On Mon, May 23, 2016 at 1:30 PM, William D Clinger <xxxxxx@ccs.neu.edu> wrote:
>
> Part of that performance advantage comes from hiding all
> manipulation of cursor-like things behind the abstraction,
> where they can be implemented in terms of low-level
> operations for which performant systems will generate
> inline code.

But cursors _are_ an abstraction.  An implementation
could in fact simply equate cursors with indexes, and
inline all operations, taking advantage of the range of
cursors to always use fixnum operations.  This has
zero overhead, and in fact the potential to be faster
than raw indexes.

--
Alex