Path: utzoo!utgpu!water!watmath!clyde!att!osu-cis!tut.cis.ohio-state.edu!bloom-beacon!bu-cs!purdue!decwrl!sun!imagen!atari!portal!cup.portal.com!dan-hankins
From: dan-hankins@cup.portal.com
Newsgroups: comp.sys.amiga
Subject: Re: The ultimate fix!!!
Message-ID: <9997@cup.portal.com>
Date: 13 Oct 88 00:28:31 GMT
References: <8810092209.AA09182@cory.Berkeley.EDU> <2791@sugar.uu.net>
Organization: The Portal System (TM)
Lines: 39
XPortal-User-Id: 1.1001.5361

In article <2791@sugar.uu.net> karl@sugar.uu.net (Karl Lehenbauer) writes:

>Do you disagree that Unix systems are better protected than totally 
>unprotected systems like the Amiga and the PC

Not to answer for Matt, but yes, I think I disagree.  It is not memory and
file protection that protects Unix from viruses.  As far as I can tell,
there is only one real protection Unix has against viral attack.  Unix runs
on such a large number of incompatible architectures that people do not
share executables.  Instead they share source and do recompiles.

Source sharing isn't perfect (it is possible to write a virus that will
infect via a combination of source code and executables), but it's much
better than sharing executable code.

It doesn't protect against viral spread on a single machine (I would
imagine that a lot less source sharing goes on on a single machine where
the architecture is obviously compatible), but it does severely limit the
spread of viruses over the net.

>are you saying that Unix is somehow wide open such that running
>mischevious code from any signon means  the system's hosed (and if so,
>how?), that people shouldn't even bother with  countermeasures, or what?  

Maybe Matt's saying that, but I'm not.  I do believe that there are
measures which can be taken to severely limit the spread and
destructiveness of viruses.  However, these methods involve entirely new
operating systems.  For existing systems, the best that can be done is to
have someone whose job it is to hunt down these suckers and come up with
antidotes on a one-per-virus-class basis.  This may not even be possible
for some classes (for instance self-mutating viruses).

I believe that the real answer to viruses (not total but very good) is to
implement an object-oriented operating system in which objects are treated
like users on a multiuser system.  Implement 'need-to-know' and 'secret
clearance' type restrictions on the objects in the system.


Dan Hankins