News for hackers. Stuff that matters.
Re:How much your computer can do in a second by nostrademons on Sunday March 26, 2017 @07:47PM (#13964083)
For people who have an exceptionally high tolerance for learning what people actually care about and why the shitty technological solutions won in the marketplace, this is probably an opportunity.

When I entered the workforce in 2005 after learning all about bygone languages (Common Lisp, Smalltalk, Dylan, Haskell, Erlang, Ocaml, etc.) in college, it was really, really painful going back to Java. Java had gotten so many fundamental language features wrong when better alternatives were well-known in research literature. Monitors instead of CSP, null types instead of Maybe, inner classes instead of closures, manifest typing instead of type inference, no first-class functions, mutability everywhere, conflation of modules & classes, not even a semblance of pattern-matching. But I could use it to write software that performed well, that could be deployed with tooling that everybody used, that had a wide base of available libraries, that had IDEs and other tooling available, and that other programmers could maintain.

It's 12 years later, and all of my recent projects have been in some combination of Swift, Rust, Kotlin, or ES6+React. And most of these languages have: promises & message-passing instead of explicit locks; Optional types or at least some form of null-safety; lexical closures; type inference; first-class functions; immutable-by-default; module systems that don't require that you shove everything into a class; and pattern-matching. But they also did the hard work of figuring out an interoperability story with some large existing body of code; of writing decent package managers and easy build systems; of making the language reasonably performant for most everyday tasks; and of (in Swift & Kotlin's case) providing full support within an IDE. All the language features that were common in research languages of 2005 but missing in industrial programming of the time are now very usable in the industrial programming of now.

I wouldn't be surprised to see a similar renaissance in tooling over the next decade, with exotic features like time-traveling debuggers, fully-inspectable systems, and migratable processes becoming standard in the next generation of IDEs.

There are no comments on this comment yet.