[HCS-D]Re: new student group

Gregory N. Price gprice at fas.harvard.edu
Wed Dec 8 20:42:48 EST 2004


  On Wed, Dec 08, 2004 at 04:05:54PM -0500, Ivan Krstic wrote:
    Gregory N. Price wrote:
    >I'm not sure what `tagging-based' is supposed to mean, but the
    >approach used by SpamAssassin certainly isn't false-positive-proof.
    
    Right. I was referring to the fact that SA/CRM114 'tag' mail headers 
    with a spam index, and then let user-side software deal with it 
    (procmail) rather than rejecting mail outright - the RBL approach - at 
    the MTA level.

Oh, I get it; when you wrote

  On Wed, Dec 08, 2004 at 12:59:32AM -0500, Ivan Krstic wrote:
    Ivan Krstic wrote:
    >hcs% mailq|grep "^[0-9A-Z]"|wc -l
    >    2923
    
    ... of which, I just found about 1000 fall in one of three roughly 
    identically-headered pieces of virus/spam mail. When migrating lists, 
    I'd like to hook automatic virus scanning into the MTA ...

you didn't mean the latter sentence as a solution to the former
(i.e. the size of our queues), but rather to the downstream problem
(i.e. the spam in users' mailboxes) it points out.

It's true that I don't really need any customization of my spam filter
beyond perhaps what numerical threshold I use.  So system-wide
tagging, if it makes learning more effective or even just saves
resources, is a good thing.

Other thoughts?

                Gregory





More information about the hcs-discuss mailing list