A comment on said poll page -
I think it's easier if we develop a Bayesian Flitering Edit system. The RC patrol can mark edits for a Bayesian filter to show what kind of changes are considered as "spamish" or "trollish"... in our case, "vandalism." Then in the RC changes, have a percentage listing how much of a chance that the article was vandalised by the edit. Factors would include large amounts of missing text, large amounts of added links, and so on, and so forth. --[[User:AllyUnion| AllyUnion]] [[User talk:AllyUnion|(talk)]] 09:13, 11 Dec 2004 (UTC)
So the questions are, can it be done, and how dificult would it be to do?