[Talk-GB] Non-intuitive addresses

[Mark Goodge]

On 12/02/2022 14:40, Colin Smale wrote:

> Don't forget there is reasonably broad agreement that OSM uses *postal 
> addresses* in cases where they conflict with addresses to locate or 
> identify a premises. When addressing a letter to Abbots Walk, Royal Mail 
> say you can omit the main street name if you are pushed for space.

In this particular case, Abbots Walk is not a substreet or, in Royal 
Mail terms, a dependent thoroughfare. It *is* the street name (or, in RM 
terms, thoroughfare) in the address, despite the fact that the houses 
actually face onto a street with a different name.

This illustrates one of the problems with postal addresses. There's no 
simple, reliable way to infer them from what's visible on the ground. 
And, apart from the post town, none of the canonical sources are open 
data. Postcodes are open data in themselves, but the linking of 
postcodes to individual properties isn't. UPRNs and theoir Point 
coordinates are open data, but linking UPRNs to postal addresses isn't. 
USRNs and their LineString coordinates are open data, but linking street 
names to USRNs isn't. It doesn't help that OS OpenRoads and OS OpenNames 
(which are open data) don't always match the names held in the NSG or 
the PAF (which aren't open, but are canonical), and the LineStrings of 
OpenRoads don't always match those of the open USRN dataset.

It's all a horrible, nasty mess, and will continue to be a horrible, 
nasty mess until we finally get the NSG and PAF released as open data 
(which will happen, eventually, but not before a lot of institutional 
inertia has been overcome). In the meantime, we just have to do the best 
we can with what we've got available.

Mark