> Kai Schaetzl <[EMAIL PROTECTED]> wrote: >> Chris Barnes wrote on Fri, 6 Jun 2003 12:03:20 -0500: >> >>> I am more concerned about the false positives. Looking at the >>> headers of those messages that are legit, it seems that the biggest >>> score comes from the BAYES test. >>> >>> Is there a place that talks about how that score is derived? >> >> I'm not sure what you mean, did you already check out >> spamassassin.org? > > Yes. Unfortunately it did not talk about how the Bayesnian score is > calculated (or if it did, I missed it). > > >> Apart from this, your ISP is the correct contact >> here. You are their customer they have to answer you. > > Sadly, like most companies (not just ISP's) - their response was "if you > pay for a higher level of service, you can turn off SA".
And if that is the attitude of your ISP (putting SA on sitewide and wanting you to pay extra for the privelledge of turning it off again) then I suggest you seek another ISP. That sort of attitude goes *completely* against the developers wishes - the developers don't like ISP's/companies installing SA sitewide without telling their users and without providing a way to easily opt in/out. I'm just about to roll out a sitewide install of SA, and after a lot of consideration and agonizing over it, I ended up deciding that while it would be on by default for all users, the default action would be to simply tag the message as spam in the subject and headers but pass it along. (report_safe 0 mode) This gives the individual user lots of control - they can see if SA is "getting it right" on their messages during an "evaluation" period without any chance of missing out on something important, and if they wish they can simply use a subject/header based filter on their email program. If they decide they trust it, they can then customize their settings by way of a web interface integrated into a webmail page, which includes being able to quarentine spam in an imap spam folder on the webmail (saves downloading it via pop3) change their threshold, add their own whitelist/blacklist entries and change a few of the more useful SA settings, and if they so choose, even disable running SA completely. (Total opt out) Never lock your users into something they don't want.... > However, they > do seem to be willing to look at the configuration settings that SA is > using, but it would help if I could tell them what configurations > settings to use. > read that to say: if I tell them what to use, they'll probably > adjust the settings. But if I don't they're going to stick with > the same defaults that are causing the problems. So, if you do their homework for them, they might, if you're very lucky, help you sort out a problem that is there because of them. Ditto - find a better ISP. > Another thread on this forum has already proved helpful. It said that > the Baynesian tests are useful only on an individual basis - on an ISP > level they are much, MUCH less useful. It also said that the default > setting is still to have it turned "on". Well actually I, and a number of people have found the reverse - that sitewide bayes using the auto-training works very well. However it means you can't do per-user manual training. Maybe one day SA will support the concept of simultaneous site-wide and per-user bayes. I'm not sure if its even possible, let alone practical however, but we can always wish :-) >> It seems >>> that when I have a friend that uses HTML formatting (or heaven >>> forbid, Incredimail) they have a MUCH higher chance of being tagged >>> as a spammer. >> >> If they make heavy use of colors, font size etc.: yes. > > Incredimail also puts a "click here" (in html formatting) at the bottom > of everyone of their email messages. Of course, a multi-word Bayes test > would check for "click here" AND Incredimail's tag line and flag it as > NONspam. And incredimail makes extremely gratuitous and unnecessary use of X-headers, html formatting, and advertising tags, so its not surprising it looks a little bit like spam. Every time I get an email from an incredimail user in Eudora, (fortunately it doesn't happen too often :) it makes me go UGH, because the default action of Eudora is to show X-Headers, and incredimail has about 10 or 15 utterly useless X-Headers which I have to scroll past to see the actual message. (Which is inevitably in html, and includes a silly background image, and that click here image and link) >From that you can probably guess I don't like incredimail :) (I suspect a few of the **ix diehards on this list that use stuff like pine etc don't like it either...) Regards, Simon ------------------------------------------------------- This SF.net email is sponsored by: Etnus, makers of TotalView, The best thread debugger on the planet. Designed with thread debugging features you've never dreamed of, try TotalView 6 free at www.etnus.com. _______________________________________________ Spamassassin-talk mailing list [EMAIL PROTECTED] https://lists.sourceforge.net/lists/listinfo/spamassassin-talk
