Hi Philip, On Friday, January 23, 2004 at 3:58 PM, Philip wrote: > Friday, January 23, 2004, 10:58:14 AM, you wrote:
T>> That's the same question I had. As I have SpamAssassin on the mail T>> server, most of my spam gets caught and filtered that way, but I was T>> hoping that this would catch the one or two that slip by every week. > If SpamAssassin on your mail server is preventing spam mails from > getting to your copy of The Bat!, then you should really disable it. Thanks for your response, but I think you must have misunderstood what I wrote. I'm not having any problems with SpamAssassin on the mail server. By "caught" I mean that SpamAssassin is doing exactly what it is supposed to do and works exactly how I've set it up - it evaluates all incoming mail for spam, adds the appropriate headers and forwards it to my mailbox where I then download all mail. I then use filters on TB to separate out the spam based on the header information. > Having two spam filters is not actually a good idea in most cases. If > one deletes a spam or otherwise stops it from being seen by the other > filter, then the other filter cannot evaluate it. Worse, if it if a > statistical filter - a bayesian one like bayesit - then that filter > will not learn about the spam's content and be able to update its > filtering. I have used two filters in the past with no problems at all. I used a SpamAssassin and popfile combination for almost a year. Popfile caught those that SpamAssassin scored low. When the plug-in became available, I switched to the BayesIt plug-in and Bayes Filter plug-in instead of popfile (one less program that's running). Neither has worked. I've read multiple posts both on this list and on the RitLabs Forum to know that I am not the only one having this problem with BayesIt. It works for some and not for others. I haven't heard much about Bayes Filter. > The only exception for this that I can think of is the way The Bat! > implements its plug-in architecture - which allows multiple filters, > all of which report a score. You can then choose to act on the lowest > (minimal) or highest (maximal) amongst the results, or to use an > average from all filters. (See the anti-spam plugin filters preference > page for details - it's not obvious at first!) > This multi-plugin capability means that The Bat! would decide what to > do after all filters have had a chance to look at the mail and report > their findings. It also means that marking mail as junk would cause it > to be analysed by multiple filters, improving the detection by all > filters in the future. I'm aware of how the plug-ins are supposed to work and how e-mails can be scored by multiple plug-ins, scores can be averaged, etc. Unfortunately, none of the current anti-spam plugins work as effectively as SpamAssassin does on my server. That's my experience. YMMV > However, at the moment, I'm only aware of one free filter plugin for > spam - bayesit - so this very nice feature of The Bat! goes unused as > yet. Hopefully you can now see why the plugin architecture is so > valuable, though. I'm very familiar with the benefits of the plugin architecture, but thanks for the review. There are three anti-spam plugins available, BayesIt and two others. One is called VampireX. It doesn't use bayes classification and only uses regex. My preference is Bayes and since I haven't heard any positive comments about this one, I haven't used it. http://www.softpedia.com/public/cat/10/3/3/10-3-3-161.shtml There is also one by Achim Winkler that is called Bayes Filter, I think. http://www.lkcc.org/achim/download/ It's described on TBDEV in this message. http://www.mail-archive.com/[EMAIL PROTECTED]/msg00532.html > SpamAssassin should at most be marking suspected spam with some kind > of subject alteration or header, and then you should let bayesit candle > it as normal. If you're filtering on modifications made by > SpamAssassin and not marking the filtered mails as junk for bayesit, > it will never learn. That's what I do. BayesIt still doesn't learn. I've also updated the kludges file on the off chance that the case of the headers mattered. I suspect, based on the number of people that seem to be having problems, there may be some issues with the plugins that haven't been worked out. It could be any number of things. At one time, there was a report that one version of BayesIt wasn't working on Win98 because the author was relying on some system variables that were Win2k and XP specific. That was fixed. Someone's reported that it took over 500 spam mails until it started working. Someone else reported it started working after 10 spam mails. Who knows? It never worked for me and I average 50 spam mails a day and I had both filters in place for several months. I don't really have the time right now to spend on trying to troubleshoot the plugins. I've uninstalled both BayesIt and Bayes Filter and will now try them out one at a time to see what's happening. Maybe something will become obvious over time. To date, BayesIt hasn't identified any spam. I've checked the logs and it is seeing the e-mail. Duplicate spam e-mail of ones already trained were received and not filtered. These are also from an account that I have that is not on my server and is not filtered by SA. Frankly, it isn't a big problem for me. SpamAssassin works amazingly well and working plugins would just be a bonus. > To be honest, because of this I would tend to say that you should run > only one anti-spam precaution - unless additional ones are also > plug-ins for The Bat!. Anything else risks not correctly training the > bayesit plugin. This would only be true if I took some action that would prevent the anti-spam plugin from seeing the mail. > Of course, this could also be a failing in a beta of The Bat! - in > which case, aren't you on the wrong mailing list? ;-) I'm not using the beta, so I am on the right list. -- Regards, Terry Using The Bat! v2.01.3 on Windows XP 5.1 Build 2600 Service Pack 1 ________________________________________________ Current version is 2.02.3 CE | "Using TBUDL" information: http://www.silverstones.com/thebat/TBUDLInfo.html