[Tagging] Housenumber interpolation with regularlyskippednumbers

Anthony osm at inbox.org
Tue Oct 13 18:33:46 BST 2009


On Tue, Oct 13, 2009 at 12:31 PM, Mike N. <niceman at att.net> wrote:
>>> TIGER obfuscates the data by declaring the entire numbering range of a
>>> zone: for example a "400 block / Even" containing houses 404 through 420
>>> would be declared as "range Even / 400-498" in TIGER.   For navigation
>>> purposes, that gets you to within one block of an address.
>>
>> Maybe they do it for obfuscation, but that has the additional
>> advantage of being able to locate an approximate address when house
>> #422 (or #402) gets added to the block.  Of course, we don't have to
>> be quite as dumb as Tiger.  We could always use three blocks,
>> 400-404/Even, 404-420/Even, and 420-498/Even.
>
>  Why use 3 blocks?    If a cursory survey shows that 404 and 420 are
> physically the endpoints of a block, why not use a single way?   Even if 420
> is not the physical endpoint, why not a single way?

Because the area from 404 to 420 is unlikely to be 16/98 the size of
the area from 400 to 498.

> What is a house number after all, if not street-level data?

It's generally postal service data, at least in the US (I believe
outside the US as well, but I'm not 100% sure).

> The house
> number has no meaning to the physical building if not attached to a street.

When I lived on campus at the University of Massachusetts, my address
was "1403 Washington".  Washington was not the name of a street, it
was the name of the building.

Furthermore, even when an address does point to a street, you don't
always get to the address by using that street at the closest point of
the street to the address (and sometimes, you don't even use that
street at all).

>> And that's just the easy case, when you're not trying to combine data
>> from both schemas on the same block (I'm not sure that any of these
>> have been mapped yet, but imagine a rural area with lots of houses
>> near the road, some houses far off the road in flag lots with long
>> driveways, and some houses both on and off the road in various stages
>> of development and not yet assigned addresses; or try to combine
>> actual addresses and potential addresses on a road in a retail area
>> with lots of strip malls with individually addressed stores; or a road
>> with lots of apartment complexes/condos with individually addressed
>> apartments/condos).
>
>   I would never use the Karlsruhe Schema ways to determine a house/building
> location.   There can be many good reasons to use address interpolation when
> the building location is unknown - no aerial photographs, blurred or
> obstructed aerial photos, new construction,  etc.

Just because you're using address interpolation doesn't mean you have
to use the Karlsruhe Schema, though.  If you have no idea where a
house is other than it's relative location on a street, you shouldn't
use the Karlsruhe Schema.  You shouldn't randomly tag locations away
from the street, if all you know is a location on the street.

>> For clarification, the direction for the purposes of right/left would
>> be determined by the start and end node, not the direction of the way.
>> The way could be reversed without breaking anything (and not all the
>> ways have to even go the same direction).
>
> Now I'm confused.   Unless the street is one-way, the only way to find
> the start and end node is to go into edit mode. Streets can be oriented in
> any direction, so left/right is often not useful for physical representation
> on the map.

My suggestion was to use a relation, not to tag the ways directly.
Tagging the ways directly wouldn't work - an interpolation might cover
more than one way.




More information about the Tagging mailing list