Just a note of support for the general idea of incorporating Taylor's extension.
1) Selfish: I'd probably write more scheme code if there was a decent (= slime :) development environment 2) It seems like it is only gently touching the existing code 3) My opinion: More implementations have, in this project, led to better abstraction, which in turn has benefited all the implementations. I don't see why the the scheme backend won't lead to the same type of improvements.
Caveat: I don't understand the full implications as deeply as Helmut does - e.g. I wouldn't have picked up the cc/repl issue (until it happened to me). However it seems like a good start that his issue has already been addressed with a patch. I'm interested in seeing more substantive criticism.
-Alan