Question: what is the current understanding of building=* tags (also the whole family, such as building:levels etc) in nodes? Wiki says that we can use in nodes, and there was a discussion back in 2014/2015 for that.
From what I understand, this was useful when imagery was really bad that they were sure there was a house/whatever building there, but could not draw the polygon for that, so building=house was used as a node.
Is this still valid?
More context: The Brazilian community is working in a future import (more info later, in the proper import channel), where we have now addresses available for the whole country. From that data we can obtain many info about the building itself (type of building, levels, flats etc), but the data come as nodes. There is some questions whether we can still use this very useful data along the addresses, as nodes.
Why wouldn’t it be? There are new buildings built all the time and imagery can take a very long time to catch up.
If the license etc on the import is all OK I don’t see why this would be a problem on the face of it. Something like this will need to be conflated with existing data very carefully to make sure that anything OSM has that’s more up to date doesn’t get wiped out or duplicated.
If it’s a small building like a kiosk you may decide to map it as a node even with good enough imagery.
you can map any building as a node, but it will always mean adding less information than with a way, you will not show the size, shape and orientation if you decide for a node. I would encourage mapping every building as a polygon.
The question arose because the original arguments for building=* as nodes were: when there is not enough resolution to map a polygon, a node is required to bring this information.
While this is still true, this is becoming less and less relevant 10 years later. Of course, there are still PLENTY of places without good imagery around the world (especially now we don’t have Maxar anymore).
Our import is expected to bring at least dozens of millions of nodes into OSM. If we bring all of them with building=*, we will heavily impact this number.
Is this a good thing? Bad thing? Doesn’t matter? I don’t know, but since we are organizing the import, this is one of the questions that we were asking ourselves, so that’s why we consulted the community.
Again, thanks a lot for all the inputs! Given nothing changed, we will propose the import with the building=* tags.
Yes, we are aware about conflation between existing addresses and existing buildings.
The building data is valuable (type of building - such as apartment, house etc -, building levels, addr:flats, building:flats). One option to recover it later would be to use the unique ID attributed to every address, but I’m still unsure if local community wants that.
I have not meant that (it is usually a poor idea to add such ids) but rather is it easy/doable/plausible to move that tags to area representing building, once mapped.
Detect whether it’s in, or nearby or if multiple candidate buildings in the external source seem to be applicable to a single building mapped in OSM, or vice versa. It might need a lot of manual intervention.
