Not surprisingly, Simon didn't like my characterization of Sun as "not open". In the mean time, Anne responded to Simon's assertion that Java is open. Her reply captures a number of my issues. But here's what I think is at the bottom. We're working with two different ideas about openness. Simon is comparing Sun and Microsoft and saying that Sun is clearly more open than Microsoft. And therefore Java is more open than the CLI/C#. In my mind it is clear that Sun is more open than Microsoft, and perhaps I should have given Sun some credit for the steps it has made towards being more open. In my mind, this isn't about whether Sun or Microsoft is more open. That's easy. Sun is. But for me the standard of openness (which Anne correctly notes is not as open as public domain), is open source style openness. And one way (but not the only way) of measuring that openness is to ask the question "Can I fork"? If I can fork, then it's open. If I can't fork, it's not. The purpose of branding and certification and so on is to prevent forking, extension, etc. But the reason that I'm dissatisfied with Java as a substrate for the future is that it can't fork. Stuff done in the JCP can't fork. It's the same reason that I'm dissatisfied with the .NET additions to the CLI (suspicion of Microsoft notwithstanding). I want something that I can fork. I want the ethos of the "platform" to be ruthless competition among similar technologies.

I believe that this can happen on the CLI or a fork of the CLI. I'm not saying that we should take the CLI and do whatever Microsoft says. I'm saying we should take the CLI and make it do what we want. And if it stops being compatible with the MS/ECMA CLI, then I don't particularly care. We fork it.

Simon pointed out that there is an open source VM. I suppose that's Kaffe, which has never performed that well, and doesn't seem to be improving that fast. The Mono people are improving at a very high rate. What they have and Kaffe doesn't have is an active, flourishing developer community that seems bent on making it usable for real production work. Kaffe isn't - just read their website. So maybe I should have said usable open source JVM. Fine. All we need is one, it doesn't matter who it comes from, Sun, IBM, Apple, GNU, I'm not picky.

Simon asked why we couldn't take Geronimo and chart our own course. For me, it's not just about the "Enterprise application framework" space. Linux needs a managed code environment like the one in the CLI. For desktop applications as well as server applications. I personally want a VM that will be or can be made to be friendly to languages outside the Java/C# space. If you want to talk about the J2EE equivalent space, I believe the minimal containers like picocontainer or Avalon, coupled with some form of Aspect Oriented Programming will prove to be much better fits for the enterprise computing space.

This isn't about whether I'm personally fond of Sun or Microsoft. It's about having a platform where you can hack all the pieces. The Dashboard demo shows that if you have the source for everything, you can accomplish very impressive desktop application integration in time on the order of weeks. When you don't have the source, when you have to answer to a compatibility board, you can't do that. And that's what I'm interested in.

There's only one thing that bothers me, and that's the GPL licensing. Anne doesn't believe that this is a problem. The only thing is, read these two faqs. They both say that GPL libraries infect applications even in programming language interpreters. That's the reason that JBoss is LGPL and not GPL. And the meaning of the LGPL is ambiguous in the fact of Java's (or CLI's) dynamic linking model. This is not a show stopper for me, but it could be for attracting support. For what it's worth, I don't consider this to be particularly open either. There, now both Sun and the FSF aren't open and Simon should feel honored to be in such company ;-).