Over the ASCII repertoire, case folding reduces to mapping to lower case, so you should use that. Unicode, but not ASCII, has many cases of a lower case character without an upper case equivalent, but not vice versa.
Only now realized to think of what ascii-ci<?, ascii-string-ci<? and the
rest do when comparing letters to non-letters. Equality and inequality
are easy enough, but how to determine whether a letter is less than or
greater than a non-letter?
In practice you probably either map lower-case to upper-case or vice
versa, right? The current SRFI text and implementation map lower-case to
upper-case, but is that right? Unicode has a standard case-folding
algorithm but I don't understand it.