Submitting your vote...
Yesterday Google give a link to very nice article on Spam Resource web site. It explains why autoresponders, non-delivery reports and other notifications are bad and gives some numbers. Read the part below and, if interested, visit Spam Resource.
Let's do some quick math on the back of a napkin. A quick check of my personal spamtrap account finds 2039 pieces of backscatter, just by searching for a few common phrases found in bounce messages and challenge/response requests. Out of the 320,000 recent pieces of spam I've got in that account, that may not seem like much. But that's just me and my tiny domains. Think about a big site, like say, Outblaze, with millions of users across hundreds of domains.
Spam lists contain a high percentage of invalid addresses, driving a high bounce rate. A normal mailing to a legitimate list will result in 3-10% of the mail bouncing, per mailing. And that's if you handle bounces properly. When working with senders, I know that if they haven't tracked bounces properly in the past, a list can have anywhere from 30-50% bounce rate. Spammers rarely handle bounces properly, so let's assume a 40% bounce rate on a spam run. If a spammer sends two million pieces of spam, that leaves 800,000 bounces. Where do those bounces go? I know from helping clients legitimately handle bounces, maybe 7-10% of email servers accept the mail then bounce it back later.
Do the math based on all of these assumptions: 2,000,000 spams, 40% bounce rate, 9% of mail servers send backscatter... That means that for that two million spam run, 72,000 bounce notifications (NDRs) are going to be sent back to the sender address. Since spammers forge the sender's address, this mail is going to be be received by people who had nothing to do with the spam. This, in a nutshell, is backscatter. And there's a lot of it floating around.
I would only add: those 72,000 bounce e-mails will come to innocent person, whose address was used by spammers! Isn't it a reason for ISPs to stop backscattering?
