[Imports] Australia - Geocoded national address data

[Vincent de Château-Thierry]

Hi,

Le 26/12/2015 19:11, Pavel Machek a écrit :
> On Sat 2015-12-26 18:49:42, Frederik Ramm wrote:
>> Hi,
>>
>> On 12/26/2015 04:56 PM, Pavel Machek wrote:
>>> Well.. we do address points as points, because one building can have
>>> multiple. Importing it soon looks like a good idea to me.
>>
>> Great care has to be taken with such large import projects and I don't
>> like to hear words like "soon" in that context. Ideally we'd like people
>> familiar with an area to add and check the address points; they would
>> then also notice things like an address point located on water or along
>> a missing road and so on. I'd rather have a slow import that takes care
>> of these issues than a quick "we can still fix later" job that will make
>> Australia look like this
>>
>> http://www.openstreetmap.org/#map=17/57.22356/10.30863
>
> This actually looks good. It is useful to know how many houses are
> there even if you don't know the shapes, and address points are
> extremely valuable.

Address points are of great value of course. But according to me the 
question should be : do millions of address points add value to OSM if 
imported as a (single) big chunk ?
In France we have started 18 months ago a project called "BANO" (= 
National Open Address Dataset)[1]. It is a compilation of address points 
from various ODbL compliant sources. The main source is the french 
Cadastre. We have collected 15M single address points so far.
Since the begining we decided _not_ to import them in OSM, because we 
consider that we are not able to _integrate and maintain_ such a dataset 
from an OSM point of view.
We deliver address point files in multiple flavours (csv, shp) and 
geographical extracts so that every one is able to use them. We also 
provide them at the street level in .osm format for integration within 
OSM but such service has only very few users.

>>> Well, I have a programs that can deduce street names from surrounding
>>> address points. So addresses can help even with that...
>>
>> I'd prever surveyed information over deduced information any time...
>
> And I'd prefer useful map over useless one.
>
> This kind of automated deduction makes it easier to add the names
> (because you don't have to actually type them), but has enough errors
> so that you need to review ever single one.

BANO's value stand in the content itself but also in the focus provided 
for french community on missing streets (geometry and names) in OSM. We 
have now a set of tools to put the focus on missing street names through 
listing and maps such as this :

listing : http://cadastre.openstreetmap.fr/fantoir/#insee=95428&tab=0
map : 
http://tile.openstreetmap.fr/~cquest/leaflet/bano.html#18/48.99569/2.32832
Map and listing above deal with the same area. Red polygons highlights a 
possible missing street name and somtimes a street shape also. It uses 
the address points to compute a convexhull at the street name level.

Such tools are heavily used by the french community. Address point 
integration has only slow usage.

vincent

user:vincent_95
@_vdct

[1] : https://github.com/osm-fr/bano