Shiro Kawai <xxxxxx@gmail.com> writes:

> I'm implementing srfi-252 from the spec and wondering if the "uniformly random
> distribution" requirement is desirable.
>
> For example, Gauche supports full unicode and it's easy to make (char-generator)
> return a generator that uniformly samples from the entire Unicode space. If we
> ...
> Will there be a drawback when we leave the distribution to each implementation?

Having uniform randomness allows us to enforce the use of SRFI 194,
which provides a standard way to create deterministic random values, by
setting the current-random-source parameter.

That is the main drawback. But I do think an ascii-char-generator would
be a good addition to the SRFI.