Need to next setup a class for reading from the block accessor.
So I can read uncompressed entries in the ZIP for now. I can dodge reading deflated entries for a very short duration. However one thing that I need to fix is the out of date check.
System suites, if loaded, need to be mapped to suite accessors or just appear in any order.
So one thing I need to determine is the best way to represent processes in the real system and in the interpreter. I suppose what I need are pseudo processes. Basically there is just a virtual representation of a process. The interpreter uses this. The kernel itself when it is running on native code does the same thing with it also. Potentially interpretation can be used too but that I will not actually do. Effectively the process will be managed by the native thread code and such. So this basically means that control of processes is completely wrapped. Then the thread manager will handle actually running them as needed. So even if a process was initialized it may not have ran yet or have had any of its initializers ran.
So yes as before, it is purely virtualized and actual execution is hidden from the kernel.
So KernelThread and KernelProcess would probably best not be final or be virtually wrapped. They can just be directly created and returned. That would simplify things greatly for example.