> For a long time neither the head nor the latest release would even build,
> and I think people gave up on looking at it.

Ah, that explains. I must be new here.

> FFI in any form will not be part of R7RS-large by chair fiat; he committee
> voted it down in the earliest days, and I am committed not to bringing it
> back.  There are just too many differences, even apart from Schemes running
> on JVM or CLR.

That's probably a wise decision at the language spec level.

> Note that r6rs-pffi is just a wrapper over the FFIs of some R6RS systems.

Yeah. I've been thinking there should be a common FFI that solves the
80% case. There's no end to how fancy you can get if you really want to
exhaust all the possibilities, but how often is there a point to that?

The elephant in the room is that C libraries are still written to be
consumed primarily from C/C++ and only incidentally from higher-level
languages, even though most applications should be written in
higher-level languages now.

> Loko is
> probably too new, though it looks cool: it can run on bare metal as well as
> Linux

Even the earliest release did that! IIRC it already runs some loops
faster than Chez. It has the cp0 compilation pass known from Chez,
Ikarus and Vicare. Also there's no C code in the implementation: it's
all Scheme. Göran has a ton of Scheme/low-level experience and writes
impressive code really quickly, everything he makes is worth a look.