1 Jan 1997 03:38
Re: Re: WaterFall [Was: XP implementation]
<cg <at> cdegroot.com>
1997-01-01 02:38:44 GMT
1997-01-01 02:38:44 GMT
jhrothjr <yahoogroups <at> jhrothjr.com> said: >I think you meant 'some of the time.' And of course they do. The 'big >bang' approach works when you manage to get the requirements right, and >they don't shift out from under you during the project's life. People who >say that it's "impossible" to get the requirements right are overstating >their case, as are people who say that one is "always" aiming at a >moving target. Indeed. I have worked in a couple of projects where the requirements were stable over many years. One was so big and complex (basically an overhaul of the German air control IT infrastructure, with many many many contractors) that changing requirements would take time only measurable on a geological timescale; the other one was basically a piece of software to help a pharmaceutical comply with all the rules and regulations of the FDA and other drug registration authorities - what the FDA wants to see tends to be quite stable over time, and therefore the data that needed to be captured and the way in which it was to be stored and displayed were almost completely known in advance (the only thing added 'on-the-fly' in that project were on-line connections to Perkin-Elmer chromatography computers; shudder). Both these projects were done in a BDUF manner; but in both these projects I had the role of fire fighter soon afterwards and in both I teared out large chunks of software and rewrote them from scratch. I think that even though theoretically these projects could have been satisfactorily completed with a classical waterfall-style model, I think that application of XP values would have helped a lot (it certainly would have prevented the double work I had to do, rewriting perfectly functioning code because it was impossible to touch it in order to strike a handful of defect reports off the list - too much spaghetti code, and I think that a waterfall model tends to 'not discourage' spaghetti code).(Continue reading)
>
That's probably why engineering-degreed middle managers love Java and C++ -
they know how to handle projects with long compile times 
RSS Feed