On May 17, 2005, at 10:22 AM, A-NO-NE Music wrote: >> in the last four weeks SS has been changing spam ratings to messages >> coming from historical senders which have never been identified as >> spammers and now, they suddenly are. > > I am glad I am not alone! I reported this a month ago, which seems to > coincide with yours. Mine started roughly around OSX10.3.9 release > time, tho.
I don't think I have any messages from you on that subject. I try to follow the list, but sometimes miss messages. If you want to be guaranteed a response, please write to <[EMAIL PROTECTED]>. > Many spam those which should had been caught are now missed, while > a few > legit ones are marked SPAM and sent to Spam folders almost every day. The best thing to do in a situation like this is to look at SpamSieve's log (or ask me to do so). It will tell you which messages it predicted to be spam or good, and why. And it will tell you which messages it was trained with and whether it recognized them as false negatives or positives. In most cases where there are a lot of "misclassifications", the log--which represents SpamSieve's view of the world--doesn't agree with what you've observed in the mail program. This indicates a configuration problem (e.g. with the filters in the mail program) rather than an accuracy problem and is generally straightforward to correct. On May 17, 2005, at 10:51 AM, Michael Lewis wrote: > Maybe this isn't the list to ask. SpamSieve seems its the culprit here > and it may be better to check the SpamSieve site or list. Here is > information about subscribing to the SpamSieve Talk list: <http:// > lists.c-command.com/listinfo.cgi/spamsieve-talk-c-command.com>. I try to keep tech support off the mailing list. > It'll be interesting to see how this holds up as I run it more and the > corpus and rules grow. Maybe it would be a good thing to rebuild the > corpus and rules once a year? It's not necessary or particularly helpful to clean out the rules, though you can if you want. Rebuilding the corpus every year or so (depending on how many messages you receive) *is* useful, though. If you've got training tips enabled in SpamSieve, it will tell you when to consider doing that. -- Michael Tsai <http://c-command.com>
