[Talk-GB] UK street addressing

James Derrick lists at jamesderrick.org
Mon Dec 28 15:00:10 UTC 2020

Hi Tim, and welcome to OSM and the GB list!

Will you forgive me if I use your post to crystallise some thoughts on a 
very interesting postal_code discussion though the last week which has 
benefited from a wide breath of knowledge from contributors?

On 28/12/2020 09:16, Tim Saunders wrote:
> However, the more relevant point to OSM is that surely there are many different ways of addressing and therefore context is all.
I have to agree that context is important, almost to the point of 'user 
centred design' where we need to both model messy reality, and also 
enable the map to be useful.

What is someone likely to enter into a Nominatim search box, a GPSr 
address finder in a car, or a Python data analysis script?

>    Postal addresses and post codes are one example, but administrative, electoral, lat/lon (with your projection of choice), What Three Words, etc. are all valid depending on why someone is asking the question and who you ask.

You mention in passing an interesting counter-example - WhatThreeWords 
is a proprietary geolocation system based around selling the 'secret 
database'. People can easily turn their location to three words, however 
to work out where that is costs money.

(Proprietary business models aside, there is of course no reason to add 
WhatThree Words as like the free and open https://plus.codes/ , both 
systems are entirely algorithmic - functions can predictably transform 
their system to and from lat/long.)

>    I don't think it realistic that OSM can (or needs to) become the defining source of one true address.

I agree, due to the difference between a 'de jure' and 'de facto' standard.

National legislation sets statutory guardians of key parts of civic 
geolocation - notice I said *guardian* and not *owner*, as power needs 
to be accompanied by a responsibility to make information useful for the 
public good.

Monopoly powers are often used as a means of collecting tolls, to help 
fund the work needed. This can work well, with some data sets looking to 
me like the 'freemium' model where commercial or volume data users pay 
for additional features or access but most people don't need to.

If a 'de jure' guardian charges too high a toll though, it risks being 
imperfectly bypassed with a 'de facto' replacement. Enter Steve Coast 
and OSM... :-)

Back in the day, £7.95 for a 'pink cover' paper OS Landranger 1:50k 
represented great value to a kid with a bike.
£200 for a fixed license for my Psion or £250 Garmin GPSr didn't... 
which is why I started contributing to OSM 12 years ago!

Originally there was no economic means of getting access to the Postcode 
Address File which meant I was forced to used to perform a postal_code 
lookup on an early map site and manually enter the Lat / Long into my 
Psion to get to a friend's wedding.
Thankfully http://www.freethepostcode.org/ happened, and there's now 
enough coverage for OSM to be useful. RM to their credit saw the change 
and started to open up. The 'freemium' business services remain.

Recently, as UPRN / USRN data became available in a limited redacted 
fashion, I added tagging to several 'local haunts' as an experiment to 
see if the post_code process will repeat.
The OSM database can now perform a 'join' or lookup from postal_code to 
ref:GB:uprn using open data (albeit with very limited data - random UIDs 
need 100% coverage).

This is a great example of OSM adding value by including useful external 
data sets which can freely be used for public good. The overhead to OSM 
is not that much - only a single 'foreign key' tag is required per 
database/ namespace.

U*RN coverage is VERY limited due to the guardians redacting the 
information. If you want 100% coverage you need to pay to get access to 
the closed data. That 'freemium' model is starting to build again. :-)

To summarise, IMHO - OSM should only add external 'foreign keys' as tags 
where the external reference is widely used and the additional data 
required is limited.

> This should include a tag showing the Post Town (or being able to deduce the Post Town by reference to boundaries) because, if you ask someone their address, they may give you their postal address (without a postcode) so that is what you would want to search for.

Whilst I can perhaps see that post town might give additional redundant 
data to help correct bad handwriting on a Post Code, or tell which Blyth 
someone is referencing, my conclusion from this thread is that post town 
is already a fossil from a previous postal age! :-)

It's been said that population change and consolidation of RM 
distribution means that post town doesn't necessarily match the route 
from post box to letter box, and has been replaced by postal_code in 
automated systems. My experience is certainly that postal_code is 
universal for human address entry - phone calls, websites, car GPSr, etc 
(often with interactive PAF lookup). Even addr:city is auto-filled.

Back to my previous point about OSM data usage though - post town adds 
data that is not useful as a 'foreign key' and is not useful outside RM; 
postal_code is enough. The real geographic addr:city is useful for 
user-centred tasks as it is discoverable and widely known.

You could almost say RM has been almost too successful in promoting the 
Post Code in the UK. It has been borrowed and is routinely abused for 
geolocation - a part of UK public life.

Thanks for your thought provoking post, for reading this far, and Happy 

James Derrick
     lists at jamesderrick.org, Cramlington, England
     I wouldn't be a volunteer if you paid me...

More information about the Talk-GB mailing list