The JBoss Group LLC is best-known as the provider of the JBoss open source application server, and it has also been known for having disagreements with Sun about licensing of Java test suites. But those disagreements were resolved in 2003. InfoWorld Editor at Large Paul Krill met with JBoss CEO and Founder Marc Fleury during the "J2EE 1.4 Kickoff Event" earlier this week in San Francisco to discuss the company and open source issues.
InfoWorld: JBoss was considered somewhat of a renegade in the Java world until recently. How did the reconciliation with Sun come about?
Fleury: We finally reached an agreement on the specification licensing of J2EE 1.4. The big innovation there was that for the first time an open source implementer could license it, so we have. Right now, we're writing the EJB (Enterprise JavaBeans) 3.0 spec with Sun and bringing back a lot of the innovation that we've been pioneering in the open source community and the Java community. We're bringing that back into the standards.
InfoWorld: When is EJB 3.0 due out and what's that going to feature?
Fleury: The 3.0 draft is out now (but it is) still private. We'll announce at JavaOne (in late-June), preannounce at the ServerSide conference (next week). And what we deal with is simply finding the programming model. And so whereas today it's fairly complex for a developer to code an EJB, we go with a plain old Java object model, we use the query language that's an extension of EJBQL, very much like Hibernate. And that's it, so (we're) simply finding the programming model, enhancing the (query) language and persistence capabilities and at the end of the day. It will be very transparent, plain old Java object-based development that will greatly simplify EJB development. We at JBoss hope to have a prototype by JavaOne.
InfoWorld: You mentioned today you didn't see much benefit from making Java open source. Isn't that kind of ironic, considering you are with an open source company?
Fleury: I sat through and (asked myself), 'What's the goodness of open sourcing Java?' And there would be marginal gains that we could have such as speed. Granted, you know there are some bugs in the virtual machine class loaders that have dug JBoss for a while and if we could fix that, that would be good, but really we offset that marginal gain with what I consider a big risk. The success of Java has been its ironclad portability across operating systems. The fact that the Sun engineers were capable of building the best Java virtual machine on Windows, which is an operating system they don't have the source code to, speaks volumes to their commitment to Java as a portable platform and so it's also thinking about how we do open source at JBoss. There is an advantage to having very large communities, but at the end of the day you have to mix this community with one responsible person that's in charge and makes the final call and these are very efficient equilibriums in my mind. We do it at JBoss, meaning we have large communities of partners that contribute. By the end of the day, JBoss is responsible, stands up, has contracts, does business, and it's a good thing to have this central figure, benevolent dictator image that that's used over and over. And I think Sun has done a tremendous job in that respect as the benevolent dictator. People are saying, 'Well, we're tired of the dictatorship.' But really what good would that do? The worst that can happen is Sun loses control of Java and I'm not sure that would be a good thing for Java.
InfoWorld: Why not? Why wouldn't it be good for Sun to lose control of Java?
Fleury: It wouldn't be good for anybody, I think, because, again, the portability is what has defined our market. Sun has committed to that, Sun has implemented Solaris, Linux, Windows and they have an excellent track record. Why change something that works today, for what gain and at what risk? I'm not even fighting for the J2EE specs to be open source. I'm quite content with the JCP (Java Community Process) as it stands today
InfoWorld: You mentioned today an interest in expanding into other product areas such as open source ESBs (enterprise service buses), workflow engines, rules engines. Why that direction for JBoss?
Fleury: We started as an app server, but (in) the model of professional open source, (the vendor) does not need to stop at the app server. We're letting our users drive this. So for example, we have a user who calls up about six months ago, saying, 'We're going to buy JBoss support from you. We use a lot of Tomcat, where can I buy my Tomcat support?' And of course it's tempting to say, 'Well, right here. Give me your money and I'll support this, it's open source.' After all, 1,000 consultancies out there sell support every day in the name of JBoss when they really have no credibility to do that. So what we did is we went out and we recruited Remy Maucherat, the lead developer for Tomcat and now we can say, 'Look, this is professional open source and as far as we're concerned, Tomcat is federated as part of JBoss,' even though that of course ruffles feathers on the Apache side, (since) it is still an Apache project. We're the largest contributor today to Tomcat. We can expand that model to other revenue streams and then from a purely financial point of view, then our gain may just be a stream aggregation.
InfoWorld: How does JBoss make money, specifically?
Fleury: We have purely a support business model, very much like Red Hat. We've successfully proved the subscription model has been a very scalable one and to be honest, the play in open source is the same as the traditional enterprise software vendors. It's always going to be the same usual suspects that buy support, right? The large companies want a support relationship in place as they go to production, they want an insurance as they go to production, and it makes total sense, so our play is really to give our technology the widest reach, mass adoption, and open source helps us distribute. We have 5 million downloads of JBoss in the past two to three years. And then out of those 5 million downloads you're maybe going to have one in 10,000, one in 20,000 -- it's always going to be the same guys: the travel industry, Wall Street, (etc.). They're going to buy from us and we'll be there to support our software just like a proprietary vendor does. There's no difference there, right? In fact, our contracts are industry-standard. There's nothing very particular about our contracts on maintenance. They're absolutely plain-vanilla industry-standard.
InfoWorld: You don't also sell your product on a dual-license format?
Fleury: We don't do dual license today.
InfoWorld: But you'd like to with the ESB?
Fleury: We don't want to do it and we can't do it on the core JBoss product itself, but on new layers it may make sense. It's still being debated inside the company. Some are for, some don't like it. I feel that there are two known business models, dual-license and support, and both are clean, open source plays. We're very close with MySQL (which offers a dual-license format).
InfoWorld: Why can't you do a dual model with JBoss?
Fleury: Well, first of all we don't want to kill the goose that laid the golden egg. I think that's the bottom line here and the golden egg in our case is the No. 1 market share in ISV/OEM (for Java application servers), beating BEA and IBM by a large margin.
InfoWorld: Is JBoss profitable right now?
Fleury: We were always profitable from the get-go.
InfoWorld: Recently, I attended an industry event where a Microsoft official basically said he didn't see how you could have a software industry if there's open source because if they're giving this stuff away, how do you really compete with that? How would you respond to that?
Fleury: Well, I think we're in agreement on that there's no free lunch. Well, let me step back. So 20 years ago, Bill Gates was working with (Richard) Stallman, the founder of the Free Software Foundation. Stallman invented free software and open source 20 years, 30 years ago. And back then he and a young Gates were going at it through to public forums at the time and Gates was arguing correctly, in my mind, that, 'Look, this free software thing doesn't make sense because writing software, marketing software, distributing software is a very expensive proposition, therefore you need licenses to fund all that.' And I think he was correct by and large. What has changed now is very simple. It's called the Internet. With the Internet we're capable -- a band of punks, daft punks mind you -- were capable of putting together a world-class application server. More importantly, we're capable of distributing it at no cost, at least no cost to us. We're hosted on SourceForge. We're capable of marketing at very low costs. I had zero marketing dollars, you know, when we beat Sun in the number of downloads and we beat the reference implementation which very we heavily marketed. Bottom line, when we're looking at an optimized business model where it's the same components, we still fund the procreation, there's no free lunch there, but it's a lot cheaper and therefore we can survive. Do we need to make money at this? Absolutely, I don't believe in non-profit open source. I want to see open source, just like Microsoft wanted to see software, as a standalone category.
InfoWorld: But if you're giving the software away, how do you make money other than by support? This Microsoft official's contention was that if you're going to try to make money on support, you farm it out to China, since American-based companies couldn't compete on a support model.
Fleury: Yes, that's a risk. Now, it's a good point. The truth is, we'll see. (But) there are in fact very strong (barriers to) entry to supporting JBoss or Linux for that matter. Let me explain that: It's called knowledge. So granted you can call some clown and in fact we compete with a lot of (them). But does the guy have any knowledge about JBoss? No. What's his credibility in doing that support? Little. So what we found out is -- and why we're successful now in federating all these little consultancies is -- yes, they're good to give first-line support. So they do customer hand-holding and there you may want to go to India or China if it's cheaper, but in the back end, who's the vendor that's really responsible for this code base? Who has the critical mass of knowledge about that code base? And today that's JBoss Inc., and it's going to remain that way with a brand barrier to entry and a critical mass barrier to entry.
InfoWorld: Which means you would be able to sustain JBoss on a service and support model?
Fleury: Yes, yes, and so the way we work with these firms, and we do have partnerships with some Indian firms and across Europe and the U.S., Unisys for example is a support partner. So what Unisys does is first-line, an implementation for large government bodies, but they boast a partnership with the open source vendor so that they can articulate a very easy sales proposition which is, 'Mr. Government, you want to buy JBoss because it's cheap, because you're not going to be hooked on the licenses. It's a friendly relationship you're going to have based on pure support and I'm going to be your frontline. Now be assured that if there's a real problem, I got the JBoss guys on the hook with clear contracts and escalation path for bugs and (etc.).' It's a very simple, reassuring message. We have the brand and we have control. We have the critical mass of knowledge about a code base within JBoss.