I beg to disagree. Its simply a matter of how you structure your code. All
the "cluster" stuff changes is adding some fields to each "object" structure
eg file now starts
struct file
{
CLUSTER_OBJECT
[normal fields]
}
and preprocessor magic does that. Also syscalls simply require we swap
the syscall table for cluster_sys_xx entry points, and provide
linux/cluster/*.c which will call sys_xx for local objects. I see no
need to seperate the trees