On Fri, 30 Aug 2002, Richard Earnshaw wrote:

> > That doesn't really help.  I had a disk crash recently that consisted of 
> > a few scattered
> > sectors on the disk suddenly being unreadable.  Before this point in 
> > time everything
> > was fine, but suddenly... BOOM.  And after that, you had to be very 
> > careful which
> > executables you tried to run.  A crunchgened /rescue would probably have 
> > meant
> > that I was hosed at that point.  But if you have an old machine you 
> > probably have
> > a floppy to boot from at that point.
> 
> If your bad sector had been in /bin/sh you'd have been equally hosed.

Nope. You also have /bin/csh, which has no relation to /bin/sh.

>  If 
> you have a probability N of a bad sector on a disk and P sectors of 
> 'critical' code needed for recovery with statically linked executables and 
> Q sectors of 'critical' code with a crunchgened executable, then the 
> probabilities of the bad sector lying in your critical code are P*N and 
> Q*N respectively.
> 
> Given that practically everything in /bin and /sbin could be considered 
> critical for some problem, you are clearly better off if Q < P.

But this is where you go wrong. Everything is /bin and /sbin are critical,
but several of the tools can be used to do tasks other tools are desgined
for.
For instance, you have two shells, echo can be used for ls, cat for cp, cp
and rm for mv, nfs can sometimes circumvent ftp and vice versa, and so
on...

So chances are that a single block blown will not really stop you from
getting things in order again.

	Johnny

Johnny Billquist                  || "I'm on a bus
                                  ||  on a psychedelic trip
email: bqt@update.uu.se           ||  Reading murder books
pdp is alive!                     ||  tryin' to stay hip" - B. Idol