November 2006


There’s a pretty good discussion going on here at Dana’s ZDNet blog regarding Fleury’s comments.

The general view is that:

  1. Fleury is not a traditional manager, but rather a programmer at heart, so give him some slack when he speaks his mind about Red Hat management.
  2. JBoss and Red Hat have two very different development methodologies & communities.

Let’s discuss the 2nd one further. While both Linux & JBoss are licensed under the GPL, the major difference is the community surrounding both projects.

If we look at the Linux community, we find several competing vendors working together to advance the various Linux-related projects. By doing so, each vendor (who pays for employees to work full-time on these projects) is sharing the cost of building a solid product. The net result is like everyone contributing $5 to R&D, but getting a product that’s received $500 in R&D investments. Good for everyone, especially the vendors who are paying their employees to work on these projects. This model works because there is no central vendor who is exclusively benefiting from contributions from individual developers or 3rd party vendors. And when contributors assign copyrights to their works, they do so to a foundation, not a commercial entity, who may choose to license that work under a non-OSI-compliant license (i.e. a commercial license). Basically, there is a vibrant community around Linux because there is no commercially driven entity ruling over the project.

Unfortunately, very little of the above holds for the JBoss case. For the most part, JBoss products are developed by JBoss employees. There is no real community of individual developers or 3rd party companies working on JBoss products. And let’s be honest, to quote Tim O’Brien of O’Reilly: “It’s tough to build a real community when you have paid committers and unpaid contributors developing code under (L)GPL with the original copyright assigned to the corporate entity that funds the effort.” It worked in the case of Linux because the early work was done by individuals, so everybody was equal. Then, Linux attracted vendor attention, and most of the leading developers/community participants got hired by vendors seeking to advance Linux. And again, (almost) everybody was equal. The individual contributors that aren’t ‘equal’, continue to contribute because they get much more value out of Linux than the work they put in.

Dana suggests: “Changing the subject from what Fleury said, or what he said he wanted, to making the JBOSS community more open and vibrant may be the right way for RedHat to launch its discussion of this internally.”

But, that’s really tough to do at this stage in the game. Community is hard to build in the best of situations. It’s next to impossible when you want to maintain control. (Ahem…Sun)

Imagine that IBM had decided to launch the Eclipse project under IBM control and required that copyrights to contributions be assigned back to IBM. Where would Eclipse be today?

The fact that Fleury was expecting greater R&D investments after being acquired should make us think twice about the common belief that open source development is efficient because everyone benefits from the work of everyone else.  This is really only true if there is a community. A true open source project, shouldn’t be classified on license alone, but more importantly, the community contributions that the project is able to attract. A truly open community benefits from community effort. (Mostly) Everything else is commercial development done out in the open.

Sys-Con is quoting Marc Fleury, JBoss founder & CEO and now SVP & GM of JBoss Division at Red Hat, as having said:

“(while Red Hat) really haven’t gotten in the way”…“the R&D really hasn’t benefited from a huge investment, which I was hoping for and was the main reason I went to Red Hat”

[UPDATED 2006-11-27: Eweek has a longer article on the Fleury interview here]

Well, if Marc was looking for an R&D bump, he should have calculated how much more Red Hat spends on SG&A (sales, marketing & administrative) vs. R&D compared to Oracle or other large IT vendors. If Marc had done this, he’d have thought twice about Oracle. But maybe he’ll get to experience that also ;-)

While JBoss and Red Hat feel each other out and Marc goes to the Red Hat executive team with his tin cup for R&D, the WAS CE team has been making significant strides against JBoss. According to an Evans Developer Study of Eclipse users released in the summer, WAS CE usage with Eclipse grew from 0% in 2005 to 16%, compared to a growth of 13% to 19% for JBoss. That’s nearly 3x faster growth for WAS CE than for JBoss. Oh, and Apache Geronimo usage by Eclipse users grew from 1% use in 2005 to 14%, not bad eh? Combined, 30% of developers using Eclipse said they developed against Geronimo/WAS CE compared to 19% for JBoss.

Goes to show you what a true open community (Geronimo) and IBM’s R&D efforts (WAS CE) can team up to do together.

Matt believes that JBoss following a Fedora/RHEL model is good news for customers.  I’m not so sure :-)

What about customers and partners that don’t want to pay for support?  What happens to them?

This move is incredibly bad for partners because it’s going to force every ISV and SI that is considering JBoss to enter into a licensing deal with JBoss (i.e. pay for support etc.).  While this may be okay for Alfresco (where Matt is part of the executive team), I’ve spoken to literally hundreds of smaller ISV/SIs that sell to SMB customers and don’t have the margin room to get railroaded by this Red Hat/JBoss revenue grab.  These partners who are using or considering JBoss are surely going to look elsewhere.

Apache Geronimo & WAS CE are much better alternatives.  Customers & partners are free to use and distribute Geronimo or WAS CE without support.  Just like these forgotten customers and partners expect. Just like all customers and partners expect, CHOICE.

PS: based on the recent WAS CE Partner Initiative results, it seems that partners are totally getting it!

IBM announced last week that more than 600 partners have joined the WAS Community Edition (WAS CE) partner initiative that was launched 6 months ago. And that over 200 ISVs have already developed applications that support WAS CE.

This is pretty big news for WAS CE, and the Apache Geronimo community also (since WAS CE is built with Geronimo inside). Comparing the way that others are treating their partners, I can bet that the number of partners attracted to WAS CE & Geronimo is going to keep growing this fast. One possibility is that these recent WAS CE partner figures will scare Red Hat/JBoss into reconsidering their heavy-haned approach to partnering.

More details on the WAS CE Partner Initiative can be found here. It looks like some of the dates have passed, but I’ll find out if the WAS CE team is going to extend the program in light of the huge success over the past 6 months.

[UPDATE 2006-11-28:] I just found out from Bob Arfman, who was leading the WAS CE Partner Iniative, that the program has expired.  Partners interested in WAS CE can still join IBM PartnerWorld, but the special benefits associated with the WAS CE Partner Initiative are no longer being offered.  BUT…. he’s working on something a lot bigger that partners are going to dig.  Look for the new partner initiative in early 2007.

[SEE Update at the end of this post}

Computer Business Review Online is reporting that Red Hat CFO announced that Red Hat is working on a plan to create a Fedora-style community development version of JBoss as well as a subscription-only RHEL-style package.

A change in the package model is part of the company’s plans to convert the 11 million free JBoss users to subscription customers, Peters added. “We have an installed user base today that are natural customers when they get to the point of needing to have better support,” he said.

I take issue with two things in the above text. First, 11 million JBoss users, ROTFL. Second, if Red Hat is truly looking out for your customer and partner base, then there is no need for a Fedora & RHEL style development & sales model for JBoss. This move is really bad for any ISV/SI that uses JBoss without support and was foreseeable over a month ago when Red Hat changed the JBoss EULA. I’ll deal with that in a second post, but first:

ummm, note to Red Hat CFO Mr. Peters, I’ve downloaded Firefox at least 15 times in the last year (different computers, reinstalls, upgrades, fixes, etc). Just imagine how many times I will have downloaded Firefox in 6 years. That does not make me 15 users. Because if it does, then man could I get a lot more done in a day!

So believing that your user base is 11 million is definitely new-math (and since you’re CFO, I’d suggest you be a little careful with new-math). You could argue that there have been 11 million downloads of JBoss packages over the past nearly 6 year, but that does not equal users. Every one of your users would have downloaded at least 20-40s of times over the past nearly 6 years. And they would likely have downloaded/used other application servers at the same time.

Also, Sourceforge stats show that there have only been 8.6 million downloads for all JBoss packages on Sourceforge since JBoss packages were available on Sourceforge in 2001. So that means every single version, release and modification binary package, the associated source packages and MD5 files would be considered a separate download inside of that 8.6 million figure (aka “a customer” using new-math). I’d bet a dinner that the number of actual JBoss users is at least two orders of magnitude less than the download figure.

Anywho, good luck trying to force your actual user base, specifically the ISV/SI base into buying support. And here I was thinking that JBoss & Red Hat were in the business of giving your user/partner base choice. Well, lucky for them that Apache Geronimo, WAS CE, or JonAS (amongst others) are more than viable alternatives. And since most of your “customers” are really only using the Servlet functionality provided by Tomcat, they’ll be happy to know that Tomcat is also inside of, and supported within, Geronimo & WAS CE, so they can rest assured that there is a future-proof migration path when your sales colleagues come knocking.

Thanks for once again reminding your users and partners that using Tomcat, Apache Geronimo and WAS CE are much better alternatives if they actually value choice, flexibility, freedom and an open community.

PS: I’m going to see if I can get Geronimo and WAS CE download figures to compare with what we can see on Sourceforge for JBoss.

[UPDATE 2006-11-20]: A comment on this post points out that the “11 million users” quote was a mistake on the part of Computer Business Review Online. The article now states “11 million downloads”, which is much better.  Not sure where Red Hat gets 11 million from, the Sourceforge data shows less than that, but maybe Red Hat is counting another source also?

There’s been a lot written about the possible impact of moves by Oracle & Microsoft/Novell on the future of Red Hat.  Some have predicted the decline of Red Hat’s fortunes, and Oracle’s future acquisition of Red Hat.  I’ve recently heard that JBoss customers are beginning to wonder whether they’ll eventually become Oracle customers.  I say to them, don’t worry, you’ll always be able to move to Apache Geronimo or WAS Community Edition.  And actually, why wait for Oracle to acquire Red Hat ;-)

So, the news is official: (and yes, I agree with IBM’s question as to why Sun didn’t join one of the community projects focused on delivering open source implementations of Java SE, ME and EE instead of trying to build their own community.)

Sun has decided to use GPLv2 for their reference implementation of Java ME, and the GPLv2+classpath exception for the Sun reference implementations of Java SE and Java EE. Sun’s Java EE implementation (GlassFish) was previously under the CDDL, and the GPLv2+classpath exception license is additional (i.e. it does not replace the CDDL for GlassFish).

They have a very detailed FAQ here. Three very interesting points after reading the FAQ:

[1] Sun still requires contributors grant Sun joint ownership of any IP that they contribute to the Sun owned ‘communities’. (More here)

[2] Sun is still going to provide commercial versions of Java SE & Java ME implementations. And just for kicks, Sun will allow 3rd parties to make modifications to the Sun owned implementations of Java (or is that community owned? I’m confused) without having to put these modifications back into the open source community. Score one for the community, yeah!! And yes, btw, this is made possible by Sun’s requirement of joint control of IP from contributors. (More here)

“Commercial source code licensees may optionally buy additional services and gain additional rights, including: … The right to make modifications to the source code without putting those modifications back into the open-source code commons. “

[3] Sun believes that they can attract customers and drive more revenue by using the “open source” tag than they’ve been able to thus far. (More here)

“…In order to gain the benefits of commercial support, and predictability, these customers may choose to use Sun’s commercial distribution of the JDK or JRE in conjunction with a support contract. However, they would not consider the commercial product if the open-source version were not available.(emphasis added)

Sun gets the glowing press and keeps control of Java, what more could they ask for?

Next Page »