To be honest I don't think modern languages are all they are cracked up to be. Back in the day, very large and complex systems were managed quite well with languages like QBASIC, COBOL and FORTRAN, and ran rock solid for decades (hell, a few are still operational). But today, programming is such a polyglotic mess and conceptual navel-gazing nightmare -- it really is no wonder that so many projects end in failure.
As an outsider looking in: programmers, fix thy tools.
My first introduction to programming was QBasic in the 90s, and I had a great time making little pictures, beeps/bloops, and text on the screen. I then moved on to some basic Fortran and other engineering-type languages in undergrad c. 2006. My impression of the workflows for Fortran at that time was "this is more annoying than BASIC, but I guess its because its more powerful and complex. I bet they'll fix the awkwardness eventually though."
Imagine my supersize when I decide to pick up the trendy modern language of Haskell in 2015 and find out that installing, configuring, and working in it was just as bad as Fortran back in 2006. WTF has everybody been doing for the past 10-20 years?
It seems to me that because programming is a new discipline, lots of people are trying to reinvent wheels. There is nothing wrong with this, in and of itself, since we'll essentially be making sure that we get really good at making wheels and end up with really high quality specimens. The problem is that wheels by themselves aren't that useful, and no one is bothering to make an automobile.
QBasic might have had square wheels, but at least they felt attached to something.
> It seems to me that because programming is a new discipline, lots of people are trying to reinvent wheels. There is nothing wrong with this, in and of itself, since we'll essentially be making sure that we get really good at making wheels and end up with really high quality specimens. The problem is that wheels by themselves aren't that useful, and no one is bothering to make an automobile.
There's a lot of truth to that. People are blazing forward, without even looking at the past that's beyond the most recent trendy languages and frameworks. End result is that there's a lot of wheels reinvented, but poorly.
Learning Lisp was a particular eye opener for me. It was a gateway drug to discovering that there were so many things, so better thought out than what we have now (maybe because people did actually think about what they're doing back then, instead of releasing glorified CRUD apps "early and often"), and we're only sometimes slowly reinventing them.
> WTF has everybody been doing for the past 10-20 years?
They have mostly been touching and stroking themselves by inventing a new programming language every 15 minutes, rather than sitting their asses down for ten years, mastering one language properly, and then squeezing the living daylights out of it in performance, simplicity and efficiency... nope, that requires actual work, it's much cooler to invent yet another way of doing the exact same thing: wheee! Look at me, I invented yet another language!!!
Fortran is still numero uno when it comes to finite element analysis and high performance computing industries: not only does the Formula TRanslator language lend itself particularly well to expressing physics and mathematical equations needed for doing finite element analysis, but intel and PGI ship ultramodern Fortran compilers which generate very fast machine code, faster than even optimizing C compilers. intel ships their Fortran compiler with support and optimizations for their latest processor features and instruction sets (for instance, AVX), and PGI pioneered a Fortran compiler which can compile Fortran directly into NVidia GPU code.
Beware of survivor bias. Most old projects failed, like most new projects.
But why is it that the rate of failure in programming projects hasn't fallen?
I've heard it said that the computing world in general operates on the edge of acceptability. As soon as the browser became fast enough to become an applications platform, we implemented the browser in the browser; as soon as good command-line interfaces develop, we replace them with bad windowing ones; as soon as dumb phones become a mature technology, we all move to highly experimental smartphones.
This is probably most severe in programming itself, where our field's most free to make its own choices. The percentage of projects that fail remains more or less constant as our level of experience grows -- because, as we become more experienced, we move to more complicated models. Any modern programming shop could churn out successful COBOL waterfall projects at a rate of two or three a year; so, as we got good at waterfall and old languages, we all moved to Agile and OO. Then, once we developed dangerous amounts of expertise in Agile and OOP, we moved to the Web (and lost most of our ability to use IDEs). And now, with even Web programming in danger of becoming a mature discipline, I'm hearing a lot of talk about Haskell.
Why does this happen? I don't know -- because I don't feel the temptation to do it. I program in C++, and I carry a feature phone...
