Hacker News Reader

Comment by ska | Hacker News Reader

Was this basically what doomed IronPython also?

I remember a conversation ages ago about how you couldn't really get a (common?) lisp running properly, irrc due to limitations in the way CLR modeled classes amongst other things, but FFI came up there too.

View on HN

Yes. The problem was that the interpreters for both languages did not have a well-defined FFI. From a native library you could reach up into the interpreter and party over the AST to your heart's content. We couldn't in any practical sense emulate those data structures from the CLR which meant that native libraries couldn't run. That meant that existing scripting code that used those libraries would need to be rewritten to use an equivalent .NET library (e.g., regex). We naively thought that was what people would do and that turned out to be a mistake :(

What do you mean by well defined FFI? Why couldn't calls to native libraries be marshaled?

Iirc They did something like this for c++ and it wasn’t pretty but I can’t remember details, it may have been mostly because of the unstable ABI.

> That meant that existing scripting code that used those libraries would need to be rewritten to use an equivalent .NET library. We naively thought that was what people would do...

AFAIK, that's what JRuby does but Java had (has?) a bigger community and is more platform independent. Microsoft was a different company back then and wasn't supported on Linux. Betting on IronRuby/IronPython meant you had to use Windows as your platform.

Reply on news.ycombinator.com