Process-based scheduling for the block I/O layer sounds wonderful. Can we look forward to systems that no longer become completely unresponsive when the disk I/O gets too heavy? (That was meant to be happiness at the thought of what is coming, not complaining about what is now...)