Diff: APlanForWikiSpam

--- Version 

+++ Version 1 

@@ -0,0 +1,16 @@ 

+### Index ### 

+---- 

+! Overview 

+Its rather simple.  Rewrite each edit as an email, and use existing spam tools to classify the edit.  Bayesian filters should work fantastically well for this application, though I can't think of any good reasons why more traditional filters such as (http://www.spamassassin.org/ SpamAssassin) won't work. 

+---- 

+! Wiki as Email 

+!! Headers 

+Most of the headers in this application are defunct, but well formed headers will help the filters work their magic in the correct fashion. 

+!! Content 

+The content is a little tricky.  Do we simply supply the raw wiki text, or do we render into HTML?  Which content do we include - everything or just the diff?  Initially I think that the diff text in raw form should be enough, rendering into HTML is probably a good idea at a later date. 

+---- 

+! Filtering 

+!! Bayesian Filtering 

+The regular benefits of Bayesian filtering over other methods should apply equally as well on a wiki as in email.  As with any Bayesian filtering, the system needs to be trained and so the training interface will probably be the most cumbersome component of our anti-wiki-spam coding. 

+!! !SpamAssassin 

+!SpamAssassin's default rules would need to be tweaked by use of a custom config file, as various tests (eg: MIME_HTML_ONLY) are useless in this context.