[Imports-us] Address imports in light of OpenAddresses.io

[Serge Wroclawski]

I'm sure people are wondering "With the advent of OpenAddresses, what
is the state of address imports in the US?" or maybe even "What is the
point of address imports in the US"

While I don't speak on behalf of anyone, since I'm so involved in
imports, I thought I'd speak to the point a bit.

There's an obvious and immediate issue around the availability of
addresses being available. This issue is related to OSM but goes
beyond the scope of our project.

To that end, I see OpenAddresses.io (OA) being a good resource to fill
that immediate need. If you're writing a geocoder, for example, OA is
probably the best resource right now to use as an address backdrop.

That does not negate the power that OSM has, for example, in being
able to provide potentially higher quality data than OA can provide.
For example, a street address may be shared by multiple businesses, or
the lat/lon that a government agency has provided may be less accurate
than what OSM has. In other words, in my view, an address manually
entered by a human being, or validated before entry into OSM has
greater value than an OA address, but an OA address is substantially
(possibly infinitely) better than no address at all.

This leads to a second question, though, which is given OA as a
resource, is it worth while to import address into OSM at all.

I think the answer to that is four fold:

1. If the import is simply going to be a dump of an existing dataset
onto OSM, then OA is a better system for storing that data than OSM
would be, because OA has superior update mechanisms than OSM does

2. If imported data is going to be conflated with other datasets (such
as buildings) and or manually validated by hand before placement into
OSM, such as is being done in NYC, then the OSM data is going to be
superior, especially as we are now starting to see tools that aid in
the ongoing maintenance of imported data.

3. Even in the absence of imports, I would be inclined to use OSM data
that has been manually collected, and the use of OA may in fact
provide a clean separation layer between external data about addresses
and validated OSM data. This means we may see less addresses in OSM,
but have greater confidence in the ones we see.

4. Combining OA with other datasets (such as store/restaurant data) we
can increase the availability and quality of OSM data in other
dimensions. For example, an external dataset (a city heath inspection
list) may list a restaurant and address, but mappers may be able to
provide additional data about that restaurant, such as its wheelchair
accessibility, dietary accommodations, website or hours of operations.

In other words, I think that AO and other projects like it will
provide useful ways for OSM to be more useful, while unburdening it
from many of the problems that we've had around imports.

- Serge