Originally Published: Monday, 27 August 2001 Author: The Staff of Linux.com
Published to: develop_articles/Development Articles Page: 1/4 - [Printable]

Interview with Eric Allman, CTO of Sendmail Inc.

Today, the first day of Linux World, San Francisco, Sendmail Inc. one of the world's most successful open source companies and IBM (big blue) announced a joint partnership to deliver powerful software for mainframes on Linux at a scale the world has perhaps not seen before. This terrific validation of the open source development method is cause for celebration. Linux.com caught up with Eric Allman, CTO of Sendmail, Inc. (who wrote the original sendmail back in 1981, arguably the most important piece of network software ever written) to ask the guy some questions.

Sendmail and Working with Open Source   Page 1 of 4  >>

As Sendmail, Inc.'s chief technology officer and co-founder, Eric Allman leads the company's technology strategy and developments. Allman authored sendmail, the world's first Internet Mail program, in 1981 while at the University of California at Berkeley. He continues to spearhead sendmail.org, the global team of volunteers that maintain and support the sendmail Open Source platform.

Today Sendmail Inc. and IBM announced a partnership to develop a suite of products for the IBM eServer z900 mainframe, including Sendmail Switch, Sendmail Advanced Message Server and Sendmail Mobile Message Server on Linux for mainframes. These solutions provide ISPs, telcos and other large network organizations the opportunity to provision high performance software that can handle more load (up to two million users on a single machine), more reliably (99.999% uptime) and at a lower total cost than ever before. It's an agreement that represents a terrific validation of everything the Open Source movement holds to be true.

Linux.com caught Eric Allman to ask him some questions intense about morality and puddings, among other things.

Linux.com: Why did you write the original sendmail, back in 1981?

Eric Allman: Berkeley had connections to the ARPAnet and UUCP, with the ARPAnet being on a PDP-11 owned by the INGRES (Interactive Graphics and Retrieval System) database management project, which I worked on, and UUCP being on a departmental VAX machine. There was also a third home-grown network called BerkNET, written by Eric Schmidt (now CEO of Google).

But none of these networks talked to any of the others. In particular, to send or receive ARPAnet mail, you had to have an account on the INGRES machine, which was quite overloaded as it was. We actually did bring up accounts for all the faculty and some of the graduate students, but the situation was untenable for the long run.

At some point I realized that these networks were (primarily) software, so all I had to do was write some "glue software" to make them talk to one another. I created a predecessor to sendmail called delivermail. Use quickly exploded at Berkeley, and after not too long many other sites started using it as well.

When Berkeley got the ARPA contract to do VAX software for the new (as yet unbuilt) "Internet" thing, designed to replace the ARPAnet, I added SMTP client and server support to that program and redubbed it as sendmail.

Linux.com: Can you tell the Linux.com readers who don't know very briefly what sendmail is?

Eric Allman: Sendmail is a Message Transfer Agent (MTA) - the software that knows how to route mail from one place to another. This is in contrast to the Mail User Agent, which is what people see on their screen, and the Message Store, which holds your inbox until you get a chance to read your mail.

Linux.com: When did you form Sendmail, Inc.?

Eric Allman: We opened our doors in March of 1998.

Linux.com: That was before many people were trying to combine "Open Source" - software development by individuals and the academic community with - commercial businesses. What gave you the idea to make the leap?

Eric Allman: Two reasons really. First, I personally viewed that as a requirement. I may have written sendmail to solve a local problem at Berkeley, but I continued working on it because I saw that e-mail was going to change the world - or rather, continue to change it - and I felt that change was going to be fundamentally positive, and that sendmail was a big part of that change. However, I did need to make money because sendmail had become such a success that a small crew of volunteers couldn't keep up with e-mail support and do development at the same time. But getting rich wasn't my primary goal, and ultimately I felt that the best way to contribute to the Internet was to keep sendmail Open Source.

Greg Olson, my business partner and the "business mind" behind Sendmail Inc. did some market research and ultimately agreed with me. He was adamant that Sendmail, Inc. needed to be a "real" company right from the start, and a big part of being a real company is making a real profit - something I very much agreed with. Being Open Source gave us market share, name recognition, and street creds: intangible perhaps, but very valuable. Frankly, I think that closing up the sendmail source code would have been suicide.

Linux.com: Sendmail, Inc. is often cited as a model commercial open source company. To what do you attribute your success in that area?

Eric Allman: As noted above, we started up the company with a goal of building a real company, unlike so many Internet Bubble startups that seemed to take the attitude of "let's get mind share, and we'll figure out a way to make money later." But maintaining true to the Open Source users was also essential. And it helped that Greg and I came from two very different perspectives. We had many "frank and comradely" discussions, and we continue to do so. Far from weakening things, these different perspectives made us stronger. Another important point is that Greg and I agreed early on that we needed experienced management. Too many bubble companies (which isn't the same thing as open source companies at all, I hasten to add) thought that dropping out of school was the best qualification to be CEO. We looked for people who had some background, and I think that was a good decision. However, I will admit that it took some time to educate some of these people about open source, which is often considered to be antithetical to making money - which of course it is not.

Ultimately, we created a "hybrid" company, rather than a totally "open source" company - that is, we have open source software in the form of the sendmail MTA, which remains the core of our products. But we have embellished that with monitoring and management tools, a message store, POP and IMAP servers, web and mobile access, and so forth, which gives us a much broader product line.

By the way, Greg and I are still friends - a somewhat unusual situation for co-founders, I'm told.

Linux.com: How did Sendmail, Inc. successfully marry the needs of a commercial entity with the needs of the Open Source developer community while maintaining the integrity of the core technology?

Eric Allman: First of all, we decided early on not to do a "pure" open source play. That is, while we are committed to keeping the core MTA technology as open source, we are selling other parts of the software in a more traditional old-school business model. For example, the graphical user interface is not open source. On the other hand, this is not software that would typically be of special value to the developer community, who want power-tools, not simple tools.

This division makes it reasonably easy to build a Chinese wall between the open source technology and the rest of the company. There have been fights (er, "discussions") about how this should work, particularly with schedules. For a while we tried to keep the open source releases synchronized with commercial releases, but that didn't work well, and we've reverted to releasing the open source MTA more-or-less when the engineers say that it's done, which seems to work best for everyone.

Linux.com: How did the engineering practices change when moving from a pure Open Source environment to a commercial environment?

Eric Allman: In some ways in the best possible way: we have more people able to work on sendmail full time than we had before. In other ways, the changes have been much more subtle.

One of the first things that I noticed is that there are far more versions of the code in play at any given time. In the pre-company days, we pretty much finished one release (that is, released it and let it stabilize) before starting the next. Today, we have both the current and next release active at any given time, plus the versions that go into commercial releases (these are very close to the open source releases, but in some cases may have slight variations; it turns out to be easier to just create a separate branch early on), plus any special customer branches, plus experimental or porting branches, and so forth. We've been using CVS for source management since we started the company, but that's starting to run up against the wall, so we'll probably have to re-evaluate soon.

There are also, necessarily, people like management and marketing involved. They fulfill a valuable role, but they have very different perspectives on things, some of which impact engineering.

Linux.com: Has the Sendmail Consortium (i.e. the Open Source initiative) enjoyed benefits from the commercial venture? How involved are you with the Sendmail Consortium?

Eric Allman: Absolutely. As I mentioned above we now have more people working on open source development than we did before, and if we need to buy commercial software to help us develop, debug, or support sendmail we have resources to do so.

Perhaps more importantly, we now hold an annual conference of the core people in the Consortium. This costs a non-trivial amount of money, because these volunteers come from all over the world. But a huge amount of idea exchange and review gets done in a very short amount of time. It's well worth the effort.

Personally, one of the down sides of founding a company is that there is always too much work to do, and sadly I find I don't have much time to code any more. I do some occasionally, but I try not to get pulled into any schedules - these days, it seems like everything I do is late.

The annual meeting of the Sendmail Consortium and friends (the "Meeting of the Minds") focuses almost exclusively on the open source MTA. However, these people are far from stupid, and they understand and appreciate the needs of a profit-driven company. Inevitably, advice about commercial releases is offered. But the focus is very definitely open source.

Linux.com: When you started the commercial division sendmail already had a massive dominance - it was used all over the world. How important do you think that was to building Sendmail, Inc. as a company?

Eric Allman: I think it was very useful to us. It certainly helped us during the funding stage - it never hurts when the investors have already heard of you. But it has also been, and remains, useful to us during the sales cycle. It doesn't automatically sell the product, but it helps get us in the door. Just like investors, customers like to buy things they have heard of to get the job done.

Sendmail and Working with Open Source   Page 1 of 4  >>