Here's the IETF Draft on the spec.
http://datatracker.ietf.org/doc/draft-ietf-marf-base/Forgive me if you are already familiar with this, I feel a little explanation may be helpful. Most email providers and ISPs have a "Customer Feedback Loop". Basically, this sends an email in ARF format to a participating mail sender, when a user clicks the SPAM button on one of the messages sent by the participating mail sender. Obviously, this only works if the SPAM button being hit is in a UI controlled by the ISP, like a Webmail interface or custom app. Standard mail readers like Outlook, Mail, Thunderbird won't generate an ARF in response to the SPAM button. The mail sender is supposed to remove the email address from the list pronto. Spamhaus covers this a little more in-depth here:
http://www.spamhaus.org/faq/answers.las ... Issues#119In my very short experience with this, some email providers send the actual email address with the ARF, and others (notably AOL) redact the email address and only send the Message-ID header from the offending email. I am still trying to figure out if Dada is sending this or if MIME::Tools is setting up the Message-ID header, and how I might setup my mail server to log it so I can honor the removal requests.