From: Robert Klemme Date: 2009-06-06T20:45:06+09:00 Subject: Re: SPAM from Usenet On 06.06.2009 08:23, Bill Kelly wrote: > This is starting to sound partly like a gmail problem. I think rather not. > I've used the same email address for 22 years, and I've > never made much attempt to shroud the address in any > online posting. As one might expect, I receive a TON > of spam. Hehe, sounds as if someone is paying you as a honeypot provider. :-) > The ruby-talk spam is a drop in the bucket. I suggest > it's partly a gmail problem, because (apparently) > gmail doesn't allow any kind of local filtering (being > web-based.) Actually, IMHO this is wrong for two reasons: first, you can read GMail via POP or IMAP. Second, GMail's spam filtering is pretty good and I believe your spam marks train the filter. I believe at least in theory a spam filter of a mail provider with web access can be much better than a local filter because there is more training input. > As an aside, here's an idea for a ruby quiz. Start > with a pre-existing Bayesian spam classifier and a > huge corpus of spam messages, and have quiz participants > augment the existing classifier by inventing new methods > of analyzing the kinds of messages that slip through > (such as ones that play HTML/CSS tricks to hide verbiage > intended to fool Bayesian filters, etc.) That sounds more like a competition than a quiz, doesn't it? The winner will be the software with least false positives and most identified ham. :-) Cheers robert -- remember.guy do |as, often| as.you_can - without end http://blog.rubybestpractices.com/