Re: [Maposmatic-dev] Nominatim released

[Emilie Laffray]

2009/12/8 David Decotigny <address@hidden>

Hello,

Emilie Laffray wrote:

I wouldn't trust the osm ids that you are getting for the street. The system uses a voronoi diagram and therefore it may be of worse quality than using administrative boundaries when they exist.

Hmmmm... I'm not sure I understood. Do you mean that the polygon we get is Voronoi ? Or that the osm_id="xxxx" or place_id="yyyy" attributes are purely meaningless ?

What I mean is that you have two possibilities in terms of polygons you are getting:
This one http://nominatim.openstreetmap.org/details.php?place_id=49962699
and http://nominatim.openstreetmap.org/details.php?place_id=10659

The second one is behaving weirdly due to the way it is generated. It is based on hierarchical model. Both links will be giving you proper osm ids but I am not sure to what extent I will be trusting the result of the second one. This is an extreme case however. Let's take a more interesting example.

http://nominatim.openstreetmap.org/details.php?place_id=49994199
http://nominatim.openstreetmap.org/details.php?place_id=41832

In the first case, I have 427 residential ways. In the second one only 261. 
This one from my home town is more interesting. I am not sure that the streets are matching. As you can see the second one has a polygon that is "weirder", which is actually a result of a Voronoi diagram. There will be plenty of spaces where no administrative area will be available and you will be "stuck" with the voronoi polygons, giving you some less accurate.
Else the osm ids are fine, and usable.

Maybe if maposmatic is using nominatim, then we should focus on getting administrative and then place if administrative doesn't exist.

Emilie Laffray