They are - at least to a subset of people including me. You can directly derive the postcode and linked USRN from them (and then usually other address details from that, which I find handy. They're also good for finding unmapped things (new houses etc).

If there is a seperate path that doesn't follow a road there will normally be a USRN record for it, and a UPRN at some point (typically where it joins a road, or other path). For pavements there usually isn't anything distinct.

Hello,

I did add it intentionally. It is the street record UPRN for USRN 34401599 (Handsworth Avenue). It basically represents a notional point for the street.

Also, thanks for adding/splitting and adding addresses to all those buildings in Sheffield - it's now much better mapped.

Hello, do you have an example to hand?

I generally split buildings where there are separate footprints and addresses (for example a semidetached house), but add to building parts for parts of a complex building (such as a hospital, with wings that are connected together).

Thanks for spotting this - I can see what went wrong. It should now be fixed in changeset/189265798

Thanks for surveying those so quickly from the note. I've added UPRNs and (as a result) postcodes to the buildings you've added.

Thanks - I'll do a few passes over to check for anomalies.

That one should be fixed in changeset/188165281

Good idea - I'll add notes to any that are similar in future

My logic on this one was that it is the only UPRN linked to that USRN that is not unambiguously linked to a building, and falls on the centre line of a road with nothing else visible in the vicinity that is a likely candidate.

In my experience most areas add these street records at junctions, but some areas (like this one) do add them somewhere along the road.

Will do. I'll try and make my chaneset descriptions more descriptive and split them out as you suggest.

I also find it strange that having created USRNs, UPRNs are assigned to represent the same object, though UPRNs seem to almost have become a general purpose way of specifying coordinates in some cases, with things like ponds often having them.

Hello,

I believe that UPRN is the street record. My basis for this was that it links to the same USRN for that street, is centred to the road, there isn't another candidate to be a street record, and there wasn't anything nearby (street furniture, postboxes, etc) that could be another candidate.

I've found that the positing of street records varies by area (though is consistent within them). They always seem to be aligned at the centre point of the street, but can either be positioned exactly at junctions, offset from a junction to the be on just the street in question, or somewhere central on the street.

I do agree that it only makes sense to add UPRNs to that confidently match things in OSM and that the list of "unmatched" is very valuable in itself. My thinking on adding street records, in the absence of any open data on the UPRN type, is that this reduces the list of unknowns to ones that genuinely need investigating.

You're right. Thanks for spotting that and fixing them. I really must write a JOSM validator to spot these.

Thanks for pointing this out - I'll review them over the next few weeks.

Good spot - that was a copy & paste error and should have been in addr:postcode. I've now fixed it.

Hello,

I generally use cadastral parcels for "boundaries" between parts of buildings, and UPRNs as (typically) centres of buildings. I would generally only re-align buildings based on the cadastral parcels - so apologies if my comment wasn't clear about this.

I do try and blend the alignment between cadastral-parcels and existing where I can, to avoid any discontinuities along the way. As you know, there is a balance to be stuck between accurate on an absolute vs relative scale, and I try to balance both - but not always perfectly.

I always prefer to fix anything I've broken myself, if possible, so I'll look over this in the next few weeks. On that note, I appreciate you bringing it to my attention. (I feel that I should point out that I genuinely mean this, not based on your messages, but on general correspondence).

It sounds like our our approaches are very much aligned, but might have queries about edge cases,

I had previously tried to group related changesets into a larger changeset, but based on feedback, these might be better broken down to smaller geographic areas. I try and do this for any new changesets.

As always, I'm happy to receive advice and recommendations on any changes.

Phil

Hi,

I generally try and align with Cadastral Parcels, and UPNRs (as a centre point to buildings).

I tend to combine changesets with UPRNs and other addressing date with geometry, as I find they are so interlocked that working with both together makes things easier.

I will have a look over the area to make sure the geometry changes make sense. I might not be able to do this immediately, but as you've asked, I'll put it at the top of my list.

Phil

I had intended to split this the initial changeset, but had left the "auto close" setting in JOSM off in error, so they all ended up in one large one.

I'd be more than happy to split them out, but my understanding is that it is not just the size of individual changsets, but their aggregate volume that concerns local mappers, so I wouldn't want to jump in and do that if it wasn't welcome.

(I'd also need to work out how to do that, but that's a lesser problem)

Maybe a building:part or distinct entrance node might be appropriate, but as you say, it needs more information or ideally a survey. I just wanted to let you know.

Thanks - that is a very good spot!

Hello,

I wonder if this might be part of the building carved out to be independent? A few sources online suggest that may be the case, but nothing usable for OSM.

Anyway, I'll defer to your local knowledge, but just wanted to point this out.

Thanks, I think we were both changing it at the same time, which might be behind the confusion here.

I generally don't add addresses to this sort of thing, but I've come across quite a lot of areas that do. I do think the postcode should be in postal_code rather than addr:postcode (if at all), as wouldn't really have a postcode as such.