Closed mflatt closed 10 months ago
According to #400 the idea was to preserve the r6rs behavior for the version of /
exported from (rnrs)
, but that seems not to have happened.
I should think the Right Thing, if R6RS is wrong, to add this to the R6RS errata. In that case, the (rnrs)
and (chezscheme)
libraries can do the same thing.
According to https://github.com/cisco/ChezScheme/issues/400 the idea was to preserve the r6rs behavior for the version of / exported from (rnrs), but that seems not to have happened.
True. My thinking is that adding r6rs:/
is just more stuff and unlikely to matter to anyone, but it's straightforward to add if someone really wants it.
If it makes sense to amend R6RS to allow this, so much the better. I agree with Kent that we got this wrong.
That said, I'm surprised to discover that I long ago made Racket's R6RS layer stick to the current spec: https://github.com/racket/r6rs/blob/master/r6rs-lib/rnrs/base-6.rkt#L264
A proposal to move forward:
/
that references this PR and #400 and serves as a place for someone to ask for r6rs:/
.I'm happy to do that if there are no objections.
The release notes mentioned a change to the treatment of exact 0 as a divisor, but not as a dividend. The release notes now also point out that this change is not consistent with R6RS.