Path: utzoo!utgpu!news-server.csri.toronto.edu!bonnie.concordia.ca!thunder.mcrcim.mcgill.edu!mouse
From: mouse@thunder.mcrcim.mcgill.edu (der Mouse)
Newsgroups: comp.unix.questions
Subject: Re: What does '*' symbol in /etc/passwd means?
Message-ID: <1991Jun18.052517.18133@thunder.mcrcim.mcgill.edu>
Date: 18 Jun 91 05:25:17 GMT
References: <27176@adm.brl.mil> <1991Jun14.002427.6120@csc.canberra.edu.au> <1991Jun14.051958.17564@colorado.edu>
Organization: McGill Research Centre for Intelligent Machines
Lines: 27

In article <1991Jun14.051958.17564@colorado.edu>, frechett@spot.Colorado.EDU (-=Runaway Daemon=-) writes:

> Two more notes.. 
> 1. I say crypt(3) because crypt(1) is totally different.
> 2. crypt(3) is purposely designed to take a HUGE portion of CPU when
>    encrypting which makes passwd cracking very slow and fairly
>    visible.  If I just run one guess through every line of the
>    /etc/passwd file on my DEC5500 (about 28 Mips) it hangs about
>    every 5 seconds for up to 20 seconds.. The machine just can't
>    afford to keep the process in memory all the time.

Then this is a deliberate crippling of your crypt(3) library routine,
or your machine is severely overloaded.  There is nothing sufficiently
intrinsically difficult about it to justify this sort of (lack of)
speed.  On a Sun SPARCserver 470, for example, I have seen code that
runs at about 1000 encryptions per second, sustained.  (I think this
machine is supposed to be very roughly comparable to your DEC5500; my
hazy memory says Sun claims it's in the low tens of MIPS....)

(And yes, you're entirely right in your note (1).  crypt(3) is a
slightly tweaked DES; crypt(1) is a one-rotor Enigma machine.)

					der Mouse

			old: mcgill-vision!mouse
			new: mouse@larry.mcrcim.mcgill.edu