Brian Behlendorf - interview

From LXF Wiki


Apache Angel

A lot of us take Apache's dominance in the server room for granted. Brian Behlendorf tells Linux Format how it came to pass.

Whether he's collaborating on the Burning Man music festival deep in the Black Rock Desert, Nevada, or working on web design and version control software back in San Francisco, Brian Behlendorf is an innovator and a true survivor of the fin de siècle craziness that ruined so many internet entrepreneurs at the end of the nineties. Thanks to his belief in an open community and an open development model, he's been a success with one of the first web design companies, Organic Online, and then with CollabNet, founded in 1999 to sell open source solutions for big business.

Graham Morrison had a chance to talk to Brian about how Apache emerged from the NCSA primordial soup and what role Linux will play in future software development.

Linux Format: Could you tell us what your role is at CollabNet?

Brian Behlendorf: CTO. I don't have quite the title `founder' on the [business] cards.

LXF: Oh, and you're listed second on CollabNet's website.

BB: Sure, well actually, Bill Portelli is our CEO. I found him three months after starting the company, so really he's a co-founder. I was really happy to have found somebody who had the business sense, who could hire the sales team, who could do all the boring bits of building a company like CollabNet ­ and allow me to focus on figuring out what it is that we're trying to sell as a business and how we communicate that to the rest of the world, as well as making sure the product we build is in line with that.

LXF: So, where did it all start?

BB: Rewinding the clock just a little bit, in the early days of Apache, it got started by a bunch of us who were building websites and we just needed [a server we had more control over]. I pretty quickly handed off development to other people who were better at it.

LXF: That seems to be what every developer says!

BB: [Laughs] Perhaps there's this debilitating humility in open source communities. I extended the focus more on the back-end systems, being the sysadmin to the stars ­ making sure that the mail flowed really well, that the CVS trees were up and happy and not corrupted. The bug tracking system was up and I was creating projects, adding accounts and tying these things together with baling wire, Scotch Tape and shell scripts here and there ­ generally doing a pretty hacky job on it, because you never wanted to spend very much time on it.

It was always something I did in my spare time, because my day job before CollabNet was at a website design company called Organic [Online], and Apache has never had a full-time employee; it's always been volunteer contributions. And so I was the infrastructure team, I was the one who, when a machine went down, would drive down to the facility, press the reboot button and make it work.

In doing that I was just getting more and more frustrated because it felt like, a team and I could sit down and write something better from scratch. Or not even so much entirely from scratch, but just re-integrate these pieces using a real system for tracking users, permissions and projects, and bringing these tools together in a more tightly integrated way. That would be kind of cool and help Apache grow; but it would also make it easier for companies to start open source projects.

LXF: Thus CollabNet was born?

BB: I was getting bored with Organic around 1998 and looking for something new. I was commiserating about this with Tim O'Reilly, and he said, "Why don't you come and join us as an entrepreneur in residence?" This being '98, '99 ­ the go-go days in Silicon Valley ­ every problem was a business opportunity worth putting a business plan around. In fact, if you weren't doing that you weren't cool.

We decided to float a collection of different ideas, this (CollabNet) being one of them. That idea was to put these tools together and provide it as a service to companies launching open source projects. One of the first conversations we had was with Sun, who wanted to build brand-new Apache-style communities around first NetBeans, and then Those are projects that we launched in '99 and which continue to be very active today. If you go to OOo, the tools behind that community ­ for CVS, bug tracking, for publishing of content, for coordinating releases and things like that ­ those are tools that are CollabNet. Sun pays us a fee, and we run it as a public site.

LXF: With CollabNet today, how do you convince the unitiated that open source is viable?

BB: One thing that I try to point out to people is that 1 years ago, those of us who were using the NCSA web server, were using it not for hobbyist purposes. Some of us were, but most of us were using that server to start businesses. Whether it was a website design company like I had, or the Internet Movie Database, or the web master for MIT... we all were using this for business reasons, and we all got together to write code.

LXF: How did you see the potential of open source so young?

BB: I guess I didn't know any better. I was a 20-year-old near-college drop-out from UC Berkeley, and I was spending a lot of spare time discovering this cool thing called the Net. It fascinated me how a system like this could work ­ it didn't seem to have any centre point. I'd been a user on Prodigy [the ISP], and here was something where there was no gatekeeper, no central coordinator, but it still worked. I just fell overboard, staying up all night reading things that I could find. I kept seeing references to these IETF working groups. So, I joined a couple, notably the HTTP working group, and I got to sit there and watch Tim Berners-Lee and Martin Gleeson debate how Mosaic should talk to the CERN web server. Or on the HTML working group, whether the image tag should be a container tag or not.

I guess I just didn't know any better but it seemed to me that design was something you needed to do as a conversation rather than as one person delivering the ten commandments off a mountain. That was a fairly strong impression, and in the early days of Apache, there were lot of us who were heavily involved in the HTTP working group. We had these businesses to run, we wanted a better web server, we were worried about Netscape being a brain-drain fo all the people from NCSA, and we combined our efforts and thought, "Hey, let's see if we can do it better. Maybe someday we'll have to buy the commercial one, but for now this is better than nothing."

But I think there was an idealistic aspect to it too. Those of us in Apache were pretty eager to see a reference implementation of the HTTP standard, that could also be considered production quality. We were worried that some single company would come to own the client side and the server side of the web. And if they did that, it would be essentially no different from Microsoft's ownership of the desktop. It wasn't motivated by an anti-company FUD, it was motivated out of concern that we liked this thing called the net that had no gateway, no chokehold, and we wanted to keep it that way. We didn't know we would get 70% market share...

LXF: Is that thinking how you chose which licence to use?

BB: We had this discussion early on. The original [Apache] licence on the NCSA code base was incredibly liberal it said: "Here's this code, do whatever you want with it. You can create derivative works, you can use it for any reason ­ just if you pass it along to somebody else, don't blame us if it breaks. But do let other people know that what you're handing off came originally from NCSA."

We thought that that was a pretty fair licence. And so the Apache License was basically that same concept. Same thing with BSD. Even in those early days, we said: "What if a company picked up Apache and incorporated into some commercial thing?" But then we said: "So what? In fact, that would be a good thing, because that would help enforce HTTP as an open standard. Using Apache, they wouldn't have any excuse not to conform to the standard." They couldn't say it was too difficult, which is what most companies' excuse was. Furthermore, we believed in the carrot rather than the stick. The GPL said, "You must collaborate, you must share your code." The BSD licence says, "Do whatever you want", but we felt that companies would eventually share their patches back with us. Or even their improvements back with us.

LXF: What impact do you think GPL version 3 will have on open source software?

BB: I've always been more of a fan of Berkeley-style licences than GPL-style licences, because of this carrot versus the stick. I worry that any language that expands the extent to which you should release source code only makes it more difficult to bring other participants in to the open source world. I have a lot of sympathy for [Richard] Stallman's position. I think that in the very long term, the 30 to 50-year term where software is inside every doorknob and we start programming life along biological means, having access to source code is a lot like having access to civil rights. And so in the very long term, I think he's right on the ball.

In the short term, I want to bring over as many people as I can and convince them of the advantages of sharing code. I prefer licences that stay simple. The solution proposed [for GPL v3] that if an application has a link that says `download the code', that link has to be preserved ­ I think that's OK. But I think that it will still come down to what it means when I deploy a GPL application here and I need to access another application over there. Do I have to distribute the whole piece, or just the original GPL piece?

Companies try to understand how much they need to give up, and are staying away from GPLed software because it's so vague. I think that's kept more participation from happening. I hope GPL v3 fixes that.

LXF: You weren't worried that companies might simply fork Apache into their own standards?

BB: Well, if they did that, so what? Even if Netscape, or even Microsoft picked up that code base, we felt it would be OK, because having an open development process and an open source licence meant that the majority of activity would come back to us. With the scale of the internet, we didn't need that many people to come back to help improve it.

The other thing we did was figure out: "What is this collection of processes, attitudes, tools that helped us build that web server?" We replicated that to other software projects, and we started the Apache Software Foundation in '98. Now there are 25 different top-level projects. A lot of this gets back to CollabNet. What I saw in '98 was this need for better tools for the companies to understand, and this need for an understanding of the development methodology, or a bias towards openness, being transparent ­ these kinds of things.

LXF: What do you think about where the web is going?

BB: There are a lot of interesting apps being built right now, and a lot of focus on wiring different apps together. I really like the apartment finder thing, wiring Craigslist with Google Maps. That's interesting. It's the recognition that the value of a database isn't created by one person in a Soviet-style command economy approach, but is instead dynamically generated by comparing lots and lots of other people's activities into a much more interesting combination. One of the biggest values of Amazon is its reader comments; the biggest value of Wikipedia is the fact that's it's built by a huge number of different people. It's not Jimmy Wales's [Wikipedia founder] views on each one of the subjects; it's a grass-roots aggregation of thousands of viewpoints. I think the most interesting apps to come are these aggregations of bottom-up data. We haven't seen that.

LXF: Have you got any good ideas?

BB: [Laughs ­ we take that a,s "Maybe, but I'm not telling."

LXF: Do you think proprietary web products like Flash are acceptable?

BB: I'm biased against it because it's taken them forever to get a plugin that works for Linux. I'm further afield because I run FreeBSD, not Linux. Without source code to that, it's been difficult for anybody to try to fix hangs and things like that. And it's still the case that maybe one out of 20 Flash animations that I see will cause some sort of hang, or stutter in some way.

It's not a problem with Macromedia per se. As a programming environment I think we're seeing Ajax pick up as just a more natural interface. What's nice is that Flash animations have a very fixed window that they operate in, whereas Ajax animations (like HTML) seem to do a better job of fitting into the window you provide, which makes it a lot more flexible.

LXF: It's taking a long time for an open source solution...

BB: Well, it's the reverse engineering. Even if it's possible, you run the risk of incurring the wrath of the DMCA. Ajax, on the other hand, is built using all open technologies, and when you have two equivalent systems, the open one usually wins.

From a broader picture, it's not like we're going to see the end of proprietary software. Like Apple, we sell proprietary software that people love, but we always have to move upstream. You never want to be what Netscape ultimately became, which was selling web servers against Apache. Have a business model that's based on maybe a combination of services, but also on something above and beyond what people can get on their own. It will be interesting to see where the software industry as a whole goes in response to that.

LXF: Do you think patents will become a bigger problem?

BB: The funniest thing is that, over the last year, which company was the one most hurt by patents?

LXF: Erm... probably Microsoft?

BB: Just a $500 million Eolas patent [the US Appeals Court has since partially overturned the verdict and ruled for a retrial]. Who was the second biggest company? RIM BlackBerry [RIM's out-of-court settlement to NTP Inc was actually worth more than the Eolas verdict].

I'm really happy that there are organisations like Free Software Foundation Europe and others that have been fighting against the encroachment of a stronger patent law, and we need to do more of that in the States as well. I think the biggest champions for that kind of fight are the existing large IP patent holders, who will realise that we've created this fool's game if it's too strongly enforced.

I'm not against the idea that in some circumstances patents make sense. Sometimes there is a really bright idea that did take a lot of trial and error to arrive at, and the inventor ­ not necessarily the company that paid their salary ­ should be rewarded for some of that. You need that to protect some degree of R&D investment. But the most innovative things in software over the last ten years have not been patented technologies. They've not come from sitting in a lab toiling away for years, [they've] come from communities of dynamic development.

LXF: Do you think there will be an increasing division between US and European software development?

BB: The thing that interests me is where is IP going to go in India [CollabNet has an office in Chennai]. When I was there last week, I read this article on a company called Cipla, who announced that they were about to start production of millions of doses of Tamiflu, the flu drug. Roche Pharmaceuticals, who owns the patent on Tamiflu, can only make a certain number each year, and the delivery time if you ordered today from Roche is 18 months. Cipla said this is a matter of national health, a pandemic, so they started making it. [Roche markets the drug worldwide but does not hold the Tamiflu patent in India.]

So, it will be interesting to see what the Indian government does about that, because the Indian government did not get their flu vaccine orders in, and they've got a billion people to worry about and Roche can make only 12 million doses a year. For some places patent law won't change gradually. There are some places so wedded to the idea that stronger patent laws are always better that it will take something like the equivalent of a patent Pearl Harbor to cause a sea change in that perception. I don't want it to be the case, but it could be that the issue around Tamiflu becomes that situation where people realise that strong patent laws cost lives.

LXF: Finally, are you still involved in electronic music?

BB: I do the John Peel thing where I'll play at parties, for friends at events ranging from 20 people to... well the largest in the last year was about 600. They're underground parties where we'll camp out, haul a big sound system out into the mountains or a beach somewhere, and play music. LXF