chalst writes:

I think FFIs in the scheme world are better, on the whole, than in the Common LISP world,

Despite being a CL hacker, I tend to agree; at least on Unix. It's an implementation issue not a language issue - the situation may well be different on the Mac (MCL by all accounts integrates very nicely in to MacOS) - and it was pretty certainly different on LispMs, where the OS was Lisp. CL on Unix seems to have inherited that "Lisp is the world" view to the point that it's more difficult than it need be to talk nicely to Unix. This annoys some Lispers more than others, and me appreciably - hence cirCLe

Of course, Unix doesn't necessarily make it easy to write performant FFIs to it either; despite CMUCL/SBCL having what's actually a very nice API for calling from Lisp to C - it lets you call any C function you can link in, specify types including structures and unions, and do automatic coercion between lisp types and C types - you soon find out that there's enough preprocessor macrology in glibc (e.g. errno and stat) that you end up having to write C wrappers for everything anyway.