> There are big fat warnings in the current draft under set-umask! and
> set-current-directory! that use in multi-threaded programs has
> unpredictable results.  I think that is the best we can do.

The main point of the current-directory conundrum is not the
`current-directory` procedure itself. It's all the other procedures.

If RnRS procedures take filenames relative to a per-thread
current-directory, but SRFI 170 takes them relative to the OS CWD, I'd
expect that to be quite confusing to users.