Talk:N-version programming

From Wikipedia, the free encyclopedia

N-version programming simply doesn't work. Programmers generally tend to make the same mistakes and assumptions, generally interpreting design requirements and specifications in the same ways. The concept of N-version programming was faulty to begin with. FredricRice 17:12, 27 June 2007 (UTC)flr

N-Version Programming is only as good as the initial spec. If the initial spec is flawed in one or more ways, both designs will likely reflect these flaws. If the designers aren't aware they are developing to a flawed spec, then there will likely be a political struggle for a time until the design flaws in the initial spec are uncovered. In this way, it is actually more detrimental that the groups don't communicate more with each other and with the initial design person(s). What N-Version programming will correct for, though, is for the scenario where one team has a very junior programmer who codes lots of logical errors into one or more algorithms. The inferior program will likely be apparent on comparison of the two final products, provided the comparison isn't being run by the person who caused the errors... again, a hard thing to know in advance. If you knew who wouldn't be good enough to code a system up front, then why would you involve them in the first place? All these are issues that are easy to discuss, but very hard to definitively solve.Fine Arts 16:19, 12 July 2007 (UTC)

It seems that one of the critical parts of N-version programming was missed: the groups must be extremely diverse. (See [1]). That post explains it far better than I could.

Thequux 18:17, 9 August 2007 (UTC)