Regarding refs, lists and arrays behave similar, but not the same.
There seems to be a possibility for further unification.
This works for both:
(setq l (sequence 1 10)
a (array 10 l))
;;
;; Applying refs works for both lists and arrays:
(setq r (ref 5 l))
(l r)
(a r)
But getting a ref does only work for lists (see above); this does not work:
> (ref 5 a)
ERR: list expected in function ref : (1 2 3 4 5 6 7 8 9 10)
What about having
Then this would work, too:
(setq r (ref 5 a))
(l r)
(a r)
The semantics of an array ref could/should be the same as for list refs.
Such unification would be good for functions operating on containers - being agnostic against the array/list difference.
Because there already is setf for arrays, I'm assuming that most machinery to get this done already exists.
In the big picture of things arrays are used rarely in newLISP, there are more a speed optimization for specific situations then anything else. I did it for apply and map because it was extremely cheap to implement, and like for some of the statistical functions, because there was a moderately, frequent use-case for them - at least in my work - and an easy way to share additional code required among those functions.
Doing it for ref, ref-all, set-ref, set-ref-all is much more involved and I just don't see people using it that much. Most of the time when I see people using arrays, I think, a list would have been as good or better. Like most requests made in this forum, it goes on a list and could be re-evaluated in the future.