dot in search PATH

ian at utcsstat.UUCP ian at utcsstat.UUCP
Fri Aug 5 04:54:16 AEST 1983


	The discussion of whether to allow '.' in the $PATH reminds me
	of an incident that occurred where I used to work.  
It reminds me of a debate in metaphysics. 
And who has been discussing whether to `allow' . in $PATH? Isn't
this a little like `allowing' people to breathe?
	When we first got our VAX/UNIX system up one of the people in charge of
	the system wrote two utilities which he called au (add user)
	and du (delete user) and which were very useful for doing those jobs.
A wonderful choice of names. If he'd been thinking UNIX instead of YAMOS,
he might have used `rm' instead of `delete' as a model, and never had
the problems described below.
	He placed these in the /etc directory.  About 3 months
	later the other system administrator was trying to determine
	how much disk space was used by a certain user.  So he typed
	'du /user/user_name', unfortunately he was in /etc at the time
	and '.' was the first entry in his $PATH.  When du didn't print
	the expected data on the screen he scratched his head and typed
	'du /user/*', fortunately a few seconds into this he aborted
	the command.  By that time though he had removed 4 users and
	their entire directory structures and partially removed a fifth
	(me).
Alternately, the other administriviator might have put in one prompt before 
removing a user's entire file structure.

	Needless to say du was quickly renamed 'removeuser' and the
	administrator paid the price by having to spend a couple of
	days restoring all our files from backup tapes.
Needless to say.

	I hope that this illustrates the problem of having '.' at the
	start of your $PATH and the problem of not choosing resonable
	names for dangerous utilities.  --
It actually illustrates the danger of having two system managers
who can't talk to one another.
And the danger of being root all the time. Why do you need to be root
to run a normal `du' command?
But butchering UNIX to disallow `.' in your PATH variable is not
a reasonable solution.

	John Vander Borght ...pur-ee!sequel!van

Ian Darwin, Toronto (utcsstat!ian)



More information about the Comp.unix.wizards mailing list