OCaml and its undying fans [pdf](cambridge.org)
cambridge.org
OCaml and its undying fans [pdf]
https://www.cambridge.org/core/services/aop-cambridge-core/content/view/02F18023B4C43BF6E53512AA7062A9A5/S095679680800676Xa.pdf/caml-trading-experiences-with-functional-programming-on-wall-street.pdf
2 comments
I assume by 'concurrent GC' you mean multicore parallelism in the same process, in which case, yes, OCaml has it now: https://discuss.ocaml.org/t/ocaml-5-0-0-is-out/10974
And Jane Street have long since moved forward into even more advanced territory. They are now trying to add Rust-like ownership guarantees: https://blog.janestreet.com/oxidizing-ocaml-ownership/
And Jane Street have long since moved forward into even more advanced territory. They are now trying to add Rust-like ownership guarantees: https://blog.janestreet.com/oxidizing-ocaml-ownership/
However, the authors point out some key weaknesses of Ocaml: for example, it lacks a concurrent GC, its optimiser isn’t very good, etc.
This paper, though, was written in 2008. I wanted to ask Ocaml engineers here on HN for a 2024 update on the state of Ocaml. Do these negative aspects of the language still exist? And if so, how do you get around them?
I’ve noticed a lot of HFT funds, market makers, etc. all migrating to Rust.. which obviously doesn’t have the above issues (who needs a concurrent GC when you don’t even have a GC).. but Jane is sticking with Ocaml afaik. I wanted to learn more about that decision.
Thanks!