From: Rob Chanter (no email)
Date: Tue Feb 01 2005 - 05:43:35 EST
On Mon, Jan 31, 2005 at 02:23:41PM -0500, Victor Duchovni wrote:
> On Mon, Jan 31, 2005 at 02:14:31PM -0500, Jason Gauthier wrote:
>
> Why centralize the job of reviewing spam for FPs? We send users a 6-hour
> summary of new mail in the quarantine and they can release anything
> that looks plausibly relevant to them, otherwise it just ages out
> of the quarantine after 31 days. We find that users release only 1%
> of quarantined mail, so most quarantined messages go unreviewed and
> (to mangle a verb) "unmissed".
PMFJI. I'm in the process of implementing one of these. It consolidates
amavisd-new quarantines from a bunch of MXes via rsync, then runs a
digest script over it to send fresh message summaries. Rescue is via
return email requesting release of the (for practical purposes
unguessable) quarantine filename. At the moment, it considers that a FP
should be rescued for all recipients, and so we tag it with [rescued] in
the subject and an X-Rescued-By header. It strips any sensitive stuff
like Amavisd's X-Envelope-(To|From) headers before release. Rescuability
is based on whether the requestor is an original recipient or quarantine
administrator.
Rescue or lack thereof also prompts feeding of a site-wide Bayes store.
From that brief description, is there anything insane in my setup that
I've overlooked?
Viktor, did you arrive at 6 hours by any vaguely scientific process? Is
that mid-morning and mid-afternoon? I was thinking that running
overnight and midday would kind of make sense, but don't really have a
feel for how people are going to use it.
ta
rob
|
|
|