+ /** @brief Gives the control from the given user thread back to the maestro
+ *
+ * schedule() and unschedule() are the basis of interactions between the user threads
+ * (executing the user code), and the maestro thread (executing the platform models to decide
+ * which user thread should get executed when. Once it decided which user thread should be run
+ * (because the blocking action it were blocked onto are terminated in the simulated world), the
+ * maestro passes the control to this uthread by calling uthread.schedule() in the maestro thread
+ * (check its code for the simple semaphore-based synchronization schema).
+ *
+ * The uthread executes (while the maestro is blocked), until it starts another blocking
+ * action, such as a communication or so. In that case, uthread.unschedule() gets called from
+ * the user thread.
+ *
+ * As other complications, these methods are called directly by the C through a JNI upcall in
+ * response to the JNI downcalls done by the Java code. For example, you have this (simplified)
+ * execution path:
+ * - a process calls the Task.send() method in java
+ * - this calls Java_org_simgrid_msg_MsgNative_taskSend() in C through JNI
+ * - this ends up calling jprocess_unschedule(), still in C
+ * - this calls the java method "org/simgrid/msg/Process/unschedule()V" through JNI
+ * - that is to say, the unschedule() method that you are reading the documentation of.
+ *
+ * To understand all this, you must keep in mind that there is no difference between the C thread
+ * describing a process, and the Java thread doing the same. Most of the time, they are system
+ * threads from the kernel anyway. In the other case (such as when using green java threads when
+ * the OS does not provide any thread feature), I'm unsure of what happens: it's a very long time
+ * that I didn't see any such OS.
+ *
+ * The synchronization itself is implemented using simple semaphores in Java, as you can see by
+ * checking the code of these functions (and run() above). That's super simple, and thus welcome
+ * given the global complexity of the synchronization architecture: getting C and Java cooperate
+ * with regard to thread handling in a portable manner is very uneasy. A simple and straightforward
+ * implementation of each synchronization point is precious.
+ *
+ * But this kinda limits the system scalability. It may reveal difficult to simulate dozens of
+ * thousands of processes this way, both for memory limitations and for hard limits pushed by the
+ * system on the amount of threads and semaphores (we have 2 semaphores per user process).
+ *
+ * At time of writing, the best source of information on how to simulate large systems within the
+ * Java bindings of simgrid is here: http://tomp2p.net/dev/simgrid/
+ *
+ */
+ public void unschedule() {
+ /* this function is called from the user thread only */
+ try {
+
+ /* unlock the maestro before going to sleep */