5 ms·

Well the article touches some good points, since it made me think... The problem is the language and paradigms used not the framework or how many levels you ab

• kabdib 13y ago
  Whether you like it or not, your world runs on C, C++ and assembly. These languages form the ninety-percent-plus core of modern computing's foundation. Lift the hood of nearly any embedded system, BIOS or OS and you'll find these in heavy use. You'll find C code in the networking layers that let you talk to the world, and in the very light switches that let you go to the bathroom at night. C runs your dishwasher, your car, your elevators, and probably your toothbrush. Then a lot of stuff gets layered on top of this. And sure, languages /do/ jit, but these are generally hothouse flowers, surrounded by an infrastructure provided by C, C++ and assembly. Few systems are native boot without involving a bunch of C. I'm happy if people are satisfied to work in the layers above all this stuff. Frankly, not many folks (as a percentage of the programming population) can do good kernel level work. But don't pretend that it doesn't exist, or that it is somehow morally inferior to hacking away in Haskell. Maybe this will change in thirty years; I think that's the time scale required to make a fundamental change in the way we program modern systems. I'd /love/ to see a native Erlang system, soup to nuts. But there's little economic incentive to make one, given that the lower layers are actually doing a pretty decent job.
  • Chris_Newton 13y ago
    Whether you like it or not, your world runs on C, C++ and assembly. These languages form the ninety-percent-plus core of modern computing's foundation. Isn’t that part of the problem? There is no technical reason we couldn’t have a language that offered the same fine control and hardware integration as C, compiled to native executable code in a similar way, but was both safer and more expressive. There is no advantage in having an awkward syntax for specifying types or in making all pointers nullable. Mainstream industrial languages today are a triumph of good enough, and they continue to dominate primarily because of momentum and the size of the surrounding ecosystem rather than technical merit in the language itself. Unfortunately, this creates a vicious circle that reinforces the status quo, and the few organisations with sufficient resources to break that cycle have limited economic incentive to do so.
    • ippisl 13y ago
      At least in embedded systems, this looks like something that can be done the open source way. The success of the arduino and the mbed in building popular ecosystems has shown us that. I can see some better version of the .net micro framework, being improved by the open source community(1), and benefiting from the ms ecosystems and tools, becoming fit for some large part of embedded systems work. Also I can see this getting adopted since it let's embedded developers in small companies, who has some power on which tools to use, learn another skill that can better their employment opportunities. (1)improved speed could be achieved through using the cito compiler project, which might be good enough for many projects. And for making the language hard real time, it could be done using specific implementation of reference counting, although at some speed/memory cost.
    • kabdib 13y ago
      There have been many contenders, notably Eiffel, Oberon and D. There are many others that I don't immediately remember the names of. These languages have /great/ technical merit. They offer safety (in various forms) and other interesting technologies that C definitely lacks. They were lauded by academics and industry pundits. So why didn't they take the industry by storm? Perceived technical merit is a terrible way to choose a language. Pascal was widely regarded as a great language, a wonderful model, and it was widely used in the 80s by various large companies. Today it is mostly dead. I believe this is because Pascal only did an adequate job of expressing stuff at the hardware and kernel level, and that C was better. Certainly nearly everyone at Apple that I worked with breathed a sigh of relief when it became obvious that it was okay to write C instead of Pascal for new projects. For the most part we'd been writing C for years anyway, just in Pascal. About the only thing that people missed were nested procedures (whereupon, C++). Your new "adult" language is going to need a set of very compelling offerings over and above "well, it's safer" in order to succeed. Take a look at things people are doing /to/ C in order to be better: - "analyze" builds that do control graph analysis and find bugs (not just ones endemic to C, but actual logic bugs, too) - declarative sugar that helps tools to reason about what things like drivers are trying to do - control extensions (commonly seen as macros providing 'foreach' like support) - ways for tools to enforce local conventions (without spending tons of manpower on parsers and so forth) Come up with a language as good as C at low-level programming, that has great debugger support, offers easy tool plugins, and that has interoperability with the gazillions of libraries already available [take a page from C#'s great interop story here], and you might have something. Go "academic" and just say "this is good for you, use it instead," and the working programmers will see nothing in it for them and ignore you, just as they've ignored or abandoned dozens of other offerings in the last 30 years.