[TUHS] Re: Has this been discussed on-list? How Unix changed Software.

steve jenkin wrote in <DF4161CD-6FFC-4957-9946-AEA1190B3DFA(a)canb.auug.org.au>: |Doug, Larry et al, | |Thanks very much for the history - unaware of those stories/ facts. |I’ve scanned the 1989 Miller et al paper, will read properly soon. |The legacy of that paper is the extensive automatic testing now commonpl\ |ace in large Open Software projects. The "overly misused automatic testing" in my opinion. I have seen commit hooks etc which start test builds and test runs for any change, whatever it may be. (Like adding a period in a manual, so it cannot have beeen a Google clang one, heh.) Like going to the fantastic Coverity.com for which' advertising i do not even get money (stupid left wing..). And then there are thirty years and more in between. And isn't it ISO 9001 who requires extensive testing?(??) You also have extensive testing (i hope) for nuclear and airplanes and what. Question is: why is wc(1) not implemented to use three distinct logical code paths, and then choose the result which is delivered by the majority of them? There are plenty of different algorithms for wc(1), i personally am still stunned by the variant that i think Ken Thompson implemented (!?), and that can still be found in plan9port of Russ Cox (and more), as well as the 9front successor of Plan9 as such. I would never have thought this, but i would never have thought Aho-Corasick for fgrep either. And then i think you have to see that by then relatively few people were working on the absolute base level with quite restricted possibilities in power. And word was received that they simply reported bugs they encountered during their daily work, so that "when Dennis came in the evening he would find them, and fix it overnight". More or less. They had a job to do, and they wrote the tools to do that. This was science, exploration, desire, and, hm, military needs, i would say. So whereas i think what you said is totally right, i also think it is not. I also would not explain it with business, as bugs are costly. I think it is simply natural continuation of a thing, which gets more and more sophisticated and masterly the longer it is done. Now whereas one might (good) testing is also that, easy re-verification of a status quo helps even the masterly experienced. Having said _that_, Plan9 did not have any tests i think... But go(1) has quite a lot, and shares many heads. --steffen | |Der Kragenbaer, The moon bear, |der holt sich munter he cheerfully and one by one |einen nach dem anderen runter wa.ks himself off |(By Robert Gernhardt)