On 7/27/07, Tim Starling tstarling@wikimedia.org wrote: [snip]
In this case a small amount of care by the deleting admin would have avoided the problem. It should have been obvious that the file was a useful inclusion to the article on the named artist.
Though the deletion itself is easily undone once the error is discovered... Almost as easily as the section removal. Asking human resources to spend perhaps 10x the amount of time on an unused image just doesn't scale well, that process is already way over taxed.
The problem could also be addressed by better handling of vandalism. Not that thats an easier problem, but solving it also solves a lot of other types of wiki-rot that more care with deletion doesn't solve.
In this case the vandalism never should have stuck: The diff was obviously bad. Any experienced human looking at it would have reverted. I'd argue that the total removal of all categories is so infrequently valid that we could painlessly have a software agent autorevert, or at least bring the edit to human attention.
We're still seeing a constant stream of bad edits which could trivially be identified as bad by software sit around until a human editor finds them on their own. (http://en.wikipedia.org/w/index.php?title=Cannabis_%28drug%29&diff=prev&...)
Can someone please tell me why this is still happening? We've had the software to identify bad edits for years now, why isn't it working?
The IRC bad-edit feed that I used run went down due to technical reasons... I didn't bother putting it back up because it wasn't fun to use since people using it were slower on the draw than folks directly skimming the RC feed. ... but the fully automatic reverting bots shouldn't have a problem with non-fun work. ;)