Path: utzoo!utgpu!news-server.csri.toronto.edu!rutgers!mcnc!duke!crm
From: crm@duke.cs.duke.edu (Charlie Martin)
Newsgroups: comp.software-eng
Subject: Re: Proofs (was: Global program state.)
Message-ID: <663805031@macbeth.cs.duke.edu>
Date: 13 Jan 91 22:17:13 GMT
References: <1070@tetrauk.UUCP> <1987@oravax.UUCP> <1991Jan11.175952.10978@pdn.paradyne.com>
Organization: Duke University Computer Science Dept.; Durham, N.C.
Lines: 24

In article <1991Jan11.175952.10978@pdn.paradyne.com> locke@nike.paradyne.com (Richard Locke) writes:
>[not redirection from discussion in comp.object]
>
>Capers Jones makes the interesting statement that there isn't any
>empirical evidence that correctness proofs are an effective defect
>removal or defect prevention strategy (in the context of "real"
>product development efforts).
>
>Does anyone have evidence to dispute this?
>

Yes -- in fact, all the empirical evidence of which I am aware directly
contradicts this.  See, e.g. the papers on the Cleanroom methods, in
which Harlan Mills documents about an order of magnitude improvement in
defect rates at a 30 percent GAIN in productivity.  My own experience
suggests that somewhat different methods give perhaps 80 times
improvement in defect rates.

I'd like Mr Jones to quote the studies that didn't show an improvement.

-- 
Charlie Martin (...!mcnc!duke!crm, crm@summanulla.mc.duke.edu)
O: NBSR/One University Place/Suite 250/Durham, NC  27707/919-490-1966
H: 13 Gorham Place/Durham, NC 27705/919-383-2256