[411111111111111]

That question can't be answered without knowing what kind of programs you're writing. You got a lot of options if you're thinking about websites/apis, almost none if it's machine learning and basically everything is better if you're thinking about GUI applications.

[Dove]

You know, I hadn't thought about it that way before. Perhaps the dream of a general purpose language was always incoherent. Perhaps the mess you get when perl tries to offer options to do everything well, and the suffocation you get when python tries to do everything the same way, are inevitable. Maybe using one language for small and large projects was never going to end well -- maybe either the language or the programmers were always going to do it wrong.

I suppose I have been assuming that the best policy is to do as the Romans for a long time now. Windows apps are C++ (or is it C# now?) because that's how things are over there. Android is Java, or Scala if you must -- at any rate, you hardly get to pick. Iphone requires learning Objective C. Jquery for frontend, or whatever the cool kids are doing; PHP on the back is good enough for most of the web, and it's good enough for you. And as you say, in ML, Python is not optional.

And as I think about it, I've been using specialized languages in specialized contexts for a long time. It seems the knowing the underlying language is only ever part of the problem -- you have to learn the specialized language for what you're doing, whether it's GL for graphics or your web framework's way of doing things, or MFC or .NET, or . . .

Yeah. Small tools for specific purposes, and follow the local ecosystem. Things have been going that way, and it's probably a good answer.

When in Rome, then. I like that answer. Thanks.