mmd
74
For minutely diffs, you need to keep in mind, that there’s not only an OSM object lifecycle, but obviously a Wikidata object lifecycle as well. Every time, one of the two sources changes, you need to create a new object “version” and distribute that via minutely diffs. Just sticking to OSM diffs and enriching them will miss all those Wikidata changes that happen in the meantime.
Now, once you start creating artificial object versions as a superset of OSM + Wikidata changes, there’s absolutely no sane way to ever feed this information back into OSM (like in the mirror db scenario I mentioned). Version checks are strict, and skipping versions is not an option.
Also, by having separate tags (one for the OSM data, and for the Wiki data), you could of course make the origin clear. However, data consumers need to be adapted to do something meaningful with those two tags, choose one of them based on some clever criteria.
This is all very tricky and really only meaningful if you want to provide something to pure end consumers who will never contribute anything back to OSM based on that augmented/enriched data.