Marc Andreessen, Act II

At Netscape, I was paranoid about second-system syndrome from the very beginning. I read the book many times. So we managed to avoid it on Version 2 of the Netscape browser. We avoided it on Version 3. We avoided it on Version 4. And then...

Where's Version 5?

Exactly. We started Version 5.10, which is now 6.0, in early 1996. Here we are, almost five years later, and it is only now coming to market. The engineers kept saying, "Oh my God, we've got to re-architect. We've got to rewrite the LAN engine. We've got to rewrite the whole thing in Java. We've got to do this, we've got to that." I didn't understand half the stuff, and past a certain point you just say, "Well, all right.'' So despite our best efforts, we didn't totally avoid second-system syndrome. We just deferred it to the fifth version.

You did something interesting and provocative at Netscape. You gave away the basic Internet browser on the belief that you could collect other revenue down the road. When does that business model work? When is there no "down the road"?

Well, Netscape was designed to be a full-range software company with servers, commerce applications, and all sorts of technical support and consulting services built around the basic Internet browser. So we knew from the very beginning that we could give away one product and then charge for all the other software. It's kind of a razor-razor blade model. You can see something similar in the cellular industry, or even with AOL, if you think about the true acquisition cost for each new customer.

But if you start thinking, "I've got lots of eyeballs, and I'm going to monetize some of that traffic," well, that's the kiss of death. At that point, you're trying to construct a revenue stream for the sole purpose of keeping you going as an enterprise, as opposed to addressing an actual problem that a customer might have.

For the past few years, you've argued that the Internet will burst open what might be called the "cathedrals of knowledge." No longer will big, powerful enterprises decide on their own what software, music, or news the public wants -- and then create it in secret before rolling it out in full. Suddenly, we'll have all of these choices, and we'll all get to help make the product. Is that coming to pass?

If you look at the interplay between open systems, where everyone gets to see the source code and refine it, and propietary software, there's definitely room for more of both. There's a lot more adoption of linux [an open operating system] now than there was a few years ago, but there's also a lot more adoption of Sun Microsystems's Solaris. There's a lot more adoption of open-source application servers, but there's also a lot more adoption of WebLogic from BEA Systems Inc. For a certain class of customers, there's no substitute for a large mainstream company backing them up. In part, that's because buying behavior is not always as logical and advanced as the open-source advocates want it to be. It's often based on risk aversion. This goes especially for highly complex things.

That said, a top-notch technical team working with an open-source environment is able to move faster, change faster, fix bugs faster, and expand faster. That's why a lot of the really advanced government agencies -- NASA and so forth -- tend to be pretty aggressive users of things like Linux. They actually have a lot of technical talent in those organizations. So they would much rather just fix the damn bug than wait for X Corp. to go through its three-month patch process, whether that's Oracle, Microsoft, or whomever. But there are many other organizations, such as your local insurance company, that just never will make that kind of decision.

In a way, almost everything we've discussed so far highlights the intensely social dimension of the Internet. That wasn't the case with early generations of information technology. Back in the 1960s and 1970s, computing was a nonsocial, almost lonely experience. Then we started networking everyone together. When did you first sense that the mere act of connecting everyone had huge implications?

When I got to the University of Illinois in 1989, what struck me the most was that people at the supercomputing center were already sharing computers, downloading images, and doing distributed software development. It was part of our lives. And yet everyone else at the university had never heard of this. That's because the system was set up to be almost deliberately hostile and foreign. So it wasn't much of a leap to say, "Well, the barriers to easy use aren't going to last forever. If this is useful for a physics researcher, it's going to be useful for a grandma to communicate with her grandkids."