One important use case of microLinux is the size minimization of dedicated single-purpose embedded systems. In cases like that, it should be possible to apply various techniques to exhaustively determine which code paths will and will not ever be executed on such a system. This is actually a superset of 'automatic syscall elimination' but recognizes that there's nothing special about syscalls in this respect - all kernel code should be subject to these techniques. I think some of the existing dynamic tracing infrastructure as well as some refactoring and reuse of certain parts of the hash triggers patchset could be applied to this problem. In my ELC talk, I called this 'Trace Guided Optimization' - basically the application of tracing techniques to optimize out unused parts of a given single-purpose system.
MicroLinux is moved out of project