Apache/BSD have been good for standardizing protocols such as HTTP/web server, SOAP (maybe), TCP/IP…etc. It also has been a good license for proliferating some web development tools. There wasn’t money to fight over in these areas for the vendors so these things tended not to fragment. Tomcat is the only exception since it is a simple platform and not a business nor technical control point. The Apache success stories are mainly about interoperability (though web services seems to have not worked so well to date…perhaps IBM figured it could take control…).
Platforms on the other hand have fragmented under Apache/BSD, notably the *BSD UNIXes. This is just the latest in a series of fragmentations under these types of licenses and arrangements. The Open Software Foundation (OSF), USL and other early UNIX efforts that were similar multi-vendor operations failed as well.
(L)GPL is best for standardizing platforms and APIs. Since platforms and APIs are business and technical control points, vendors will fight over control. GPL Linux and LGPL JBoss have not fragmented due to the strength of the license protecting the community. Further, the leaders of these communities did not have conflicting agendas - such as large competing proprietary-based revenue and profit streams.
While J2EE has not been open source, Sun managed the community under rules that had were more similar to the proprietary and GPL license model than the much more easily fragmented Apache / BSD license model. Hence, J2EE only fragmented at the periphery, that is, beyond the standard that Sun licensed.
BTW, Microsoft gets this. They work with IBM on web services protocols and other protocols for interoperability. But won’t work with IBM on platform or API standardization. They recognize that ISVs et al, need a stable platform to aim at and support which only Professional Open Source, (L)GPL, or, in Microsoft's case, a proprietary license can give.
Red Hat's founders also got this. That's once of the reasons why they picked GPL Linux over one of the *BSD distributions. Because of this choice, they are one of the key operating system leaders today.
Hence, I believe there is significant risk in fragmentation in the IBM - Apache Geronimo effort. Assuming any real industry support materializes (none has to date), multiple vendors will likely drive multiple, inncompatible implementations to compete. The hypothetical vendor(s) and IBM would fight and fork to compete, as previous BSD / Apache-style licensed platforms have in the past.
Actually, I have a great deal of difficulty seeing IBM invest much beyond some development resources in Geronimo, perhaps encourage fragmentation, and then point to WebSphere as the stable and robust J2EE answer. IBM's business and stock price cannot afford a self inflicted shrinkage of WebSphere revenue that a truly successful Geronimo would create. What would be the reward for IBM WebSphere executives to miss revenue goals by 10% or 20% in 2007 due to their own strategy? That is the biggest conflict of interest of all.