Greg A. Woods writes:

>Ah, no, sorry, that's not right.  Swap handling is enabled and modified
>by a very non-arbitrary user "process" (swapctl).  Disabling it for
>shutdown is, and should, also be done by that same program.

But swap is a kernel entity; surely your argument could be extended to
the buffer cache aswell; since it's mostly (only?) user programs
writing to the disk thru the bufcache in the first place, shouldn't
syncing the cache with the disks before reboot/halt done by a user
process aswell?  Yet the kernel does this automatically when unmounting
the filesystems (to my knowledge).  IMHO it would be just a lot more
practical if the kernel would properly unconfigure swap devices itself,
especially when raidframe is being used.

-- 
Matthias Buelow
home: mkb/at/mukappabeta.de
uni:  mkb/at/informatik.uni-wuerzburg.de