[Tagging] Housenumber interpolation with regularlyskippednumbers

Anthony osm at inbox.org
Wed Oct 14 13:39:31 BST 2009


On Tue, Oct 13, 2009 at 10:59 PM, Anthony <osm at inbox.org> wrote:
> I've done some tagging of some different methods of tagging strip
> malls under the current (and slightly extended) Karlsruhe Schema.
> Unfortunately, Mapnik hasn't gotten around to rendering all the tiles
> yet, so I'll wait until later to reveal them.

Okay, here we go:
http://www.openstreetmap.org/?lat=28.07859&lon=-82.506151&zoom=18&layers=B000FTF

The northernmost building is tagged by giving every single unit its
own point.  Just south of that I tagged the building with units
"14445, 14443, 14441. 14425. 14407. 14405. 14401" with
addr:housenumber=14401-14445.  South of that I tagged the building
with an interpolation from 14341 to 14391.  Finally, the southernmost
building is tagged
addr:housenumber=14301;14303;14309;14313;14317;14339.  South and north
of the buildings, and across the street, I put three more Karlsruhe
Schema interpolations next to the highway: 14001-14399, 14501-14899,
and 14000-14898, respectively.

The northnernmost building is the only one mapped strictly correctly
according to the Karlsruhe Schema.  It suffers from the disadvantage
that when units get renumbered or split, no address location will be
found.  Also, it is extremely tedious, and only possible because I
found a handout from the property management (which arguably is a
copyright violation to use?).

The second building was easier to tag.  It doesn't render very well
right now though, and probably will never render well.  Makes
geocoding relatively easy though.

The third building looks the best with the current rendering, and will
geocode just fine with any software which works under the current
Karlsruhe Schema.  I would recommend it for tagging of buildings in
this type of situation.  It wouldn't work quite as well on the first
building, and it doesn't capture quite as much information as the way
I tagged the first building, though.  And it's technically incorrect,
because it includes "housenumber"s that don't exist.  Even if one had
the individual unit numbers, I still think the range information
should be included.  This could be done by adding nodes for the other
units but keeping the way.

The fourth building was an attempt to extend the Karlsruhe Schema to
allow for information about which "housenumber"s were valid and which
were invalid, without resorting to tagging individual units of the
building.  It renders terribly right now, and I'm not quite sure how
it could be rendered better.  Geocoders could work with it well with
simple extensions.

The way across the street is tagged with Tiger-like information, using
the Karlsruhe Schema.  The ways just north and south of the right side
of the street are also done with the same.  I left out the block of
addresses assigned to the four buildings I mapped, but this leaves a
problem.  When those reserved addresses get used, geocoders are going
to have no clue of even their approximate location.  I suppose I could
include *another* interpolation, overlapping the other one, to handle
that case, but this would likely be confusing.

Comments appreciated.  If we can come to a consensus about which of
these is the best solution, that'd be wonderful.  Otherwise, I'll
probably just start using the method used by the 14341-14391 (third)
building, since that already works well with the renderers and likely
works with the geocoders as well, except for the (IMO minor) issue
that it can't tell people when an address is invalid (they can/should
use the USPS website for that, which unfortunately uses proprietary
data so we can't export it, but contains way more accurate information
than we'd ever hope to have - they're the ones delivering the mail, so
they're going to be the first to know when an address becomes
valid/invalid).




More information about the Tagging mailing list