On Tue, Dec 3, 2013 at 8:16 PM, John Cowan <xxxxxx@mercury.ccil.org> wrote:

I've also made a new version of the SRFI itself. Most of the changes
are editorial, including a substantial reordering of the procedures
to minimize forward references. Among the few new facilities are the
ability to ask a comparator whether it implements comparison procedures
or hash functions or both.

Very clean and practical. My only question is about the decision to make comparison of NaNs have meaning in make-inexact-real-comparator. I don't have much experience with NaNs, but it seems risky to give meaning to comparing them. On the other hand, throwing an exception seems like overkill because callers can always check them for NaN-ness first.

Just thinking out loud.