| Date: Sun, 10 Apr 2005 23:23:05 -0700
 | From: Per Bothner <xxxxxx@bothner.com>
 |
 | Aubrey Jaffer wrote:
 | > My focus is to get multidimensional arrays incorporated into R6RS; and
 | > SRFIs are allegedly the way to do that.  R6RS will not incorporate
 | > both SRFI-25 and SRFI-63; so concerns about their interoperations is
 | > at most secondary for a standards track SRFI.
 |
 | Ok.  But don't expect expect at least my Scheme implementation to
 | put effort into implementing SRFI-63 - at least until we get a
 | preview of R6RS.

Fair enough.

 | (I do find the lack of openness in the R6RS process rather
 | unsuitable, FWIW.)
 |
 | > ... Yes it was.  It was the SRFI-25 authors who decided to be
 | > incompatible.  See
 | > http://srfi.schemers.org/srfi-25/mail-archive/msg00090.html
 |
 | Hm.  Ironically, it was I who pointed out the incompatibility.  But
 | nobody who was actually was using Bawden-arrays spoke up, it
 | appears.  And it wasn't just "SRFI-25 authors who decided to be
 | incompatible" - others supported that decision.  I was the only one
 | (? - I haven't check the entire discussion acrhive) to argue for
 | compatibility (as I do again), but using Bawden-arrays myself I
 | could hardly object too strongly.
 |
 | However, in the current situation I myself have implemented SRFI-25
 | arrays, so I *am* in a position to object.

SRFI-47 (which is upward compatible with SRFI-63) is implemented in
SCM, Guile, and SLIB.  So renamed or not, SRFI-63 would lack
interoperability with at least one existing, finalized SRFI.

 | ... What is the usage *today*?

SCM, Guile, and any Scheme using SLIB have Bawden arrays.  Every
recent Linux distribution includes Guile, so that should count for a
lot.

But this pissing contest should be largely irrelevant to R6RS --
SRFI-63 is more capable (uniform arrays), better integrated with R5RS
(specifying vector, string, and EQUAL? behavior), compatible with
SRFI-58 array syntax, and better designed than SRFI-25.