Re: Arithmetic issues - feedback Andre van Tonder 18 Oct 2005 18:17 UTC

  John Cowan wrote:

 > Andre van Tonder scripsit:
 >
 > >  - Should a minimum precision be required for fixnums or flonums?
 > >
 > >    I would prefer not (principle of sufficient reason - von Leibniz ;-)
 >
 > Fixnums and flonums are *about* machine architectures.  It is not
 > clear that there is a sufficient reason for the machine architectures
 > we currently have, but it is clear that we have them.  Pretending we
 > don't is for higher-level operations.

 Sure.  I was simply stating my discomfort with hard-coding a particular
 value for the current architecture into R6RS, as opposed to, e.g., providing
 procedures for querying the architecture, as you yourself suggested in your
 prior message.

 There are usability reasons (flexibility, the possibility that
 any particular value might seem quaintly dated in 5 or 10 years) and
 philosophical ones ("there are only three numbers: 0, 1 and infinity"
 that should be hard-coded in a design/specification).

 Cheers
 Andre