Re: [Scheme-reports] Formal Comment: R7RS 'eqv?' cannot be used for reliable memoization

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

On 21-11-12 09:26, Per Bothner wrote:
> On 11/20/2012 11:43 PM, Mark H Weaver wrote:
>> But it's abundantly clear that you don't care.
>
> I think the point is that while a number of people agree with you
> technically that the current eqv? definition is not quite right,
> it's late in the R7RS process; there has been too much back and
> forth on the issue; and the current solution will have to do for
> R7RS: It provides the right answer for almost all implementations
> and if it doesn't, just do what you think is right. Perhaps we can
> tweak this for an Errata or if not for R8RS.

With all due respect for the hard work of the people doing the hard
work, I am quite disappointed with the way an arbitrary deadline seems
to have been set and several problems brought up recently (or not so
recently brought up but a proposed solution brought up recently) seem
to not be getting the attention they deserve because ``there is no
time'' to properly consider them.

We are actively fixing flaws, which so far have been limited

to minor editorial changes.

The problem with the eqv? issue is nobody agrees on the right

definition. Some people will be unhappy no matter what we

specify. I think it's a slightly odd spec only in that it ties us to

IEEE 754, but disagree that it's "broken" in any serious way.

As pointed out before, Mark is simply willfully bending the spec

to his own interpretation in claiming that MPFR must differ from

IEEE, when in fact MPFR is based on IEEE and any reasonable

interpretation of the standard would require it to return the same

results.

So I fail to see the issue. We could of course revise the wording

since some people find it confusing, and say that the rule applies

to something to the affect of "IEEE or any approximation thereof".

The mutable binding issue is much more involved, but in a

nutshell the R7RS library system was designed to be compatible

with all existing library systems. Most specifically, it was

designed so that it could be a simple wrapper around R6RS

libraries, and _requiring_ mutable bindings would break this

and be a rather severe breach of the charter. This isn't really

up for consideration.

If there is anything else you'd like to bring up, please do so.

Better late than never. If there truly is a serious flaw we can

make a rush vote, but it will be too late once the ratification

process starts.

Alex