Message-ID: <2016082181.22351.1417020046481.JavaMail.firstname.lastname@example.org> Subject: Exported From Confluence MIME-Version: 1.0 Content-Type: multipart/related; boundary="----=_Part_22350_272787897.1417020046480" ------=_Part_22350_272787897.1417020046480 Content-Type: text/html; charset=UTF-8 Content-Transfer-Encoding: quoted-printable Content-Location: file:///C:/exported.html
Rather than testing every possible combination of things, all pa= irs simplifies the exercise to testing every pair of things which reduces t= he complexity significantly, for example instead of 700,000 possible combin= ations, all pairs would be about 500 combinations.=20
While this reduces the number of tests, does it help find bugs? Al= l pairs works because when things break, they have a tendency to break beca= use of the faulty interaction of two things rather than 3 or more. &nb= sp; A long term study of medical device failures found a strong correl= ation for this. For all the failures reported, a quarter of the= m would have been found with all pairs testing. The true result is pr= obably much better than this, because a lot of the failure reports did not = have enough detail to allow proper analysis. Of the detailed reports,= 98% of the failures would have been found with all pairs testing! Th= e paper, "Failure Modes in Medical Devices", is at csrc.ncsi= .nist.gov/staff/kuhn/final-rqse.pdf=20
If you do all pairs testing, you could still use minimal pairs to = get better test effectivess. By starting your tests with all the mini= mal pairs first, this would give a good broad coverage of combinations.&nbs= p; The remaining all pairs combinations could then be used to finish t= he exercise.