On Fri, Mar 17, 2006 at 11:38:27PM +0100, Santiago Vila wrote: > However, I bet that if people cared to report each and every of these > "bugs" using the BTS (specially, for every spam email which isn't stopped > by the filters and it's distributed to the lists), the Debian listmasters > would be completely unable to manage the huge amount of bugs that would result > from this interpretation of what is a bug.
False positives are a much more serious problem than false negatives. > > > i suggest that you add all your posting addresses to out > > > pseudolist 'whitelist', so you won't get penalized for not being > > > subscribed to a list. > > I agree with this suggestion. I should not have to take special steps to work around broken filtering rules, in order to allow them not to be fixed. > If those rules exist at all, it's because they are useful, and if they > are useful it's because many spam emails get those flags, and many ham > emails do not. "Useful" doesn't always mean "correct". I'm sure one could create a whitelist of ISP and corporate relays, and reject everything else (including non-"dynamic" systems). That would probably give nice-looking filtering statistics; an overwhelming bulk of mail is probably Average Joe using Outlook Express through those relays. That doesn't mean it would be an acceptable filter. Similarly for filtering mails that talk about viagra and get-rich-quick- schemes--many spam mails talk about them, and many ham mails don't; the overwhelmingly vast majority of mail that does is spam. That doesn't make it correct to have a 6.4-point ".*viagra.*" filtering rule (or overlapping 4.4 and 2.0-point rules). That's turning off SpamAssassin's heuristics entirely; it's no longer a heuristic filter, but an old-school "blacklisted IP ranges" filter. -- Glenn Maynard -- To UNSUBSCRIBE, email to [EMAIL PROTECTED] with a subject of "unsubscribe". Trouble? Contact [EMAIL PROTECTED]
