R4RS compliance: negative numbers

[William D Clinger <will@ccs.neu.edu>]

I wrote:

> I do not see why inexact numbers are any more misleading than
> encoding "2 or more" as -3.

Aubrey Jaffer wrote:

> I do.  R4RS compliant implementations are not required to support
> inexact numbers, but they are required to support negative numbers.

Four remarks:

  1.  I understand the argument that encoding "2 or more" as -3
      may be less likely to run afoul of an implementation
      restriction than encoding it as 2.0.  I do not understand
      the argument that encoding as -3 is less misleading than
      encoding as 2.0.

  2.  R4RS compliant implementations are not required to support
      negative numbers.  Furthermore an implementation may be
      able to represent inexact negative numbers but not exact
      negative numbers.  See R4RS 6.5.3.

  3.  This is 1996.  As a practical matter, we can assume that
      almost all implementations will support inexact numbers,
      just as we assume that almost all implementations will
      support negative numbers.

  4.  As evidence for remark #3, I observe that these very
      assumptions have been made tacitly by those people who,
      for the past several days in this mailing list, have been
      defending the honor of Scheme's generic arithmetic against
      Matthias Blume's usual ML-is-better view.

Will