[Talk-GB] Non-intuitive addresses

[Mark Goodge]

On 12/02/2022 20:38, Colin Smale wrote:
>> On 02/12/2022 8:29 PM Tom Crocker <tomcrockermail at gmail.com> wrote:
>>
>> For what it's worth, I'm inclined to agree that having only one tag 
>> would be ideal. The reason I'd go with addr:place is various tools 
>> like nominatim won't use the address if addr:street doesn't match a 
>> highway. So we'll tag it but it won't show up in search results, which 
>> seems a bit self-defeating.

This seems to me to be a significant flaw in nominatim. Although 
relatively uncommon, there are still lots of locations that have an 
address where the "street" part of the address isn't an actual street in 
the sense that would be mapped as a highway.

> This sounds like "tagging for the renderer"... And addr:place in 
> particular was rather frowned upon in the discussion in December 2020...

I don't like addr:place in this context, for all sorts of reasons. One 
is that it doesn't match what, I think, would be the normal English 
language use of "place" - we'd typically apply that to an area, such as 
a village or suburb, rather than a building. And the wiki agrees with 
this, giving examples of small rural villages where the streets are 
either not named at all or are not part of the postal address. The wiki 
lists European countries where this is common, but I can think of 
examples in the UK too. So I think that addr:place should only be used 
in those circumstances, and not used as a workaround for issues caused 
by flawed software (as above).

Part of the problem is that the wiki isn't particularly helpful. It says 
that addr:street should normally contain the same name as that of an 
adjacent or nearby highway (which is what nominatim is expecting), but 
doesn't give any helpful indication of what should be used if the 
address doesn't contain a highway name (other than addr:place, which 
isn't appropriate here). I suppose addr:block would work, but I'm not 
sure what usage that has in UK addresses.

Personally, I'd prefer it if addr:street contained the street level part 
of the postal address, even if that's not actually a highway that's 
mapped in OSM, and software that deals with addresses (such as 
nominatim) should follow the robustness principle when deciding what 
tags to accept.

There's a separate problem in this particular case, which is that Abbots 
Walk does exist as a street in the NSG (it's assigned to the pedestrian 
access path behind the terrace). But we can't map that in OSM, because 
there's no open data source for it (it's not in OS OpenNames) and no 
indication of it on the ground. And that's another reason that I think 
nominatim and other software should be more tolerant of apparent 
discrepancies in the data, because sometimes, those discrepancies may 
simply be because we don't have an open data or on-the-ground source 
that can be used to fill in the gaps.

Mark