WebMink
Simon Phipps's personal commentary

Open source and forks

Robert Scoble comments on the fork of JBoss by the Elba project but doesn't dig any deeper. In particular, one might come to the conclusion that forks like this are an every-day occurrence. In fact, external forks of established projects like JBoss happen once in a blue moon and when they do are a sign of a much deeper malaise. Yes, all society is driven by human interactions (otherwise known as 'politics') but a substantial fork like this is a real rarity and deserves scrutiny.

Elba was intended by its creators to be part of Apache's Geronimo project, announced at LinuxWorld - the FAQ explains:

This project was originally going to provide code that allowed Geronimo to run in JBoss™, but since there seems to be a policy of removing anyone that works on Geronimo from the JBoss™ project, we decided we had to fork.

Apache aren't happy with this idea according to Noel Bergman, and Apache seems likely to insist code gets rewritten or closely scrutinised to avoid any commercial taint, but even if Elba isn't part of Geronimo its creators are and their intent is clear. I've been anticipating the creation of Geronimo ever since JCP 2.5 was ratified and now it's started wish it every success.

It would have been obvious for this project to have worked hand-in-hand with JBoss - after all, pretty much every other J2EE-related open source project has joined the party. But in this case, it turned out that wasn't going to happen. Not just because of the JBoss LGPL license usage cited on the Geronimo FAQ though. In fact, some of the committers from the JBoss project had their ability to participate in the JBoss community revoked unilaterally because of their offer of their own code to Geronimo (I also have this first-hand). As one comment says,

The fact that so many key developers have left the JBoss project over the years indicates that at least for some, there is a significant problem with the JBoss project.

Sure the code is open - but the process is not. Even when on the JBossGroup mailing list, many discussions took place behind closed doors and were not negotiable.

I do not wish to diminish the massive job that MF has done to get JBoss to where it is today - that is something that only incredible hard work and dedication could have achieved. However, in return for his organizational contributions he has obtained a near monopoly on the development process and commercial exploitation of the project.

So it seems this forking rarity is a response to widely-recognised issues, not with the JBoss code or even with the JBoss project, but with the eponymous company pulling the strings behind the scenes. As Stephen commented, prima facie the Apache project was getting special treatment with regard to J2EE certification but in fact the issue was deeper and darker.

What do we learn from this? Well, the big lesson I'd send back to Robert is that, when it's open source, there is a remedy. It's rarely used as it's extreme medicine, but I predict it will work in this case. Now, what's our remedy for the code-base that's causing most of us extreme e-mail stress at the moment? Seems we just have to keep on trusting or switching...

[Modified to reflect Noel's comment]

posted at 10:35 AM (UK) | |