Hi Alexander,
Monday, September 6, 2004, 9:45:09 PM, you wrote: Alexander> Hello Doug Weller, Alexander> 06-Sep-2004 20:37, you wrote: >> I'm thinking particularly about whether it is best to have it edit the >> subject line or add an x-header, but any other tips will be welcome! Alexander> I don't like the subjects to be altered (you have to think of it each time Alexander> when you reply/forward unless you clean the subject in TB! with a Alexander> regex-enhanced template for replies and forwards) - obviously, I'm using Alexander> the x-text-classification header. :) I agree. My ISP filters spam and adds a header which I can configure, but I think I shall turn that off as I hate having ham which has [probable spam] in the subject line! Alexander> This has other advantages: you can configure TB! to show the header, as Alexander> well as the reclassification URL of PopFile in the header pane of the Alexander> preview window, thus checking wether the Alexander> classification is correct and, in Alexander> case PopFile made a wrong classification, Alexander> reclassify is only one click Alexander> away. *nice* Good idea. >> Can I expect it to be faster then the current default plugin in V3.0? Alexander> Depending on how much mail you get (well, being Alexander> on this list alone brings Alexander> quite some each day ATM...) it may be slower (but I used BayesIt only a Alexander> very short time). After all PopFile is written in Perl, and Bayes analysis Alexander> requires a database (mine is 1.4MB right now). I hope it's not slower, filtering seems to be slowing down email delivery right now for me. And my wife's is terrible, 23 emails taking 3 minutes, I have no idea why as she has a very small database. She's using v.3 also but that made no difference, nor did getting rid of the dozen .tmp files, compressing, etc. Alexander> And be aware that PopFile can be quite a resource hog - using concurrent Alexander> POP connections results in 100% CPU-load peaks Alexander> during mail retrieval, and Alexander> memory usage is something like 20MB here, always. Alexander> Another note: of course, every Bayes filter Alexander> requires training - it needs to Alexander> know which messages is spam, which is ham, at Alexander> least (but keeping only spam Alexander> and ham apart is not all PopFile will do for you, of course). That is so Alexander> much easier in PopFile with the use of magnets. For example, set up a Alexander> magnet that will put every mail from this list Alexander> into a bucket "genuine" (or Alexander> whatever you want to call it) - PopFile learns Alexander> from every magneted message Alexander> and you have a big "ham" database in no time. I hall do that, good idea. [SNIP] Thanks to everyone who replied. Doug -- Doug Weller Moderator, sci.archaeology.moderated The Bat! 3.0 Doug and Helen's Dogs: http://www.dougandhelen.com Doug's Archaeology Site: http://www.ramtops.co.uk ________________________________________________ Current version is 3.00.00 | 'Using TBUDL' information: http://www.silverstones.com/thebat/TBUDLInfo.html
