Reply to comment

Sorry for overlooking your post on the old forums for so long. Do you mean training a mail client, such as Thunderbird or Apple Mail to recognise spam, or training SpamAssassin on a shell account? You do not have a per-user SpamAssassin profile the way we do things – meaning, with luck, that all mailbox owners benefit from any training, regardless of whether they do it themselves (also potentially allowing it to be poisoned, which is why we don’t publish the method of training it.)

Basically, we don’t rely on the Bayesian filter. URL lists and the rules we have written – specifically with the needs of civil society groups and activists in mind – are more important. However, you can see how the Bayesian filter is doing by checking the X-Spam-Status header added by SpamAssassin. Non-spam (vege-ham) should get BAYES_00 or BAYES_20, while the “spammiest” messages get BAYES_99.

If you want to contribute to the training of the communal filter, which is most useful for unrecognised or conflicted (BAYES_50) messages, you can forward spam and nonspam messages to fairly obvious email addresses that can give you. These need to be sent from a GreenNet account and as an attachment to be accepted.

(This reply is largely duplicated at http://www.gn.apc.org/forum/i-get-so-much-spam?page=1#... )