I sat down for an interview with Carl Quinn during Javapolis, and the video is now up on parleys.com. You can also find an audio-only podcast version on the Java Posse's web site. In it, we talk frankly about:

• The history of and philosophy behind Guice, what problems it solves, how we implemented it, the secret sauce in Guice's stellar performance, and how Guice compares to other approaches.
• How we in the Java community need to take a step back and rethink API and framework design after Java 5 and 6. Reflection heavy and convention based Java 1.4 designs won't fly for much longer. Readability doesn't necesarily mean fewer characters. Java 5 made static type safety a reality and introduced explicit meta data. Java 6 lowered the barriers to compile time code generation, but good documentation of these new features is still much needed.
• The future of Guice and dependency injection in general. I'd like to see dependency injection on mobile devices and deeply integrated into Java SE. One day, many of the problems we're solving in a separate framework will be addressed seamlessly by the core platform, and we'll be free of integration headaches and able to focus on higher level abstractions all around.
• Google Collections, an excellent open source project led by Guice co-lead Kevin Bourrillion. Check it out if you haven't already. Kevin and the rest of the team try to hold the Google Collections to the same standards as the core Java SE Collections, and it shows.
• Weak vs. soft references, which you should use, and specifically when not to use soft references.

Thanks again goes to Carl and Stephan and the rest of the BeJUG team. Javapolis was one of the most fun, inspiring, community-oriented conferences I've been to in a long time. Getting back and forth between your hotel and the conference was a little tricky at times, but the venue itself absolutely rocked.