[OSM-talk] Yet another street number scheme

[Elena of Valhalla]

On Thu, Oct 16, 2008 at 7:28 AM, Karl Newman <siliconfiend at gmail.com> wrote:
> [...]
> You might call it a workaround, but it's certainly a common approach. Many
> major GIS systems (including the TIGER data set) use this exact methodology.
> Not including this information will make OSM data less useful on these
> "legacy" devices.

To get osm maps on these devices one would need a converter script
anyway: creating a script that checks for a few common interpolation
rules to compress the informations and then converts the numbers on
buildings to a few additional nodes on the ways should be quite
feasible; it may be slow, but it is an one-time operation for each
dataset

the reverse just can't be done, because it would need to create
information out of nothing [1]

> [...]
> Just to reiterate my perspective, the Karlsruhe schema is fine for what it
> is, but it's not sufficient for all uses. You can pretend that address
> numbers don't belong to the street they're on (!)

yes, they don't _belong_ to the street, they are _related_ to the street

actually I believe that here in italy numbers belong to openings
(doors, gates, etc.) on the street, but since we don't (currently)
have a way to mark those on the map, the building they are on is a
better aproximation of what exists in reality

>  but there are a ton of
> existing navigational devices and software (probably all of them) that do,
> in fact, treat them like that. [...] It's a reality
> we should accept and accommodate.

Indeed it is a reality we should accomodate: not in the dataset but
with proper conversion scripts

adapting the dataset to compression tecniques used in current
navigation devices would be mapping for the navigator, and that is
just as bad as mapping for the renders

adapting the dataset to something that is both accurate and easily
_converted_ to the needs of navigators, renders and other uses is a
better approach

> I mean, really, are we trying to guide a smart bomb to someone's doorstep?
> For navigational devices, locating the spot a relative distance along a
> street is exactly what users expect. They're not going to drive to the front
> door. (I'll grant you that there are other uses for that data than just for
> navigational devices.)

renders for areas where buildings are individually marked however are
probably better suited by a nice number to place on it as a label, and
that does not work with the interpolation method

also, what if I want to create a robot that uses osm data to navigate
to someone's doorstep and deliver flowers and chocolate?

[1] with one exception: if the local rule mandates that numbers are
the distance from one point on the road, the interpolation method does
not really lose lots of informations, except near crossings (on which
road does a corner building relate to?). Here in Italy it this method
is used only for buildings on primary roads outside urban areas, and i
suppose that at least some of them is big enough to need more
informations than a simple outline to know the number

-- 
Elena of Valhalla

homepage: http://www.trueelena.org
email: elena.valhalla at gmail.com