I agree with that. But, maybe we shouldn't abandon Jonathan's idea, but use it for soft security. Similar to a prior suggestion of mine:
Those who want can activate a user option that marks "suspicious" edits. That would include
- edits by IPs (anons) that remove more than 20% of the article (in bytes)
- edits (by IPs, or anyone) that add certain keywords (f**k, etc.)
- edits by IPs whose edits have been "rolled back" (with the function)
lately
- edits by IPs who are listed on the Vandalism in Progress page
These are just some ideas that come to mind. More can be added. This won't find all vandals, and will have a false alert once in a while, but could improve the "hit rate" on malicious edits.
How about giving sysops the ability to flag certain IPs as "suspicious"? I wouldn't want to extend that ability to signed in users for various reasons, though.
Personally, I think that having talk pages and a "new message" indicator for anons would go a long way to resolving many problems. I've put that in my pipeline.
Regards,
Erik