> When I press the power button on an x86 PC, acpibtn(4) receives the > event and shuts down the machine.
that is a absolutely needed thing to have. this functionality does a smooth shutdown without need for fsck on restart. i have depended on this feature whenever i test something, and the machine hangs for any reason whatsoever. hasn't happened much but it si a lifesaver whenever it has happened (fsck problems on bigmem machines a couple of months ago, vmmap testing etc). > When I press the power button on my Blade 100 (sparc64), power(4) > receives the event and by default ignores it. Only if the > machdep.kbdreset sysctl is set to 1 will power(4) proceed to shut > down the machine. > > I don't have a preference either way, but this is inconsistent. > Either acpibtn(4) should also be protected by a sysctl, or power(4) > shouldn't. this should be a sane default (i.e halt & power off on power button press) if acpibtn() is protected by sysctl. Windows gives you a choice and so does Linux buried somewhere in control panel/center.