To follow up, here are clarified/corrected paragraphs describing
COND-IMPLEMENTS, relative to the editors' suggestion for revising SRFI
0:

Definition:

   If one the clauses has a satisfiable implementation requirement,
   COND-IMPLEMENTS must expand into the body of a clause with a
   satisfiable requirement. (I.e., it must not expand into the ELSE
   clause in that case.) It is unspecified into which body the
   COND-IMPLEMENTS form expands if several clauses have satisfiable
   requirements.

Example:

   To demonstrate the utility of the conditional construct, consider the
   following example:

   The programmer is implementing some abstraction which can use function
   aaa from SRFI a or can use function bbb from SRFI b.

   If the programmer does not care which implementation her code uses,
   she can just write:

    (cond-implements
       (srfi-a ... aaa ...)
       (srfi-b ... bbb ...))

   In this case, the Scheme system, if it implements both SRFI a and SRFI
   b, will pick a branch based on its own preference, supposedly the more
   efficient or easier-to-load implementation.

   It may also be that the programmer has a preference for SRFI a,
   maybe because the program can provide more functionality in this
   case. Only if SRFI a is unavailable, she wants SRFI b. Here is the
   code:

    (cond-implements
       (srfi-a ... aaa ...)
       (else
         (cond-implements
           (srfi-b ... bbb ...))))

   This works because COND-IMPLEMENTS must choose a non-ELSE branch if
   possible.

--
Cheers =8-} Mike