Also, I've just noted that now (ice-9 nice-9) exports only 8 bindings, so there's surely a slot for defining some new ones.
Perhaps it would make sense to liberate letrec or letrec-syntax, so that it would be possible to use it in a manner similar to let and letrec. However, I don't think that using letrec in programs is ever a good taste. I'd personally prefer to use internal "define", because it is much more readable. As to letrec-syntax, there'd clearly be a little point in defining recursive macros that have only a single syntax-rule clause.
Some time ago [1], I presented a following form:
(publish
(define (f x) .... (g x) ....)
where
(define (g x)
....))
It is useful if one wants to define a few functions that are externally visible, that share a common internal definition, but it also improves readability when one defines even a single function that makes use of some internal definitions. For example, I dare to claim that the usage shown above is more readable than
(define (f x)
(define (g x)
....)
.... (g x) ....)
because the extent of internal definition is more apparent, and the body of external definition is more concise. This is especially visible when there are more internal definitions. If anyone's interested to see how the publish-heavy code looks like, it can be viewed in [2].
So anyway, if anyone thinks that it could make sense, then I could add this "publish" form to the (ice-9 nice-9) module, so that it would export exactly 9 macros.