I'm thinking that the restriction on vector->range ("it is an error to
mutate <vector>") is actually a bad abstraction leak.  Not being able to
mutate a vector after passing it to this procedure is unfortunate, as
mutation is a fundamental operation on vectors.  I wonder if this
means that the only real use of vector->range will be as a sort of
shorthand for constructing a discrete range, by passing literal
vectors.  If so, perhaps it would be better to *remove* the
restriction on vector->range (i.e.  copy the vector argument), and
provide a discrete range constructor (conventionally, this would be
called `range'):

    (range* 1 2 3 5 7) == (vector->range #(1 2 3 5 7))

This way, the immutable vector underlying the new range isn't exposed,
and it's still possible to convert an existing vector to a range
(without the onerous restriction).

(Perhaps Marc has already suggested this, and I've just overlooked it.
If so, apologies.)

--
Wolfgang Corcoran-Mathe  <xxxxxx@sigwinch.xyz>

"If one is searching for a needle in a haystack, look in the part
of the haystack that contains more needles." --Bird & Wadler