I completely agree. The sequence of events should be that we convince
ourselves that MPC is ready for prime time on all of the major host
platforms, and then that we check in support for it, along with the
requirement that MPC be available to build GCC. I don't think we should
check in this code until we are ready to require MPC.
Of course, we might decide that the benefits of MPC are outweighed by
disadvantages if it doesn't work on important host platforms. But,
hopefully we don't have to make that decision.