happens when some adjusts/alters the road. If they are doing it based on GPS trace from driving the road the won't have a direct basis for shifting the address nodes, but if the address nodes are derived solely from the road then they probably should move as well. Part of me prefers a scheme that adds address tags directly to the nodes that make up the road. The downside of that is * It isn't karlsruhe (I don't see a lot of objection to karlsruhe and its being used; that kind of is like consensus) * The renders might not support rendering the house numbers to the to the side of the road * An import script that adds tags to existing nodes is harder to write than one that creates new nodes * Intersection nodes would have multiple addresses But it would mean a lot fewer nodes+ways in the database. > > Wonderful news that you are looking at this. I haven't yet decided what I'll work on next, it might be this, or it might be trying to match up roadnames from statscan. > > Best regards, > Richard > > > _______________________________________________ Talk-ca mailing list [email protected] http://lists.openstreetmap.org/listinfo/talk-ca
