Hacker News Reader

Comment by dgreensp | Hacker News Reader

There's no clear value proposition given. What are we sacrificing Turing-completeness in order to gain? For example, one thing a programming language might gain by sacrificing Turing-completeness is that every program provably halts; no infinite loops! This is called "total function programming." There are probably other interesting properties you can gain by sacrificing Turing-completeness as well.

Crema can restrict the computational complexity of the program to the minimum needed to improve security

"Computational complexity" is a technical term that is related to performance, but not security.

View on HN

> There are probably other interesting properties you can gain by sacrificing Turing-completeness as well.

Not the author, but I wrote another sub-Turing language. In my opinion, there's two useful properties: a) type inference -- you can infer the type of the whole program b) you can infer memory use and thus avoid the need for garbage collection or manual memory management. Both properties are very useful for performance, of course. :)

Hi, I'm writing a total functional language as well. I think that garbage collection is still useful however, as it can collect garbage as the program runs, not just at the end.

I've managed to make it so my memory use is completely deterministic -- no garbage at all.

That sounds interesting. Any references or notes on that I can read? (or source code :) )

I would say that just because your memory use is deterministic, doesn't mean you don't get garbage. Garbage is just memory that isn't needed any more.

for example, in

  let
    a = f(x)
    b = g(y)
  in
   a + b

If there is working memory used in the body of f(), it will be garbage when f() has finished executing, because it won't be referred to any more. So it could be freed after f is executed and before g is executed, which may reduce the maximum amount of memory required by the program.

From the Crema wiki that I just glanced at, it is indeed not possible to write infinite loops:

"The only type of loop supported by Crema is the foreach loop. This loop is used to sequentially iterate over each element of a list. Crema specifically does not support other common loops (such as while or for loops) so that Crema programs are forced to operate in sub-Turing Complete space. In other words, foreach loops always have a defined upper bound on the number of iterations possible during the loop's execution. This makes it impossible to write a loop that could execute for an arbitrary length of time."

This is called "total function programming."

Thanks for bringing it up, I was not aware of this: https://en.wikipedia.org/wiki/Total_functional_programming

Recently I had a need to design a small language for a constrained environment, and I spent a bit of time worrying about termination problem. Now I know where to look for more research.

Well, from my understanding, I would say "security" is a very broad term, being influenced by a broad range of factors including "computational complexity", don't you think?

If all you're doing is sacrificing the possibility for infinite loops, just hack a c++ runtime that kills a process after 24 hours, and blammo non-Turing complete.

Reply on news.ycombinator.com