[r6rs-discuss] Spec Size and Numeric Tower

From: Blake Coverett <scheme>
Date: Thu Sep 28 01:24:49 2006

I'd like to pick up the earlier concerns about the size of the new draft.
I enthusiastically concur with the idea of splitting things into
R6RS and R6RSSL, or some such name.

Interconnected with this issue is the question of compliance, and whether
it is a binary issue or one with defined levels. In particular, the
draft stance on the numeric tower concerns me.

>From R5RS, 6.2.3:

    Implementations of Scheme are not required to implement the whole
    tower of subtypes given in section 6.2.1, but they must implement
    a coherent subset consistent with both the purposes of the
    implementation and the spirit of the Scheme language.
    For example, an implementation in which all numbers are real
    may still be quite useful.

As opposed to the new section 2.3:

    Implementations of Scheme are required to implement the whole
    tower of subtypes given in section 2.1.

This puts a large burden on various specialized implementations. In fact,
I would argue that the overall domain of applications for which nothing
but exact fixnums would suffice is larger that the domain requiring any
complex numbers at all.

Certainly the whole tower must be specified, but why the radical
difference in implementation requirement?

regards,
-Blake
Received on Thu Sep 28 2006 - 01:24:44 UTC