[Imports] [Talk-es] [Cat2Osm2] Tool for exporting Spanish Cadastre data in OSM suitable format
penorman at mac.com
Fri Mar 1 23:43:44 UTC 2013
> From: Cruz Enrique Borges Hernandez [mailto:cruz.borges at deusto.es]
> Sent: Thursday, February 28, 2013 2:30 AM
> To: Paul Norman
> Cc: imports at openstreetmap.org
> Subject: Re: [Imports] [Talk-es] [Cat2Osm2] Tool for exporting Spanish
> Cadastre data in OSM suitable format
> 2013/2/28 Paul Norman <penorman at mac.com>:
> >> From: Cruz Enrique Borges [mailto:cruz.borges at deusto.es]
> >> Sent: Wednesday, February 27, 2013 11:36 PM
> > The concerns being expressed are about using multiple ways to
> > represent one building. How big do you estimate the import will be
> > total, both in number of buildings and number of ways tagged as
> We do not know because its not going to be a massive import an will
> depend on the decision that the local mappers that will work on their
> respective areas. However, we estimate that:
> - For every house more or less it would be a parcel. In big cities lots
> of houses have the same geometry as the parcel; in this case only one
> geometry is created. Son It will be less than double the number of ways.
> - We guess that the 3D information will add more or less 3 times the
> usual number of ways in a only houses import.
But some estimate would be nice. 100k ways? 1m ways? 10m ways? For reference, there are on the order of 35-40 million ways from French building imports. I have no solid statistics on the number of buildings that is, but based on a sampling I did, it's at least 8 million buildings, likely 15-20.
> >> > And I suspect that they are going to give you a big headache once
> >> > you get around to mapping addresses.
> >> I don't think so. The address is already in the building or in the
> >> parcel (it's an step of the import process).
> > It would better reflect normal mapping if you had the addresses on the
> > buildings and a conventionally mapped landuse=residential way.
> We have discussed that in the Spanish list and the consensus was that it
> is better to put that information in the parcel because:
> - If there is more than one build, yo what build will you put the
> address information? Will you duplicate the info?
> - If the parcel is big and the build is located far from a road (like,
> for example, in rural areas) we think that it would be easier to the
> routing service to find the entrance=yes of the parcel than to guess hot
> to reach to the building.
> We think that this way is the most useful way of doing, but if you can
> provide us a better solution we happily change the import program and
> the suggested import wiki.
My general practice is to try to make imports like normal mapping. For the cases you've shown so far this would be addresses on the house.
You haven't shown any rural cases so far, but what you're doing can actually make it worse. The common rural case is a lot where the building is some distance from the road. With all geocoding software I've seen, it'll return an area with what you're proposing doing. This could put you anywhere on the lot boundary. The entrance=yes node won't be used for routing. If you have the address on the building, it will put you at the closest mapped road to the building. This may not be where you want to get routed, but the solution is to map long rural driveways. If you do what you're proposing with the import then mapping the driveways won't help this problem.
The issue of rural addresses and emergency routing has been considered by Richard Welty for rural New York state. This is also essentially the same as the choice between representing address points at building entrances, lot centroids or the road-driveway junction.
More information about the Imports