Re: [Pan-users] Re: Big XML files... (was Re: Re: Better processing of very large groups?)

[Ron Johnson]

On 2009-07-04 20:58, Steven D'Aprano wrote:
[snip]
> The only downsides to SpamAssassin are that you need to be running your 
> own mail server,

Setting one up takes a little research, but Debian, at least, does a 
lot of the configuring for you, making you select 1 of 5(?) 
different kinds of service (everything from ISP to local-only), then 
prompting you for the relevant info.

All I needed to to were some minor tweaks to integrate SpamAssassin.

>                  it is relatively weak at "picture spams", and by 
> default is has the unfortunately habit of learning from it's own 
> mistakes. Not in the good way of "well don't do that again", but in the 
> bad way of "I think that's spam, so I'll flag anything that looks like 
> that as spam" -- even when it isn't.
>
> Given that last factor, SpamAssassin requires some administration, so 
> it's not just fire-and-forget, but I'd recommend it.

There's a teaching app, sa-learn, which you feed it an mbox of ham 
or spam.  Do that a few times over the space of a month or two, and 
you're good to go, with very few false positives.

False negatives will always be a problem, though, because there are 
an infinite variety of ways to write spam, but a finite number of 
heuristics.

--
Scooty Puff, Sr
The Doom-Bringer