Re: Adding of peter gutmann's paper concept to /bin/rm

[Marcus Watts <mdw_(_at_)_umich_(_dot_)_edu>]

Jim Rees <rees_(_at_)_umich_(_dot_)_edu> writes:
> I think that Garfinkel and Shelat's claim that "51 had been freshly
> formatted; 19 of those still held recoverable data" is probably wrong.  I
> suspect they are making the common mistake (reinforced by Redmond) of
> confusing "formatting" with newfs.

Their paper is here:
	http://www.computer.org/security/garfinkel.pdf
They define quite clearly what they mean by "format", and yes, it's
the MicroSoft definition == what we know is really "mkfs".  Even
what is called a "low-level" format in the MicroSoft universe may
not really do much more than zero data blocks on many modern disk
drives - bad things happen when you erase embedded servo data.

The point I think they really make is that if you're going to throw
away a drive you should sanitize it first.  Rather than trying to "fix"
rm, slightly more useful utilities for the OpenBSD world would be tools
to zero unused space in filesystems, and tools to do the multiple
overwrite thing on a partition or whole drive basis.  But I can't say
that I see this as being all that exciting.

I sometimes wonder just how effective the "overwriting" strategy is in
"kdestroy".  At least "rm -P" does an fsync.

				-Marcus Watts

[ I still get a kick out of the difference between "boot" and
"system" disks in Windows XP. ]