On an off-topic and statistical note, I'd like to say that this data is fairly meaningless in a statistical sense. 2 out of 1 million test cases were pathological enough to skew a statistical analysis. This is not surprising. However, I don't think that the kernel logic contains any pathological cases. And as for the 500 cases where 2 or more tests fail, a quick C writeup shows that if 27 independant processes are choosing 1% of 1 millon tests to fail on, there should be closer to 27,000 tests where 2 fail. If the rate of fail is 0.1% then it is at about 300 tests where 2 fail. This with a completely random distribution. Lies, damned lies, and statistics. Sorry, tis a slow morning for me. Easily distracted. -Titus > Date: Mon, 21 Jun 1999 09:28:23 -0700 > From: Edward Berner <bernereat_private> > Subject: Diversity > To: BUGTRAQat_private > > On the subject of diversity and reliability, I found the following in RISKS 3.41: > Q11: True or False? Computer programs prepared independently from the same > specification will fail independently. > > A11: False. In one experiment, 27 independently-prepared versions, each > with reliability of more than 99%, were subjected to one million > test cases. There were over 500 instances of two versions failing > on the same test case. There were two test cases in which 8 of the > 27 versions failed. (Knight, Leveson and StJean, "A Large-Scale > Experiment in N-Version Programming," Fault-Tolerant Computing > Systems Conference 15) > > RISKS 3.41 can be had at the following URL: > http://catless.ncl.ac.uk/Risks/3.41.html > > > _______________________________________________ > linux-security-module mailing list > linux-security-moduleat_private > http://mail.wirex.com/mailman/listinfo/linux-security-module > _______________________________________________ linux-security-module mailing list linux-security-moduleat_private http://mail.wirex.com/mailman/listinfo/linux-security-module
This archive was generated by hypermail 2b30 : Mon Jun 04 2001 - 10:15:37 PDT