After beating a dead horse once, you have to do it again.

Quite some time ago now, I ran in to an article where Wietse Venema was interviewed about security focused programming. He is the guy who initially started writing postfix. While he does have some good insights in to why the internet is getting more insecure, there are some things that I think he is missing.

While I understand that reinventing the internet is pointless, there are some things that need to be reinvented. The SMTP protocol is one of those things. The protocol itself was designed for a completely different type of a network. It doesn't do any authentication or verification for the sender. It basically trusts everything that is fed to it. Now, Wietse offers nice ideas on counteracting spam, but the inner problem still persists. The reason why we have so much spam in our mailboxes is because of SMTP.

I was about to write some code and put my money where my mouth is, but today I ran in to yet another scheme to stop spam and realized that I would eventually forget to write this all down before I get the time to write some code. The scheme I ran in to is which allows you to register the mail servers that are allowed to send mail for your hosts for a nominal(?) fee of $20. While this nominal fee will keep some of the spammers off the lists, it's still reinventing the wheel and trying to fix a symptom of the problem.

There are similar projects, that work with different types of protections. There is SPF, DomainKeys, Sender ID and many more that aim for similar solutions.

I'm not saying that we don't need those kinds of solutions, but rather the solutions are solving the symptom and not fixing the problem. The only real solution is to replace SMTP with something more suitable for the task. The thing that comes in to mind is XMPP. XMPP is a protocol designed for XML packet routing. It's mostly used as an instant messaging platform, but it's not a huge leap to transfer mail through the protocol.

XMPP as a protocol is designed on an age when spoofing and spamming was already a problem and it has safeguards in place to prevent malicious activity. The protocol is suitable for transferring e-mails already so no real modifications are needed for the protocol. Only thing that needs to be done is to document the common practice.

Changing an internet protocol is a large task and one can't take it lightly. The beauty in it all is that most mail servers are already capable of supporting multiple transport protocols. Initial versions of Sendmail delivered the mail through FTP (with some obscure extensions), so switching protocols isn't really that far fetched. Today, Sendmail supports various protocols while the most commonly used one is SMTP. Same goes for postfix and other mail servers. So implementing a new protocol isn't really out of the question.

I'm sorry that I wasn't able to write the code to back this all up, but at least the idea is out there in written form. It shouldn't be too complicated to implement this and get things started. The change won't happen over night, but it has to start somewhere.

Comments on this page are closed.