> On Aug 12, 2016, at 11:13, D. Hugh Redelmeier via talk <[email protected]> > wrote: > <http://www.itworldcanada.com/blog/phoenix-payroll-report-by-michael-wernick-the-clerk-of-the-privy-council/385370>
Some of my thoughts: 1. I don't like the literary form of this article. Too much theatre. I prefer non-fiction, not info-fiction. Even the URL misleads. 2. (Why) was there no parallel test period? At least for a stratified sample. That might have shown the unanticipated problems (which should have been anticipated). 3. When a sufficiently complex system is replaced, it should not be assumed that the old system was completely or consistently correct. (CF: the allusions to interpretation differences.) Systems that operate for a long time tend to become corrupt. Errors, ambiguities, interpretations, may creep in. Some of these are never discovered. Some are discovered and corrected. Others are discovered and covered up -- the error is propagated forward to avoid having to deal with the consequences. A consistent replacement system for an inconsistent system will act differently, and efforts to retain inconsistencies may result in instabilities and failures. 4. The vendor shares in the responsibility for taking on a task that contained "unanticipated complexities". There should have been vendor-penalties for such failures. And a responsible vendor would do enough due diligence to anticipate "complexities" early. -- Peter --- Talk Mailing List [email protected] https://gtalug.org/mailman/listinfo/talk
