Chapter 1: Introducing qmail

---

Highlights

Andy wants to send an e-mail message to his friend Josh. He opens his mail client, clicks on New Mail, enters Josh's address in the To field, fills in the Subject field with a short description of the message, and types the message into the large editing area of the form. When he's done, he clicks on the Send button. As far as he's concerned, the message is sent, but behind the scenes, complicated machinery whirs to life. A thousand tiny steps will be executed on Andy's behalf by processes on various systems between Andy and Josh—who could be in the same room or half a world away.

The Internet Message Transfer Agent (MTA) is the key player in the behind-the-scenes e-mail infrastructure—it's the machinery that moves e-mail from the sender's system to the recipient's system.

Before the Internet explosion in the early 1990s, one MTA, Sendmail, was responsible for delivering almost all of the mail. But Sendmail was designed for an Internet unlike the modern Internet. At the time Sendmail was created, there were only a handful of systems on the entire Internet, and most of the people online knew each other. It was a friendly, cooperative community that consisted mostly of the people who wrote the software that made the Internet work or managed the hardware that it connected. Security was not a major concern: There was not much that needed protection, and there were few potential "bad guys" from which to be protected.

The modern Internet is very different. It's millions of times larger, so knowing all the other administrators and users is impossible. In fact, it's accessible by anyone with access to a public library. Billions of dollars in business and consumer commerce takes place annually over the Internet. Large corporations exist whose entire business model relies on their Internet presence. As such, the stakes are high, and it's no longer possible to treat security casually. On top of all this, servers are being subjected to staggering loads—a typical mail server today might send more messages in one day than a mail server ten years ago sent in one year.

The Sendmail developers have worked hard over the years to enhance its security and performance, but there's only so much that can be done without a fundamental redesign. In 1995, Daniel J. Bernstein, then a mathematics graduate student at the University of California, Berkeley, began designing and implementing an MTA for the modern Internet: qmail.

While Sendmail is one huge, complex program that performs its various functions as the superuser (the all-powerful Unix root account), qmail is a suite of small, focused programs that run under different accounts and don't trust each other's input to be correct.

While Sendmail plods through a list of recipients delivering one message at a time, qmail spawns twenty or more deliveries at a time. And because qmail's processes are much smaller than Sendmail's, it can do more work faster, with fewer system resources. Further, Sendmail can lose messages in some of its delivery modes if the system crashes at the wrong time. For reliability, speed, and simplicity, qmail has one crash-proof delivery mode.

---