The other Dave Chappell is suggesting that the Java community is fracturing on whether Java is a useful programming model for SCA. Let me give my perspective on this, since I've commented on this extensively in the SCA working group and in other forums.
There is no doubt that adding support for imperative programming models is important for SOA composites. This is where the Java/SCA model makes sense.
But does it matter? My argument all along has been that we should not teach a new model around Java. Everyone is already converging on a common model: the whole Java world is moving toward pojo programming stripped free of the framework cruft. In the first SCA f2f I sat with Rod Johnson and watched him implement some of the annotations in the original SCA JCI spec in a few minutes. SCA frameworks will bootstrap Java code. Write it for Spring and it will be useable in SCA. Or vice versa.
Put it another way: we can bootstrap basic BPEL processes into an SCA environment without changing the BPEL definition because you just use the language. We expect to see the same thing for Java platforms moving forward. If anything, this is the ultimate form of industry convergence.