I am getting into this discussion a bit late (and didn't figure out how to reply to the last message in the thread "[talk-ch] Proposed mechanical edit..."), but maybe some info from our experiences with fountains in Zurich can help:

- P2795 has already been used in many instances for describing the location of fountains. See for example https://www.wikidata.org/wiki/Q27229889

- we are working on updating the fountains in wikidata with information from Open Data Zurich. A first import of the data was performed using the following script: https://github.com/mmmatthew/import2wikidata This script uses coordinates to match fountains between the two datasets to avoid creating duplicates. See https://github.com/mmmatthew/import2wikidata/blob/master/zurich_fountains_preprocess.ipynb, block 13. A similar approach can be used with OSM data

- We plan to write a bot that will keep wikidata up to date with Zurich's data, but there are some open questions

- our end application is a map that displays data conflated from OSM and Wikidata: https://beta.water-fountains.org and https://github.com/mmmatthew/proximap. Therefore, we are also working on a backend that matches fountains between the two sources and combines information between the two. This is still work in progress, as you can see in the issues: https://github.com/mmmatthew/datablue/issues

- a question that arises for our specific case is: when osm and wikidata provide conflicting information, which information do we display? For this, we introduce trust ranking on a property basis. See for example https://github.com/mmmatthew/datablue/blob/4bb1dc185aac5738375d12bf84c97881854ded89/config/fountains.sources.osm.js#L93 indicates that the 'name' information coming from osm is not given highest trust ranking.

Hope this information is of some interest to you.

Best,

Matthew