[r6rs-discuss] Re: [Formal] eq?/eqv? misbehave around NaNs

[Marcin 'Qrczak' Kowalczyk]

Alan Watson <alan at alan-watson.org> writes:

>>>No. Consider an implementation has multiple versions of, say,
>>>symbols, arranges from them to be eq? and eqv?, and also provides
>>>a non-standard means to distinguish them. Symbols with the same
>>>spelling have to be eq?, but these symbols are indeed "treated
>>>differently" by other (non-standard) functions.
>> IMHO it should not do this.
>
> Can you tell me which part of the standard (draft R6RS or R5RS)
> forbids this behaviour?

I don't mean that they currently forbid it. I mean that providing
consistently distinguishable values which are not eqv? violates my
theory about the spirit of eqv?. Thus it's a bad idea, and should
be forbidden. Even if it's not forbidden, e.g. because it's hard to
define eqv? precisely in the presence of non-standard extensions,
implementations should not do this.

-- 
   __("<         Marcin Kowalczyk
   \__/       qrczak at knm.org.pl
    ^^     http://qrnik.knm.org.pl/~qrczak/