I've written two new spam cleanup scripts, an aggressive one and a targeted one. The
aggressive one,
which edits as "Spam cleanup script", checks every article on the wiki against
[[m:Spam blacklist]],
and reverts any matching articles back to the last revision which didn't match, or
blanks the
article if they all match. It drew complaints when I ran it on some wikis, for blanking
articles
which contained links to free domain services since their creation, and rightly so. It
takes many
hours to run so it's difficult to monitor closely. I don't think I'll run it
again on Wikimedia,
except perhaps on small, heavily spammed wikis.
The second script is a targeted script which takes advantage of the new externallinks
table to
quickly find links to a given domain. It then removes the spam in the same manner as the
first
script. It edits as "MediaWiki spam cleanup". In my initial test, it took 12
minutes to remove all
links to *.f859.info on all wikis. It should be possible to make it a fair bit faster than
that.
I'll leave Brion to tell you about his CAPTCHA extension.
-- Tim Starling