[Imports] Queens, NYC addresses (was: United States Poultry Import)

[Kevin Kenny]

On Fri, Apr 22, 2022 at 10:43 PM Jmapb <jmapb at gmx.com> wrote:

> On 4/20/2022 6:52 PM, Kevin Kenny wrote:
>
> The NYS E911 address point data has a 'zipcity' column that's supposed to
> contain the 'postal city'.  I recently redid about 130,000 buildings in NY
> State to use it, because `addr:city` was giving unacceptable values
> relative to the ZIP code - usually because it was an enclosing village
> rather than the serving post office. There were a bunch of other address
> defects, like dropping the street name prefix and suffix, giving 'Church'
> in place of 'West Church Street', or whatever.
>
> The only ones that I hit were the ones that `miluethi` and his horde of
> sock puppets imported.  If there are systemic problems with NYC address
> points, we could consider designing another mechanical edit to address them.
>
>
> Well then maybe this discussion *does* belong on the Imports list ;)
>
> We're having a NYC meetup tomorrow and I'm hoping to wrangle some
> opinions. If you happen to wake up in Brooklyn tomorrow, please drop by (
> https://osmcal.org/event/1286/) , but I'd appreciate your input
> regardless.
>
> Personally I'm skittish about broad mechanical edits, especially over data
> that's heavily in use. What will the addition of addr:city tags gain us? Is
> the current situation enough of a problem to warrant mass-tagging of E911's
> zipcity into addr:city? In Queens only, or citywide? What are the possible
> downsides? And how do we evaluate zipcity's accuracy, when there's nothing
> to check it against? (Spot checking against USPS is possible, of course, if
> that's deemed useful.)
>
> In my experience, the issue of Queens' hyphenated housenumbers is a much
> bigger problem. How do we let mappers and data consumers know that 30-34
> 42nd St is the same as 3034 42 St in Queens, but anywhere else in NYC it
> means a range from housenumber 30 to housenumber 34? What's the best way to
> tag this, and how do we help mappers get it right? (This is a real problem
> right now -- compare a search for "3034 42nd St Queens NY" on Nominatim, or
> any OSM-based mobile app, versus our corporate analogs.)
>
> As I said, I hope to bring up these issues tomorrow, but I don't advocate
> any swift action. Whatever our course, I'm hoping it's a cautious and
> deliberate one.
>

Sorry, I was hiking in the Helderbergs yesterday.  With the CoViD-19
numbers having yet another surge, I'm also reluctant to travel - hiking
hardly counts since I think I saw three other people all day, all outdoors
at a distance.

I surely wasn't planning to do anything half-cocked.  `miluethi` hadn't
edited Queens addresses, and I didn't touch them.  The only thing I changed
was buildings that he imported, that hadn't been changed since the import.
I figured that it couldn't make anything worse! Tens of thousands of
buildings in that import had addresses damaged in other ways. It just
occurred to me that merging in information from E911 addresses might be at
least a partial solution to the Queens postal address dilemma.

I think it's useful for a geocoder to be able to encode the correct mailing
address of a building.  '159-13 97th Street, New York, NY' is surely wrong.
Only Manhattan addresses use the 'New York' name.  '159-13 97th Street,
Queens, NY' is slightly less wrong and would surely be delivered if the ZIP
Code is right, but it still doesn't match the postal preference.  '159-13
97th Street, Howard Beach, NY' would surely be preferred, since that's how
you'd address a letter.  I have a strong preference that the `addr:*` tags
on a building have enough information to compose its street address, or
else that the other information can be obtained algorithmically in a simple
manner.

Given that postal addresses in the US follow carrier routes and not
political boundaries, I resist the idea that "you can find the city by
looking at the smallest containing administrative region." It doesn't work.

Even more important is that if a map user searches for '159-13 97th Street,
Howard Beach, NY' - a well-formed, and indeed a correct address, a geocoder
has to be able to find it! Looking for nearby `place` nodes is likely to be
pretty dodgy.

--
73 de ke9tv/2, Kevin
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.openstreetmap.org/pipermail/imports/attachments/20220424/fa8bafbf/attachment.htm>