[net.unix-wizards] recovering lost files

dan (05/26/82)

References: sdcsvax.2016

A problem with UNIX?  I have not accidentally destroyed an important file
in several years on UNIX, though I have on RSX.  I guess a klutzy user
interface creates a need for file versions.  UNIX may be a bit eccentric,
but it is not klutzy.  Programs like cc can (and have) been changed to
make sure the output file specification is reasonable.  I recommend daily
(incremental) backup in case an important file is lost.  Any file not worth
going to a backup tape for is probably not worth the effort of maintaining
a file system with file versions.

	Dan Strick

bcw (05/27/82)

From:	Bruce C. Wright @ Duke University
Re:	Recovering lost files

I suspect that the reason Dan Strick (mcnc!idis!dan) hasn't destroyed
an important file in several years on Unix but has on RSX is that hs
is more familiar with Unix than he is with RSX.  My experience has
been otherwise;  I've destroyed MANY important files on Unix (and it's
a lot easier to do with a terminal which mungs your input before sending
it to the system!).  This doesn't excuse the rather klutzy user interface
on RSX, but the solution proposed doesn't solve the problem very well
either:  usually the reason you destroy a file is that you are trying to
do something to it, which means that probably significant information
has been added to it since the last backup (even if it's done daily).
Maybe we should have hourly backups?

			Bruce C. Wright @ Duke University

markp (05/27/82)

When I give an order, I expect it to be OBEYED, not questioned.
If I type "rm *", I want those files removed NOW!!

A knife is a wonderful tool--it cuts what you tell it to.
It's up to YOU not to cut your fingers.

					Mark R. Paulin

ARPAVAX:CAD:ESVAX:Cory:cc-18 (05/29/82)

"A knife is a wonderful tool--it cuts what you tell it to."

So does a chain saw.  "rm foo*" versus "rm foo *" is a slip of the
finger, or of the mind, and is of the order of using an unguarded
chain saw.  Would the author of the above show any fear if I habitually
used a chain saw in his living room, trusting in me not to fumble
to preserve his furniture or his limbs?

Furthermore, fingers cut with knives heal.  Links cut with "rm foo *"
don't.  Hence the chain saw is the better analogy.

~r .sig
~p
(Aw, goddamit!)

Michael Chastain
Cory.cc-18@Berkeley (ARPA)
ucbvax!Cory:cc-18   (UUCP)

rjk (05/29/82)

I suspect that the overhead involved in idiot-proofing the "creat" syscall
(eventually used by shell ">") or the "rm" program would be prohibitive.
Here are some suggestions:
1) File is "foobar.c", you are developing it and want to protect it.  Every
   so often you type:  save foobar.c  which stashes it in a safe deposit
   directory.  (I'll be glad to post the source to this command)
2) File is writable data.  Ok; same thing.  After modifying it, save it.
3) File is readonly data.  Easy:  chmod 444 data_file.  At least rm will
   now ask you and shell will not create.

As said before, if YOU deem the file THAT important, then take a couple of
seconds to ensure that YOU protect yourself.  It's like asking auto
manufacturers to ensure that we can't run into another object unless we
REALLY mean to. (wreck -y??).  You learn to drive, no?  Then learn to be a
careful UNIX user.  It's more powerful that way.
							Randy King
							..we13!rjk
PS> Look in net.sources.  I'll put "save" there.  It's been running on
    our system for years.

geoff (05/31/82)

Enough already about rm! This has been endlessly discussed in the
Datamation article and its replies. The obvious solution is to have an
environment variable (say, CAUTIOUS) to select alternate behaviour such
as asking for confirmation.

Geoff Collyer, U. of Toronto Computing Services

pavel (06/02/82)

Better than trying to remember to use a command like 'save' is to set the
csh variable 'noclobber', which will at least prevent you from '>'ing
into an existing file.  If you don't use the csh, then here's another
reason to do so.  BTW, I don't buy the explanation that people necessarily
must remember to be careful; they will always make mistakes and a good
environment should either notice dumb things and ask for confirmation or
provide a method for 'undoing' such errors.

	Pavel Curtis

Douglas@sri-unix (06/27/82)

Date: 10 Jun 1982 at 1058-CDT
Probably editors should keep ONE backup of any edited file, for non-
SCCS users.  If you can't recover from sources then you should deliberately
back up (data?) files.  I can't think of any general backup/recovery scheme
that doesn't have at least one application for which it is ill-suited.
-------