Alright, I give up.  I'm still not exactly sure why my implementation
is as slow as it is; I have a hunch that I'm taking too long
processing the list of actual parameters, but I haven't been able to
glean many specifics from statprof.  Marc also employs some nifty but
un-Scheme-y (to my mind, at least) tricks that give his version an
edge, like emitting code that destructively modifies the argument list
as part of determining where to insert default values.

At any rate, find my version attached.  I think it's probably a dead
end in terms of going forward, but maybe it's salvageable by a more
experienced Schemer? ...Or maybe a more experienced Schemer could make
another attempt at doing an implementation from scratch?  I don't know
what the right course is -- I think it would probably be easy to do
something performant in C, but I also grok why that's not the
preferred solution.


Regards,
Julian