Replies: 1 comment
Fascinating discussion! Both have valid points. Torvalds is correct in saying that evolution, or directed evelution, or trial and error, are valid ways to end up with a product. However, it is a mistake to say that there have never been successes with requirements -> design -> build -> release. Example: new code for payroll system. The specs for a new w-2 form are pretty exact, you need a design, it is coded and tested, and released.
The determining factor may be the precision of the requirements and narrowness of the application. The more narrow and deterministic, the more likely that the "waterfall" process will result in a useful product. The less certain the requirements and ultimate use, the more likely that the "build, try, and re-build" process will work. Evolution is a feedback mechanism! If you need lots of feedback to sort out what's useful and what will work, it's hard to do this via the waterfall methodology. On the other hand, if its a well defined problem, the chances are better that you can plan your way to the solution.
Posted by mcw @ 12/05/2001 04:14 PM EST