[Talk-GB] OSM UK address project: tags

[James Derrick]

Hi,

Sadly attempting to model the rich complexity of addressing human 
society feels like xkcd 15 standards:
https://xkcd.com/927/

There's lots of ways to do it, all with corner cases, all degrees of wrong!
I wish I had an answer, sadly only comments are forthcoming. :-(

I personally prefer a published national standard to Royal Mail - as 
long as it doesn't cost £280 a copy!

The idea of a documented hierarchy appeals - but what ever we iterate 
to, please can we **document at least 20x examples** covering dense 
city, industrial, flats, sparse rural, and all our UK countries?


The issue with the RM model is they (used to?) structure addresses, post 
towns (defunct?), and postcodes based on their delivery infrastructure 
and not geography. Many rural villages have a crazy post town - 
sometimes in another _country_!

Friends of mine moved close to the English - Scottish border.

An NHS England PCT asked for the postcode and on hearing TD15, they were 
told they don't serve Scotland. Errr....

They regularly get the same confusion from geocoders that don't 
understand that address areas span borders. TD15 is a physical area 
around Tweedmouth spanning borders, city/ hamlet/ shed likely based on 
the driving time of a small red van. The same hit them posting covid 
tests - too far away for Priority Postbox due to a big river.
https://osm.mathmos.net/postboxes/progress/TD/TD15/

If I understand the current RM model, only addr:post_code and 
addr:housenumber are required, making all other fields needed only for 
resilience. (suspect this is incomplete as some areas have BIG inbound 
codes)

I've been adding addr:province=Northumberland for this resilience reason 
as geocoders can get confused by common hamlet names like Newbiggin, 
Blyth, etc. Yes, the lat/lon _can_ derive the boundary area, but do 
_all_ data consumers implement this? How about OSMand+ on an off-line phone?

You can probably find out which border village I'm thinking of simply by 
looking for UPRN tags - it was actually easier to add arbitrary 
references than wrangle structured addr:* fields!


Even if we had a well-understood set of fields, physical history can 
also screw things up. There are many find pre-1900 colliery terraces 
where the name is the **block of houses**, NOT the street as the street 
was originally unsurfaced ground. The naming only makes sense if you add 
addr:street to the building=terrace. Personally, I don't remember using 
addr:terrace as the name is just on a different entity.

You might be able to fight with name:left and name:right tags on an 
adjacent highway=service... but one local terrace purposefully only has 
a garden path! It's never easy is it? :)

Happy Mapping,


James
-- 
James Derrick
     lists at jamesderrick.org, Cramlington, England
     I wouldn't be a volunteer if you paid me...
     https://www.openstreetmap.org/user/James%20Derrick