[Talk-us] Feature proposal: proposed expanded address tagging scheme for US

[Serge Wroclawski]

On Tue, Nov 20, 2012 at 10:05 PM, Carl Anderson
<carl.anderson at vadose.org> wrote:

> A "Colloquial core phrase" is something we all use everyday.  We shorten
> names down to a useful, but still meaningful, core.
> If I were to say that I was at 14th and K, many of my DC friends would know
> that I was at the intersection of 14 St NW and K St NW.
> My friends who are not familiar with DC could guess the location given a bit
> of prompting.

> Steven's proposal creates a mechanism for local knowledge and local
> colloquial use to be added into OSM.  In turn this data, when present, will
> allow people who interact with the public to better understand the intent of
> the public in a more precise fashion.
>
> The parsing steps move the bits that are not part of the core into well
> known tags that can be unambiguously dealt with.
> The unambiguous aspect is equally important as abbreviation usage is often
> lossy.  For instance some US jurisdictions use BL as an abbreviation for
> Boulevard and others use BL for Bluff.  (In the emergency services world
> hilarity does not ensue).  If OSM had such names as "Braided Blanket Bluff"
> in the proposed tagging scheme

Carl,

First, in your example of "14th and K", there is no such address as
14th and K, but rather this is a geocoded location. This is important,
but not related to the proposal. In other words, the proposal doesn't
change this fact (nor does it address the fact that OSM doesn't have a
geocoder that can handle intersections).

Secondly, regarding street names, we have existing extensive support
for multiple names, including local names, international names,
previous names, etc:

http://wiki.openstreetmap.org/wiki/Names

In fact this proposal would complicate the addresses by taking each
name's street and breaking it up into many more tags.

> If we were to use the proposal as additional tags to the current existing
> tags people could add to OSM data to the limit of their local knowledge and
> when they knew the common local usage could, correctly, completely and
> unambiguously fill out the parsed tags that Steven has proposed.

After weeks working with the OSM data in the US, looking specifically
at the interaction of TIGER tags with the "name" tag, I can say
definitively that this is a bad idea.

On the one hand, Steven's proposal is saying that the "name" tag is
derived data- that it is the result of a summation of different
components (the direction prefix, the name base, the road type, the
direction suffix). And on the other, you're suggesting that the
existing use of "name" could be preserved.

We have an existing test case for this using the TIGER dataset in the
US, and the fact is that people do not keep the two synchronized. We
have tiger  tags which have not been touched, but the name is updated,
and we have tiger tags which have been changed, but are changed
incorrectly.

- Serge