[r6rs-discuss] Comparison procedures' number of arguments

[Ray Dillinger]

Just a note.... 

Comparison procedures returning #f (or throwing exceptions) for 
single and zero-arity cases would cause errors in code I've written 
that sorts directories of files into order based on their timestamps.  

The basic logic is, "while unsorted, do this."  If one-file and 
zero-file directories are never considered to be sorted, then the 
code goes all infinite-loopey. 

I've assumed that the predicates are true *unless* there is a pair 
in sequence that violates the condition, because that seems to me 
the obvious and only reasonable way for them to be defined.  My
personal principle of least surprise is such that I've been 
utterly amazed to hear people seriously arguing for different 
views. 

And this has made me go complicate and clutter my code (grr) to 
guard against the astonishing and heretofore-unconsidered possibility 
of language implementations where different views are in fact
implemented. 

                        Bear