Michael Tiemann - interview

From LXF Wiki

Unlikely warrior

Michael Tiemann entered free software through his extraordinary work on the GNU compiler, but his lasting contribution may be his business brain. He discusses Red Hat, revenue and The Art Of War with Linux Format.

There are few Linux users who won't have heard of Cygnus Solutions. It's a company famous for maintaining many key pieces of GNU software, as well as making it relatively easy for Microsoft Windows users to harness some Unix power though Cygnus Solutions' Cygwin. Michael Tiemann was one of the Cygnus co-founders, taking his cue from the GNU Manifesto which he believes is really just a business plan in disguise. When Cygnus Solutions merged with Red Hat in 1999, Tiemann became Red Hat's chief technical officer. This was a role that forged his commitment to open source development, eventually leading to his current position as Red Hat's vice president for open source affairs. Graham Morrison recently had the opportunity to talk to Michael, coaxing some brain-repository memoirs and meanderings from one of the most interesting, and influential, personalities in the world of free software.

Linux Format: What does vice president for open source affairs involve?

Michael Tiemann: What that means is that I talk with private and public sector policy makers about open source technology and strategy. So, I'm getting ready for a trip to Sri Lanka and India, where we're going to be talking with folks in both governments and also the private sector about their increasing awareness and adoption of open source. I was in Tokyo and Osaka in June, talking with people who are looking at educational policies, government policies, procurement policies and of course their own IT infrastructures.

LXF: So you're the perfect person to talk to if you want to know how to make open source work?

MT: If you want to understand, I think, what open source means at a more strategic level. As you may know I started the world's first open source company, actually the first free software company, back in 1989 ­

LXF: Cygnus Solutions.

MT: Yes. So I have been doing it for a while. I have seen a lot of successes, some failures, a lot of rationalisations and realisations, and so my best is to connect the dots and keep those dots connected.

LXF: What is it that led you to form Cygnus?

MT: I started hacking on the GNU C++ compiler in 1987. Actually, in 1987 I starting hacking on the GNU C compiler ­ there was no C++ compiler at the time. What led me to that was that I was always interested in compilers. When I was in school I thought it was just magical that you could take a high-level language and turn it into something that a machine could execute. What I then did was... I thought it was just a miracle that I could get the source code to a compiler, you know, for free!

LXF: So that was the main reason why you entered free software?

MT: Yes, that was the big boggle. You know, what if Keith Richards could walk into a guitar shop and just take a guitar off the rack when he was a boy? That would be the same kind of magic. To me, this was just unbelievable.

The very first thing I thought I would do was, I was working on a computer that had a chip, a National Semiconductor 32032, which the GNU C compiler did not support. It supported a Sun workstation ­ I didn't have a Sun workstation. It supported a VAX ­ I didn't have a VAX. But it supported this other thing and I thought to myself, "Why don't I try to port this compiler so that it can be run as a native compiler on this machine that I have access to?"

At the time, the business of doing a compiler port was run by a small number of very esoteric companies, and they would charge, you know, millions of dollars and it would take them years to do this effort. I sat down and two weeks later sent a message out to the mailing list saying, "I've done it, and it generates code that's 20% faster than the native compiler that came with the machine."

LXF: What were you doing right?

MT: [Laughs] You know, these compiler companies at the time, they were small shops. I don't know ­ I really literally don't know what made it take such a long time or cost so much money. But I just said to myself, "Hey, if free market economics is a valid theory, then I've got a better mouse trap" right? "If I can do something that's this much faster and this much better and do it so much more easily, there ought to be a business there."

I didn't want to be the businessperson, I wanted to be the guy doing the hacking. So I spent two years trying to talk people who would be entreprenbecause they couldn't conceive how you could make money with free software.

LXF: Obviously you were very skilled at doing what you did. You could easily have created a company doing what the other guys did in half the time, and kept it proprietary ­ so why didn't you?

MT: But look at the two parts. Part one was, in two weeks I did something that would have eurial into going into business with me. Nobody wanted to do it taken two years. Part two was, I got helpful suggestions from people that made my effort twice as good as the original. Why would I want to cut that out? At the time I was, and still am today, far more interested in what is possible with a capital P than just doing something better than the next guy and taking their money off the table.

LXF: What happened after the compiler?

MT: After two years, the blinding insight that I had was that if everybody says it is a great idea and nobody thinks it will work, I would have no competition. And I figured that the technical advantages on the one hand and the lack of competition on the other made it foolproof, and that's what inspired me to start Cygnus. I found a couple of other people ­

LXF: You did get some support, then?

MT: I found two other people with that insight who were willing to give it a go. One was John Gilmore and one was David Henkel-Wallace. John's claim to fame before Cygnus was that he was employee number five at Sun. He did the heavy lifting of porting a lot of BSD... Bill Joy [BSD author and Sun co-founder] obviously did his fancy stuff at Berkeley, but John was one of the people who did the job of getting it ported to the Sun platform.

LXF: How did you manage to convince them that there was a business model around this?

MT: I went to companies who used compilers and basically said, "Here's what we can offer, do you see a benefit to this?" One of the first companies to see a really big benefit was Intel. Steve McGeady was the project lead for the i960 microprocessor, which was a very important microprocessor in Intel's telephony business ­ their clients included companies like Nortel, who built big phone-switch and data-switch products. And McGeady said, "I want a great compiler; I want to be able to give developers a compiler that they won't have to pay any money for. If they want to pay you for support services you can go and convince them, but I want to make this chip popular by lowering the cost of developer tools."So we said, "Great: you pay once, and you can distribute many, and if any of those ongoing customers want to do business with us they can."

LXF: So Intel paid for the initial hardware ­

MT: Correct: they paid what we called non-recurring engineering (NRE) and their cost to us was probably not only one-tenth of what proprietary vendors would have cost, but probably about one quarter of what it would have cost them to do internally. We gave them an extremely cost-effective deal, but that wasn't the reason why they were so excited. They were so excited because they'd actually changed the dynamics of the market, and the result was that Intel built a very successful embedded business. And other companies came to us: NEC, Hitachi, Fujitsu, Motorola ­ gosh, I can't even think of all the names, but all these guys who were in the embedded business were very keen to offer very competitive software development tools.

As a consequence of this, the GNU compiler toolchain became a kind of standard in the embedded business. We literally went from being a niche, strange business model to being the defining technology of the embedded software industry.

LXF: How big was Cygnus around that time?

MT: We built the company up to over $20 million in revenues. When Red Hat bought Cygnus [in early 2000] we were about the same number of employees and the same amount of revenue as Red Hat. They had a successful IPO. The bad news for us is that the financial market values operating systems far more highly than they value compilers. It's like beach front versus farmland!

LXF: I'd always read it as a merger.

MT: Well, you can read it... Literally, we did a stock exchange and at the end of the day we all owned Red Hat stock. But the Cygnus business was a very strong business, a lot of people from Cygnus are still at Red Hat, we still have major contributors to the GNU compiler working at Red Hat and doing good stuff. There was one other point I wanted to make that has slipped my mind...

LXF: How did the Windows part of it come about at Cygnus?

MT: The Windows platform became a very important platform for embedded developers. They loved GNU tools but they wanted to use low-cost PCs. The thought of spending $20,000 on a proprietary workstation ­

LXF: This is before Linux?

MT: Oh, that was my point, thank you! My point was that you ask yourself, "Michael, why were you so stupid that you did a compiler business instead of an OS business?" And the answer is that AT&T was busy suing Berkeley [Berkeley Software Design, over BSD], so that didn't look like very safe legal ground, and it would be two years before Linus Torvalds sent out his famous "I'm building an operating system" message. It did not occur to us that we could build revenues around a non-existent suite of technology! GNU existed ­ well, the GNU kernel did not exist but by golly the GNU compilers did, so we built out business around that first technology.

LXF: You weren't tempted to work on Hurd?

MT: Not really. Actually there's a famous expression in America: `The pioneers get the arrows and the settlers get the corn' That phrase `pioneer' is intentional in my bio [Red Hat describes Tiemann as "a true open source software pioneer"]!

LXF: So, the Windows version...

MT: Yes, what ended up happening was, you looked at the cost of a PC, and also the Intel processor had really begun to overwhelm the performance capabilities of the Unix processor. So you'd look at these guys who had these Windows platforms, and they were a lot cheaper and faster than these expensive Unix workstations. One of our developers one day just announced: "To hell with it. I'm going to build a Unix emulation environment so that we can run GNU tools on a Windows platform." He created the Cygwin library, which remains to this day. Through the entire history of Cygnus, by a factor of ten more people have downloaded Cygwin than any other single thing from the company.

What it goes to show is that the Windows market was a large market, and even today, in spite of all our success at Red Hat, Windows remains a large market. But I've been running Linux on my desktop since joining Red Hat and I'm fascinated and delighted by all the good stuff that is available there. I don't feel the need to run Windows for any reason.

LXF: How did the merger with Red Hat come about?

MT: One way it came about was that Cygnus was this incredibly valuable property. We were the toolchain that basically defined what was possible to build from an application perspective. An OS without applications is not a very interesting environment. We began looking at how to... Cygnus was the first company to take venture capital, in 1997: we closed in January of that year with Greylock, which later invested in Red Hat, and August Capital, which was a kind of spin-out from Kleiner Perkins. These venture people were beginning to consider what was the exit strategy, and you look at doing an IPO of a compiler ­ naah, not going to happen. But a potential merger was interesting, so we began talking to various Linux distributions. Things precipitated very quickly. We had a lot of interest from Red Hat and we liked them best.

LXF: How have you seen Red Hat change?

MT: Red Hat has migrated its business model from a focus on retail ­ you know, putting a box on a shelf ­ to enterprise supplier. We have defined the business model for open source. Obviously we don't claim to have the only answer, but there's no company I'm aware of that has demonstrated a hundred million dollar ­ let alone a multi-hundred million dollar ­ revenue stream based exclusively on open source software and services. The way that Red Hat has obviously changed is that we have changed our release model, our business focus and the clientele with whom we work. Some people have not yet forgiven us for taking away the choice of buying a $79 product in the remainder bin at Fry's Electronics.

LXF: It's obviously been a beneficial decision for Red Hat with the benefit of hindsight, but at the time what was the motivation for dumping the boxed versions of Red Hat Linux and entering the enterprise market?

MT: If you read books on business, if you read books by Michael Porter on competitive strategy, they'll say you can fundamentally pursue one of three competitive strategies. You can go low-cost; high-value; or a niche, focused product. He talks about something called the strategic dilemma: companies that try to do more than one inevitably fail because the attempt to do one undermines the other. Albert Einstein paraphrased this dilemma when he said you cannot simultaneously prepare for war ­

LXF: ­ and broker peace.

MT: Exactly right. Everything about the boxed product defeated any reasonable attempt to do business with the enterprise. For example, if you want to sell boxed products, you want to sell a lot of releases. You don't want to sell something that's going to sit on the shelf for seven years, you want to sell something where every six months there's something new. And if you looked at Red Hat's revenues when it was in the boxed distro business, it was all about, "Come out with this new release." "Why?" "We need more revenue." You go to the enterprise, and the last thing they want is gratuitous updates. They want to be able to choose which technologies on which time frames, and they highly value a platform that remains stable over a long period of time.

LXF: It is surprising that nobody else has made the jump in quite the same way.

MT: One of the things that I've learned throughout my almost 20 years of selling open source is that you don't win with "me too" Sun Tzu in The Art Of War says that if you want to attack a defended position you want at least four times the strength of your opponent and preferably ten times the strength.

Go back to my GNU C compiler story: who would do business with a fresh-faced 23-year-old? Somebody who needs that factor of ten. If I can deliver four times the performance at one-tenth the price, that will get attention that 20% faster and 20% less cost never will. You don't always get the super-size result, but generally speaking, if a sales guy brought me in to an account to try to convince somebody why a 10% saving is going to save their business, I'd take him outside and say, "I'm never doing this again."

LXF: That's tough.

MT: The beauty is that the market gave us that as an opening, so we established our competitive position. Nobody else has yet figured out how to do four times the performance at one-tenth the cost of Red Hat. LXF