[OSM-talk] OSM the mediocre alternative

[Martijn van Oosterhout]

On 4/23/07, Christoph Eckert <ce at christeck.de> wrote:
> Sorry I was not clear enough. I meant that a segment can belong to more
> than one way(S), and then it might be difficult to determine to which
> way the numberings belong to.

Definitly. A segment can belong to any number of ways, which is why
attaching it to segments directly is a bad idea.

> * A house is sited far away of the street which gives it the address
> (found on coutrysides). We will draw a serviceway or track to this
> house, but we will not give the serviceway the name of the street which
> forms the address

Techincally though it's the block of land that has the street number,
not the house. If there are many houses on one block you start getting
subnumbers (house a,b,c, etc). Really you want to assign the number to
the entrance to the block and not worry about where exactly the house
is.

The other extreme you have in Australia is rural numbering, where the
number is determined by the distance along the road from some point.
You don't need to put them in the database at all, just the point
where the numbering is measured from (usually some town somewhere).

> Both cases can somehow modeled by tagging the segs of the streets, but
> if we want to achieve proper routing (and that's IMO the main purpose
> of collection house numbers), we need a smarter solution.

On the whole routing only cares about the section of road closest to
where you want to be, since from there it's by definition a straight
line. So assigning the numbers to the ways adjacent isn't really a bad
solution.

Have a nice day,
-- 
Martijn van Oosterhout <kleptog at gmail.com> http://svana.org/kleptog/