Thinking about it, libraries and trust groups are very part of the core SquirrelJME and really should not be implemented as services since I can see many things dependingon them. So I will split them off and into the core system accordingly.
Now, what I need again is a way to install and utilize libraries and such so I can run and launch the test framework from the build system which would be quite useful for the most part.
Although it is not truly needed in a way. But one thing I do need is a system call argument encoder and decoder which checks things and can be used as a means to check and ensure that system call arguments are correct and well formed before they are sent over the wire.
But, I definitely want to work on the JIT and such so the test framework would be a big thing to do. Also I have an idea for less duplicated code and such. Instead of having there be separate clients and servers for everything instead there is just a single server in the server and where it is needed in the client the code is there. That way there are no two duplicates of the same code such as the library code for example.
But I think the first thing will be the array classes and such, but that can be done via a template for the most part.
But thinking about it. I think it might be best to work on the compiler and WinterCoat at the same time. The first thing would be testing and a basic interpreter and such for compiled WinterCoat code.
In the long run the Java SE environment rather limits me in implementation if I try to utilize it to the max. I think it would be best only for build system usage and not something that can launch real applications to do things. There is hosted launch for such environments though. This would simplify things and I can design the kernel around real-ish hardware rather than having it work on a Java SE environment where I really cannot do some things reliably.