John, thanks for your comments! (And please excuse my late reply.)
1) Do continuation marks make it possible to provide delimited continuations? If so, they should be included at least in the examples and maybe in the spec as well, so as to have only one non-portable SRFI rather than two.
Alex Shinn wrote on the Chibi Scheme mailing list that he planned to implement delimited continuations and delimited dynamic bindings built on continuation marks.
Continuation marks may be helpful here but I don't know whether delimited continuations can be solely built upon continuation marks.
It seems that delimited continuation are implementable in a portable way, at least sort of; see Taylor Campbell's implementation:
http://mumble.net/~campbell/scheme/shift-reset.scm. I am saying "sort of" because such a implementation may leak space by capturing too much of a continuation.
2) You should spell out the relationship between the Racket and SRFI 157 APIs.
Added a note on compatibility with the Racket implementation.
3) Inferior Scheme is a good thing to have for doing non-portable SRFIs. I think it should be removed from this SRFI and promoted to a separate repository. However, it needs comments to explain how it does what it does (both "interpret Scheme" and "provide continuation marks"). That way other SRFI authors can extend it as needed. By tagging a specific release in the Inferior Scheme repo as associated with this SRFI, people can see it the way it is now when understanding SRFI 157 while allowing continued evolution.
I agree that "Inferior Scheme" should be developed into a general usable library/program that provides the full of R7RS and that is pluggable so that otherwise unportable SRFIs can be implemented with it.
I will do this but not in time when SRFI 157 shall be finished. So I will leave the current implementation of SRFI 157 as is (which can later be switched when a general usable "Inferior Scheme" is ready).
--
Marc