On Wed, Apr 1, 2026 at 6:44 PM Per Bothner <xxxxxx@bothner.com> wrote:

> It appears to me that SRFI-268 solves no problems that SRFI 163 doesn't solve,
> but fails to solve some problems that SRFI-163 does solve. So what is the purpose?

SRFI 163 is based on lower bound and length rather than lower bond and
(exclusive) upper bound, like the rest of Scheme.  it also requires
the array-literal-header to be a single token, which is problematic
for high-dimensional arrays. As WP points out s.v. "Arrays": "A single
1000 x 1000 pixel color photo acts as a dataset with 3 million
dimensions (10^6 pixels. 3 color channels), where each pixel value is
a feature."  Also, IMO it is ugly.

I would be willing to simplify the notation to just "#Atag dimensions
datum", where "dimensions" is a list of either non-negative integers
representing the upper bound or 2-lists of non-negative integers
representing the lower and upper bounds, per Bradley's earlier email.
That would allow representing every shape of array.  (User-written
storage classes and generalized arrays would still be not
round-trippable.)