[Gllug] Spamassassin problems

[Nix]

On Thu, 30 Jun 2005, Ian Lewis murmured woefully:
> # Clear the existing headers and make way for the new ones below...
> clear_headers

This is not generally a very good idea.

> # Adds string to Subject:  Header tag

Well, that's what it *should* do. In 3.1, that's what it does.

In 3.0, it won't add a subject header that isn't there. (Whoops.)

> # Whitelisting on auto (off as of 9th June 2004). 
> use_auto_whitelist 1 

YOu said it was off, and then turned it on.

> auto_whitelist_path /root/.spamassassin/auto-whitelist
> auto_whitelist_file_mode 0700
> auto_whitelist_factor 0.55 

This is (almost) the default...

> # Where to place the Spam Report. 0 is default and thus places the report in
> the email body.
> #report_header 0 
> 
> # Do the headers get 'folded' 0 means NO
> fold_headers 0
> 
> # Report Length. Default is to use the Long form thus 0 is default here
> #use_terse_report 0

You're wiring the long report, unfolded, into the headers?!

> add_header spam Flag _YESNOCAPS_
> add_header all Level _STARS(*)_
> add_header all Status _YESNO_, hits=_HITS_ required=_REQD_ (bayes=_BAYES_)
> autolearn=_AUTOLEARN_ tests=_TESTS_ version=_VERSION_

If you just didnd't clear_headers, you wouldn't need these lines at all.
(Note in particular that a number of outside-SA things are sensitive to
the exact form of the X-Spam-Status header, and you've changed it.)

> # You may set the bayes autolearning threshold (def 0.1)
> bayes_autolearn_threshold_nonspam 2.1 
> autolearn_threshold_nonspam 2.1 

The first of these is called bayes_auto_learn_threshold_nonspam. I can't
tell what config option autolearn_threshold_nonspam is meant to be at all
(it doesn't correspond to an SA option).

> # You can set the bayes autolearn spam threshold (def 12.0)
> bayes_autolearn_threshold_spam 5.6 
> autolearn_threshold_spam 5.65 

Ditto. I think these rules need a linting.

I'd also call these dangerously aggressive values: you're flirting with
a poisoned Bayes database, I'd say.

> # Here you can set the quantity of ham/spam learnt before bayes activates. 
> # default figure is 200
> bayes_min_ham_num 200
> bayes_min_spam_num 200

... so why reset it?

> # When reporting (spamassassin -r) bayes can automatically acquire the result
> bayes_learn_during_report 1

This is also the default.

-- 
`I lost interest in "blade servers" when I found they didn't throw knives
 at people who weren't supposed to be in your machine room.'
    --- Anthony de Boer
-- 
Gllug mailing list  -  Gllug at gllug.org.uk
http://lists.gllug.org.uk/mailman/listinfo/gllug