[Tagging] Street and Sub-Street in Address Tagging

[Colin Smale]

On 2021-01-16 17:27, Paul Allen wrote:

> On Sat, 16 Jan 2021 at 15:51, Colin Smale <colin.smale at xs4all.nl> wrote:
> 
>> This is why RM say that if you only have room for one street name, use the Dependent Thoroughfare ("Bar Court") as this will be unique within a postcode.
> 
> Theoretically the house name/number is unique within a postcode 
> and nothing but the house name/number and postcode are needed. 
> A little redundancy helps with error detection and correction, though. 
> 
>> There are two sorts of house names: Firstly (particularly in rural settings) where a house does not actually have a number. In these cases the name is managed by the local authority together with Royal Mail and can't be changed at a whim.
> 
> According to Royal Mail, it is solely the purview of the local authority 
> and any changes must be approved by the local authority.

The LA are the ultimate authority for house names/numbers and street
names, but they will consult RM in the process. 

>> Secondly there are "vanity names" that people add to a house that has a number. In that case the number must still always be displayed on the property and used as part of the address, and the house name is "optional".
> 
> In theory there is no difference between theory and practice.  In 
> practise, there is.  I've surveyed houses that display neither name 
> nor number.  I've surveyed houses that display only a name but 
> are between two numbered houses in a range of numbered houses 
> and the number can be inferred (and confirmed by other means). 
> And I've encountered houses that may once have had a number in 
> the distant past but it is not displayed nor can I find anything to 
> say what it might have been, nor can it be inferred (when only 
> only one house of three on a dependent thoroughfare has the number 
> 2, it's impossible to tell which 1 and 3 were).

Not sure I understand what point you are trying to make here. Is it that
"not everybody follows the rules"? 

>> Because "Foo Towers" is a Dependant Thoroughfare and not part of the house name/number perhaps? Or maybe it is a "Building Name"?
> 
> Dependent thoroughfare if it has multiple occupants, building 
> name if it has a single occupant.  Probably.

Maybe somebody with access could look up some examples in the PAF to see
exactly how they are modelled? 

>> We have partial kluges for all three categories, but none work 
>> well.  A field for dependent thoroughfare would be a full solution 
>> for all of them (but there are probably weirder examples that not 
>> even that would fix).
> 
>> Indeed, the suggestion of addr:street mapping to the Dependent Thoroughfare and addr:parentstreet mapping to the Thoroughfare fixes this, and because it is a direct mapping to the address model used by RM in the PAF, it is likely to accommodate the "thoroughfare" part of all official addresses in the UK.
> 
> I think I would have preferred addr:sub-street for the dependent 
> thoroughfare.  I think it a better fit to how people (I'm using a sample 
> size of 1 for "people") think about things.  But maybe a slightly worse 
> fit for tower blocks.  But I'm losing the will to live, so parentstreet is 
> looking good.

I agree, the semantics of street/substreet would seem to be a better
match to thoroughfare/dependent thoroughfare. However the relationship
we are accommodating here is a simple two-level hierarchy, and as long
as it makes clear which term is at which level, it seems good enough.
Apparently addr:parentstreet is already in use, that's why I mentioned
it. 

>> A slightly nerdy explanation of all these data elements can be found here: https://ideal-postcodes.co.uk/documentation/paf-data#thoroughfare
> 
> It started well enough, though I noted it didn't explain why a 
> thoroughfare might be split into more than one postcode (it seems 
> to be because the postcode corresponds to the route a postperson 
> uses).  But as I read further it completely sapped my will to live.

That's a shame. But why is it so important for us mappers to understand
the rationale behind the system? If we start with the accepted premise
that the addresses we are trying to represent with the addr:* keys are
postal addresses (as opposed to addresses for any other purpose) then
our task is to understand how the system works, and not why it was
designed in a certain way. 

Unless of course OSM is intending to come up with a "better mousetrap"
and avoid errors that Royal Mail have made in their system. 

The major use case here is for an OSM mapper: "the postal address is
XYZ, how do I represent that in OSM". 
There are other geocoding-type use cases like "what is the postal
address of this building" or where is the postal address XYZ located." 
Any others you can think of?
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.openstreetmap.org/pipermail/tagging/attachments/20210116/3a385999/attachment-0001.htm>