Porting 58000 lines of D and C++ to jai, Part 0: Why and How

Re:Porting 58000 lines of D and C++ to jai, Part 0: Why and How by cyber_kinetist on Wednesday November 23, 2022 @02:08PM (#33723302)

I think his disdain towards academia isn't really anti-intellectualism (he's not ignorant of compiler theory knowledge, otherwise he wouldn't have been able to write a compiler from scratch in the first place! And his interest in programming languages seems to span decades, from what I remember from one of his streams). I think he's critical of the fact that the primary focus of academia and industry in computer science have diverged so much over the years, such that most academic CS works haven't really helped developers in building better programs and tools. Thinking about the countless hours compiler academics have spent on all sorts of esoteric parser theory in the past, while all the production compilers today are using recursive decent for pragmatic reasons... maybe this kinda makes sense.

[ Parent | on HN ]

Re:Porting 58000 lines of D and C++ to jai, Part 0: Why and How by klodolph on Wednesday November 23, 2022 @02:41PM (#33723613)

I’d have to say that academic research into programming languages has provided immense benefits for actual programmers over the years, it just has such a long lead time that people forget that the research came out of academia in the first place. People also don’t really understand the process of how these ideas come from academia into mainstream programming languages.

It only seems like academic research has diverged because the benefits of current research haven’t materialized as real, usable features in programming languages that people use to get work done. But if you look at features that we use in day-to-day programming right now, you can trace the heritage of these features back to research programming languages (like Self or Haskell) and then farther back to more abstract research into esoteric subjects like category theory and substructural logic.

The esoteric parsers that people invented in the past were, in a sense, necessary because people were ignorant of how to design languages in such a way that they could support a rich syntax without using a complicated parser. It took a lot of academic research for us to figure out that, say, you could probably use an LALR parser for lots of existing languages, and you could stick to LL(1) for new designs.

[ Parent | on HN ]

Re:Porting 58000 lines of D and C++ to jai, Part 0: Why and How by cyber_kinetist on Wednesday November 23, 2022 @03:17PM (#33724010)

That's a fair point. Maybe researchers needed to delve into all sorts of weird parser theories, to later figure out that they weren't that needed. Practicioneers mainly know the final result, but aren't usually interested in the full history of how people got to that conclusion.

But I think during the recent decades, there has been a consensus that perverse incentives in academia are degrading research quality and preventing papers from becoming actually usable in real-life applications (mainly with the focus of paper metrics and the constant need to apply for grants). So although I think academia is still important long-term, I understand why some people would think it's becoming less useful.

[ Parent | on HN ]

Re:Porting 58000 lines of D and C++ to jai, Part 0: Why and How by tialaramex on Wednesday November 23, 2022 @09:15PM (#33727140)

It looks exactly like anti-intellectualism to me.

Actually that's probably under-selling the problem. It's not just that Jon doesn't value the opinions of academics, he doesn't really value the opinions of anyone except Jonathan Blow.

Which is probably a healthy way to attack your first video game project, it's not as though Braid would be more likely to be a success if Blow stopped believing in it himself. But I would be surprised if that's true of a programming language.

> most academic CS works haven't really helped developers in building better programs and tools

It's doubtless possible to measure "most works" and "really helped" in ways which allow you to either draw this conclusion, or not, as you prefer but I don't think that's a useful way to think about it at all.

[ Parent | on HN ]