I appreciate Russell's defense of the gh_ interface. A more honest
statement is that I am not a language implementor at all, and my first
stab at gh_ was reflected a lot of ignorance: it was just what I
wanted for my own application :-).
I got many comments from Mikael and others, and have cleaned it up
over some iterations. It needs another pass for sure. Since it is
mostly documented in its own chapter of the reference manual, I invite
someone to propose stuff. Keep in mind that the ice-9 loading issue
is addressed by Maciej's (spelling?) proposals.
I do not address smobs because I am quite happy to work with R4RS
scheme. Still, I think there should be gh_ hooks for a type system,
since most people seem to disagree with me.
But I think that Alexander's comment on gh_ is reasonably fair because
he was talking about version 1.2. He might tone it down a bit with
the pre-1.3 releases.
jimb has already addressed (and broadly agreed with) most of
Alexander's other points.