Comment by pron

But software isn't developed for its own sake. It's built to serve some purpose, and it's through its purpose(s) that the selection pressures work. It's like Betamax fans saying that people were wrong to want a longer recording time than better picture quality. It's not enough to say that you like some approach better or to even claim that some untaken path would yield a more desirable outcome. You need to show that it actually works in the real world, with all of its complex variables. For example, in the nineties I worked on safety-critical software in Ada, but we ended up dumping it in favour of C++. It's not because we didn't recognise Ada's advantages, but because, in addition to those advantages over C++, it also had some very significant disadvantages, and in the end C++ allowed us to do what we were supposed to do better. Ada's build times alone made it so that we could write and run fewer tests, which hurt the software correctness overall more than it helped. We also ended up spending more time understanding the intricacies of the language, leaving us less time to think about the algorithm.