Open Source Blog

Fleury seems to have been particularly angered by recent reports that IBM considered acquiring JBoss but decide to pass on the company because it did not like the company's attitude to community development.

Adam Jollans, IBM software worldwide Linux strategy manager confirmed the Cnet story to me this week, explaining that IBM believed that community-based projects like Apache are a better way of fostering innovation than vendor-led communities such as JBoss or MySQL.

"That's why we decided to go with Gluecode," he said. "We clearly looked at the market, but we believe an open community is very, very key."

Fleury has another take on the decision, suggesting that in fact IBM preferred the Apache license to JBoss's chosen LGPL because it enabled the company to combine the Geronimo code with its own to create the WebSphere Application Server Community Edition product.

"While the resulting product is sometimes free, it is no longer 'open source' by any stretch of the imagination," writes Fleury. "This… is why competing commercial vendors prefer BSD-style licensing. At JBoss, we prefer FSF licenses, such as LGPL, that prevent the strip-mining while at the same time allowing end users and ISVs to reuse our code."

While the WebSphere Application Server Community Edition product is available for download and use at no charge, and users have the right to make unlimited copies for internal use, the code is certainly not all open. Only the source code for the underlying Geronimo components is viewable and modifiable.

Hence this statement from IBM's product FAQ: "If you have a brilliant idea for improving the application server implementation or for enhancing the application server for your environment, choose Apache Geronimo so you can obtain all the relevant source and build your own application server. Although it is not necessary, you may also consider contributing your changes to the open source community."

With that in mind, you can perhaps understand why Fleury thinks IBM is guilty of double standards in criticizing the JBoss business model. "JBoss has always been about pure open source. We started in OSS and we will die in OSS. It rubs me the wrong way when our competitors mischaracterize us, the real developers of FOSS, for not being OSS-enough according to their view of the world."

It all goes to underline how important it is for potential users to understand licensing before they take a decision to use a particular project or product.

Fleury also takes a swipe at BEA, accusing the company of 'waste dumping' code into the open source community. "Eager to be part of the open source wave, the vendor identifies some technology that is inferior or of limited value to them, and they dump (oops...sorry...they 'donate') it into open source," he writes.

"There has been so much waste dumping going on lately that we may very well need an 'Open Source landfill' to deal with the cleanup of all of this waste and its damage to our environment."

Whether this is fair criticism depends on how much value one sees in the code that has been made available. While it is undeniable that some vendors seem to think that open sourcing an aging product will prop up shrinking market share, it is perhaps a little unfair to criticize all such donations.

The very fact that the code is out there (if it has been licensed under an appropriate license) means that it has the potential to provide an opportunity for enterprising developers or businesses to make use of it.

If the code is of no little or no value then it will be ultimately be neglected, but this should be of more concern to the donator that fails to maintain market share than the open source developer who is free to ignore it.

One of the great things about open source is that it can be seen as natural selection in action. I wrote before (in discussing a very different approach to open source being taken by IBM as it happens) that the key to the initial success of an open source project is not vendor support and marketing dollars, but end user acceptance and individual developer involvement.

The same is true for code donations, and to some extent what I'll call 'code adoption' (although marketing dollars do provide an unnatural short-term buffer to natural selection.

Which of the approaches taken by JBoss, IBM, and BEA will prevail? They can argue all they like, but it is the user community that will eventually decide.