John Cowan wrote:
> 1) Should R7RS-large require arbitrarily large (up to implementation
> restrictions like memory) exact integers?

It should at least require some sort of well-behaved overflow. Bignums
are ideal, but throwing an exception is almost as good and easier to
implement. (But not overflowing to inexact, because it silently gives
incorrect results.)

There are a lot of bignumless Schemes, so it would be nice to not make
trouble for them. On the other hand, bignums are Scheme's only escape
from small fixnums, and >30-bit integers are needed often enough that
they're maybe worth the trouble.

> 2) Should R7RS-large require support for exact rational numbers?
>
> 3) Should R7RS-large require support for exact complex numbers?
>
> 4) Should R7RS-large require inexact complex numbers?

No, because few programs need any of these. Exact rationals are
sometimes handy, but it's easy to substitute other types. Complex
numbers are seldom used outside math/physics programs, and AFAIK exact
complexes are used only for computer algebra. Programs that do need
them can already use them semi-portably, since Scheme already
standardizes their interface. Requiring them would consume
implementation effort for little benefit.

The numeric tower is one of the areas Scheme does best, and least
needs improvement. When I want to use Scheme and can't, it's because
of problems with I/O or basic data structures or system calls or
reflection or verbosity, never numbers.

(This isn't a vote; I don't think I use Scheme enough to get a vote.
My opinion is relevant only because I'm one of the marginal practical
users R7RS-large is aimed at, who might use Scheme more if it were
less painful.)

_______________________________________________
Scheme-reports mailing list
Scheme-reports@scheme-reports.org
http://lists.scheme-reports.org/cgi-bin/mailman/listinfo/scheme-reports