evaluating switch from JunkMatcher

I’m evaluating SpamSieve as a replacement for JunkMatcher; JM was doing wonderfully, but the lack of movement on the Python RegExp issue has forced me to look elsewhere.

I’ve got SS set up and it appears to be working wonderfully; I do have a few questions however:

  1. I use Apple Mail.app and get mail from two sources: entourage (IMAP, apparently) and a POP account. If SS gives me a false positive, and I tell it to train as good, I’d like it to correctly move the mail back to the corresponding inbox, depending on the account used to retrieve the message in the first place. It’s not clear it can do this…

  2. When I look at my statistics, I see an ever increasing number of messages filtered, but my corpus stats aren’t changing. Since I have SS set to learn from new messages, shouldn’t all that learning become part of the corpus?

  3. is there any way to change the colors that SS uses to colorize spam? I find the colors very useful, but very non-obvious. Blue just doesn’t scream spam at me in the same way as red.

  4. Can SS add a header to messages with it’s spam scoring? I’d very much like to be able to show all headers and see what the score a message received was.

Thanks,

  • a.
  1. SpamSieve makes an effort to move the message back to the correct inbox by (a) asking Mail which account the message came from, and (b) seeing if the To address matches the addresses of any of your accounts. Unfortunately, these methods sometimes fail, in which case it defaults to the inbox of the first account. You can drag the accounts in Mail’s Preferences window to re-order them.

  2. No. SpamSieve will learn whitelist rules from all the messages, but it doesn’t fill the corpus with every message that comes it. Instead, it chooses which messages it wants to learn from, based on their contents and how different they are from the messages it’s already seen. This allows it to adapt more quickly to changes in spam, because the corpus isn’t full of lots of nearly identical messages.

  3. Sorry, the colors are limited to the ones that are built into Mail. They aren’t the ones I would have chosen, either. :slight_smile:

  4. No, SpamSieve does not modify the contents of messages. You can see the approximate score from the color, or look in the log to find the exact score. (And, in Eudora and PowerMail, there is a column in the message list for showing the score.)

Great, thanks for the reply Michael.

A couple of follow-ups:

  1. For the color, I see that you’re choosing from the colors listed by default in the “Set Color of Message…” action, but that action also has an “other” option, which lets me set any color of my choosing. Is that just not available programmatically?

  2. Plug-ins like MailTags add columns to the view, which track their own variables per-message. No chance for SS to do something similar for the score?

Finally, I’m a little confused by rule ordering. I subscribe to a bunch of mailing lists (30+ per account) and have those messages moved to folders based on rules. Rather than have my spam blocker, whatever it may be, do work on every one of those message I simply accept the fact that some spam may make it into those folders, and only have my spam blocker deal with the 400-500 messages of spam I get to my personal inbox.

I’ve got my SpamSieve rules at the end of my filters; all the rules that handle lists are at the beginning, and include the “Stop Processing More Rules” action. Yet I’m seeing some of the messages that should be moved into folders in my spam folder (and colored, so I know they’re being handled by SS)… any thoughts?

Thanks,

Andrew

The interface that the plug-in uses to set the color does not support custom colors. There’s probably a lower-level way that would allow custom colors, however an important consideration is that Mail’s sort order for the colors must match their associated scores, and I don’t think sorting of custom colors is well-defined.

Mail doesn’t actually have a plug-in interface. Anything that SpamSieve or MailTags does involves overwriting bits and pieces of Mail’s code in ways that its developers didn’t expect. The more the plug-in changes, the more risk there is in breaking something in a subtle way, especially if you try to add extra data to a message. Also consider that there will be other plug-ins, such as MailTags, trying to change the messages at the same time. The current SpamSieve plug-in tries to make the smallest possible modifications to Mail. It hooks in at some key places, but it doesn’t really change anything. This safe approach is what allows it to work reliably from Jaguar to Leopard, even though Mail has undergone tremendous changes. So that’s my thinking. I’m not saying that the SpamSieve plug-in won’t ever do more, just that there are trade-offs, and that adding a feature in a reliable way might not be at all as easy as it appears.

It sounds to me like your list rules didn’t match the messages. You can test that by disabling the SpamSieve rule and re-applying the rules. Also, I don’t think you need a “Stop Processing” action of you have a “Move Message” action.