so SRFI 14's sample implementation is "right", not the formal
specification. That it is a typo in the formal specification can be
seen from the examples following the description of
char-set-unfold(!). They also use the "right" order.
In addition to the evidence already pesented, I did a search of the various SRFIs with unfold procedures, and the rest all have the "p f g" (or "stop? mapper successor") order. I therefore propose the following PFN:
"As originally written, this SRFI placed the stop predicate P as the second argument of <b>char-set-unfold</b> and <b>char-set-unfold-right</b>, and the mapper procedure F as the first argument. However, the examples, the sample implementation, and the unfold predicates in SRFI 1 and other SRFIs with unfold procedures all agree on placing P as the first argument. Implementers are therefore urged to treat the formal specification as erroneous and P-first as the correct order of arguments. This will happen automatically by using the unchanged sample implementation."