*BSD News Article 89448


Return to BSD News archive

Path: euryale.cc.adfa.oz.au!newshost.carno.net.au!harbinger.cc.monash.edu.au!news.cs.su.oz.au!metro!metro!munnari.OZ.AU!uunet!in3.uu.net!194.221.49.2!news.julia.de!erbse!ralf
From: ralf@infko.uni-koblenz.de (Ralf Baechle)
Newsgroups: comp.os.linux.misc,comp.os.linux.advocacy,comp.unix.bsd.misc
Subject: Re: Measuring "stability" (was: Linux vs BSD)
Date: 13 Feb 1997 19:19:48 GMT
Organization: Uni Koblenz, Germany.
Lines: 38
Message-ID: <5dvpgk$td$1@alles.intern.julia.de>
References: <32DFFEAB.7704@usa.net> <slrn5fejrn.353.bet@onyx.interactive.net> <ywtn2te3olm.fsf@math.ucsb.edu>
NNTP-Posting-Host: erbse.uni-koblenz.de
To: Axel Boldt <boldt@math.ucsb.edu>
Xref: euryale.cc.adfa.oz.au comp.os.linux.misc:159094 comp.os.linux.advocacy:84369 comp.unix.bsd.misc:2535

In article <ywtn2te3olm.fsf@math.ucsb.edu>, Axel Boldt <boldt@math.ucsb.edu> writes:
|> 
|> I think crashme is a good start for testing stability, the only
|> problem being that it does not test the networking code and the
|> filesystems. Does a similar tool exist that rapidly sends garbage to
|> random ports? Should be trivial to write in any case.
|> 
|> We could then have a stability test that works something like this:
|> 
|> - two crashmes running in parallel
|> 
|> - repeatedly tar and untar a big directory (dir should not be
|>   writeable by the user running crashme)
|> 
|> - one outside random network attacker
|> 
|> - one httpd constantly being queried for a big doc (again, not
|>   writeable by crashme user).
|> 
|> Clearly, we would have to specify which versions to test
|> (I vote for the "stable" releases: Linux 2.0.x and FreeBSD 2.1, soon
|> 2.2), which distributions (void in the case of FreeBSD, for Linux
|> I would recommend Debian), and which network services to run
|> (ftp,email,http,telnet,ping?) 
|> 
|> Since all of this depends heavily on hardware and network speed, the
|> only thing that would make sense is to run the test twice on the same
|> machine, switching OS's in between. Also, the test would have to be
|> repeated several times because of the randomness involved.
|> 
|> What do people think?

I think that suggestion makes sense.  I however'd like to see that
project as a development tool that helps to ensure continued stability.
Also bug that are in one OS tend to tunnel into the other's code, so
including a as many tests of old and current bugs would be nice.

  Ralf