It is a very common occurrence that iD users (and potentially other editors) accidentally move a node halfway across a city when trying to drag the map. I occasionally encounter it when running my bus stop update script and getting a warning that a stop is too far away from its expected location. Most recent example: Node: Pharmacy (5210718203) | OpenStreetMap at Changeset: 174040515 | OpenStreetMap. Of course, I would only catch a very small subset of these mistakes this way.
I would hope this is a known problem. Is there any past or current work to try to catch and mitigate this?
Edit: the accidental move, visualized by better-osm-org:
That’s a good point about roads, but node objects would never be detected this way. Benches, bins, address nodes, trees etc. are all equally susceptible.
Of course it’s not easy to tell whether a move is legit or not, but there can be some good heuristics.
Perhaps osmcha or similar QA tools should detect “suspicious” node moves for human review? Do they currently do this?
AFAIK not, and problem here is that detection rate of actually bad edits would be far from 100%
and we have many QA detectors of certainly bad data that wait for processing, so I am not sure whether spending effort on making one more QA and operating it is worthwhile
though if someone is interested and otherwise they would be doing something less useful or this is more interesting to them etc… Feel free to do so!
This was discussed elsewhere and yes, there is an open issue with iD:
For what it’s worth, I prefer the solution proposed in that iD issue report (reduce the possibility of accidental moves by requiring selection before move) rather than some kind of post-hoc validation.
Until that is implemented, here’s where I always take a closer look:
large bbox
new mapper
Editor “iD”
When there’s not HOT-Hashtag in the comment I often find dragged nodes in those.
But this problem isn’t restricted to new accounts. Happens to seasoned iD mappers as well. But since I won’t go over to git hub and make an account I can’t upvote the issue.