I certainly agree with Tom that it would be a shame if support for
SRFI-50 became de rigueur for Scheme implementations.  I think SRFI-50
essentially dictates second-rate interaction with system threads, and
constrains implementations too much in various other ways, as well.
So I think it's important to put together a SRFI describing a more
opaque interface.

But I don't think which SRFI comes first really matters.  People
writing C code will choose which FFI to use by looking at the FFI's
supported by the Scheme implementation they want to use, not by
evaluating the FFI specs themselves.  So the authors of the most
popular Scheme implementations are really the ones who will decide
which FFI SRFI becomes most popular.  That group is more likely to
carefully weigh the tradeoffs of the various interfaces.  If the
discussion here is representative, opacity will be of concern to them.

As long as SRFI's of both sorts become available within the next year,
the right thing will happen.  I don't think it's necessary for SRFI-50
to be withdrawn until a more opaque SRFI is written.

Like any other SRFI, an opaque FFI SRFI would benefit greatly from the
participation of the developers of widely-used Scheme implementations,
and other people with established reputations.