Relay-Version: version B 2.10 5/3/83; site utzoo.UUCP
Posting-Version: version B 2.10.1 6/24/83; site mit-athena.UUCP
Path: utzoo!watmath!clyde!cbosgd!ihnp4!mhuxn!mhuxr!ulysses!allegra!bellcore!decvax!mit-athena!jc
From: jc@mit-athena.UUCP (John Chambers)
Newsgroups: net.crypt
Subject: Re: Why no hardware random numbers?
Message-ID: <155@mit-athena.UUCP>
Date: Mon, 1-Apr-85 11:21:00 EST
Article-I.D.: mit-athe.155
Posted: Mon Apr  1 11:21:00 1985
Date-Received: Thu, 4-Apr-85 05:48:49 EST
References: <868@utcsri.UUCP> <464@lll-crg.ARPA>, <442@x.UUCP>
Organization: MIT Project Athena
Lines: 32

>>GEE,did I discover a new effect with my Monte Carlo program or was the random
>>number generator a little off last tuesday.  I can't find out because I can't
>> generate the same sequence again.
>
>                                        ...  Take the hardware random number
>generator and dump a few million numbers out to disk.  You have a repeatable
>sequence (and if you have good buffering, it might be a faster repeatable
>sequence than the RANDU you're probably tempted to use instead).  

This does give you the same sequence of random numbers, but fail to
solve one important aspect of the original suggestion:  It can't be
used to test the original program.  Note that the original program
called the random-number generator; it didn't read from a file.  To
use the file, you have to write a new "random-number reader" routine,
compile it, and relink the program to include the new routine.  You
now have a new program whose behaviour will probably not show the
same set of bugs as the original.  

This approach would be a good one for producing a reproducible, truly
random sequence.  But to be useful for debugging purposes, it would
have to be incorporated into the library routine from the start, so
that when his bug pops up some Tuesday, he can test it on the same
sequence without changing any of the code.  

(Yes, I am truly paranoid.  It comes from years of trying to debug 
programs, many of which worked just fine with the debugging output 
enabled.)
-- 

			John Chambers [...!decvax!mit-athena]

If you're not part of the solution, then you're part of the precipitate.