I just ran across a changeset that removed addr:city from a bunch of POIs, on the grounds that it was “redundant with boundary data.” Some of these POIs had a Canadian postcode that would indeed imply city, but others had no postcode. The city could of course be calculated by the POI’s presence inside the city relation, but I hadn’t seen this as a rationale to remove information before.
Would people consider this valid? By extension, should I stop documenting the city when I create a POI; is it considered noise?
addr:city is very important and I advice to always add it to anything which has a address.
When somebody deletes this, I would count it as vandalism and revert the deletions.
My reasoning: The OSM project and community have not implemented, documented or defined such calculations yet. My mind could be changed when it’s in place on osm.org and the logic available under a free license.