That's not necessarily the right way to do it. To formalize the problems, assume that Wikipedia was intended to be subjective and that editors are potentially

That's not necessarily the right way to do it. To formalize the problems, assume that Wikipedia was intended to be subjective and that editors are potentially malicious. So some edits in the middle of a changeset might be malicious, but children edits are legitimate. So you want to delete the malicious edits but preserve the dangling child edit. What is the correct way to reconcile this changeset? Once you answer the above question, you can soften the intepretation of malicious to: Users that I would rather filter, because their opinions are stupid to me or because I consider them spammers.