> Mike Richardson wrote:
> > I personally don't like the idea of content filtering on the hubs because
> > a) users have no control over what test are done and for what.
>
> Not true. One doesn't have to make it a blanket reject.
>
> > b) users have no control over the threshold we pick for them
>
> Not true. Again, one doesn't need to make it a blanket reject.
>
> > c) we can't teach the system what individuals class as spam
>
> Not true. There are methods to collect such data from the users.
>
> > d) false positives happen
>
> While this is true the highest I've seen is 5.6. I reject at 8. If
> you're worried, set the reject threshold higher and pass it on tagged. I
> doubt you'd ever see a false positive at 10 or so (esp. with Bayesian turned
> on) and even if you did it would be so extremely rare that it is worth the
> loss (thousandths of a percent if not millionths of a percent).
>
> > e) there is a move to stop tagging and start rejecting 'for the good of
> > everyone'.
>
> Which is far preferred to the movement towards C-R and other systems
> which severely break how mail is supposed to work.
>
> > Mix all these together and you have censorship with no ability to
> > opt out.
>
> Not true. Whitelists exist. Give the users the ability to add their
> name to it.
>
> > I know that some of the above could be fixed but we really don't have
> > the time or effort to do the work necessary on the exim config etc.
>
> It has little to do with exim configurations and a whole lot to do with
> Spamassassin configuration.
>
> > Desktop tools return the control to the users.
>
> Most desktop users don't know what to do with that control. It is far
> easier to reject the blatent cases, tag the marginal and tell them to filter
> on a single header than it is to try to educate the masses on proper email
> management. If given the choice of setting up Spamassassin (or spamprobe, or
> spambayes) versus a C-R style system most will set up the C-R system because
> it is far easier for them to set-up and maintain.
>
> Sensible defaults with the ability for the end user to opt-out is not
> censorship; it is being responsible and realistic about the capabilities of
> the average user.
We have 33,000 users each with a minimum of three possible email addresses,
most with at least 10 not including generic addresses and shared aliases.
This makes per user configuration (rules, whitelists, bayesian) in
Spamassassin on the hubs very difficult requiring database lookups. We use
exiscan which handles the threshold setting thus again requiring some sort
of database lookup.
Thus why i said it was possible but too much work to actually implement and
would require big changes to the exim conf. We'd have to change our entire
mailing routing policy just to make it easy enough for people to use. We
have 2 mailhub support people, several dozen desktop support people and lots
of trainers. The desktop approach makes more sense.
Mike
--
-----Plain text only please - attachments stripped on arrival.------
Copyright 2003 Mike Richardson, Room G98, Manchester Computing
University of Manchester, M13 9PL doctor@??? Int: 56009
Left through main doors. Right then left at end of corridor.
First door on left. URL http://kira.mcc.ac.uk/ Ext: 0161 275 6009
--------------------------------------------------------------------
"If I want your opinion, I'll **** it out of you!" - Chuck Norris
"If anything happens to my daughter I have a ** and ******" Clueless
