> Integer#divz, which truncates towards zero. Associated with these are
> corresponding modulo routines, #mod (the current behavior) and #modz
> (which is equivalent to remainder).
> Then, we have two 'personalities', which are actually just modules.
>    include ZeroTrunc
>         aliases Integer#/ to divz and Integer#% to #modz
>    include SchemeTrunc
>         does the opposite
> Have ZeroTrunc be the default, but allow people to load SchemeTrunc
> for the traditional behavior.
> Library routines which can potentially run in either environment
> should use the underlying .div and .divz routines.

This leads to either inflexible code or the default division operator to
be condemned.

Perl once did that with base array index variable $[ so that people can
choose either the C style 0..n-1 indices or Ada style 1..n; this is now
widely regarded as bad style, especially since the fast raise in the
number of publicly available modules. Some versions of Basic had a similar
OPTION BASE thing that was similarly unused, and became even less used in
more structured versions of Basic.

Following my observations I think any person truly believing in modularity
will religiously avoid such compromises. To me, rounding towards the
bottom is the right thing; and i can live with rounding towards zero as
long as there's an alternative, albeit slightly longer to type. But i
would certainly not like the compromise you suggest.


Mathieu Bouchard