Apps of the Year and SpamSieve Scores
I’m honored that John Gruber’s list of Apps of the Year, 2004 includes SpamSieve. Gruber does a good job of explaining the utility of SpamSieve’s scores feature, which makes it easier and quicker for you to scan your spam for false positives by telling you how spammy it thinks each message is. Most people, I suspect, don’t use scores because they don’t get that much mail, but if you do get a lot, they’re quite helpful.
SpamSieve’s scores have a different meaning than scores from other anti-spam software. In the traditional approach, each message is assigned a score (say, from 0 to 100). The user can then set a threshold saying, for instance, that messages with scores 90 or higher are spam; those with scores less than 90 are good. Adjusting the threshold lets you adjust the aggressiveness of the filter, and by looking for messages with scores near the threshold, you can tell which ones the filter was uncertain about.
For a variety of practical and philosophical reasons, SpamSieve works differently. If it thinks that a message is good, it assigns it a score from 0 to 49; if it thinks it’s spam, it assigns it a score from 50 to 100. Thus, looking at a score, you can tell (a) whether the message was classified as spam, and (b) how sure SpamSieve was, without needing to know the current threshold. It’s always 50. You have two controls:
- If you want to vet a lot of spam messages, you can look at those with scores less than, say, 92. If you want to vet only the really borderline ones, just look at the ones with scores 75 or below.
- If you want to make SpamSieve more or less aggressive—thus affecting the scores that are assigned to your messages—you can adjust its bias.
E-mail clients show scores in different ways. Eudora and PowerMail show the score in its own column in the message list. In Mailsmith, you can use a script to label messages according to their scores; you can set the uncertain threshold in the script. In Apple Mail, SpamSieve uses colors to show the score. In Entourage, very spammy messages are assigned the “Junk” category, and less spammy messages are assigned the “Uncertain Junk” category; you can set the threshold in the Notification tab of the preferences.
Personally, I use Mailsmith. I could use the above-mentioned script to label spam messages by their scores, but I prefer another approach: I have SpamSieve notify me with Growl when a borderline message arrives. Every suspected spam message goes into my (spam) mailbox. If a message with a score less than 75 arrives, Growl shows a notification window that tells me the message’s subject, its score, its sender, and the first few lines of its body. This way, I get instant notification that a possibly good message has arrived (just as I get a notification in the Dock when genuine good mail arrives). If, as is usually the case, SpamSieve was right and the message is spam, I can dismiss the notification with a single click. This way, I only have to bring Mailsmith to the front if there’s real mail to read.
