J3/06-324

Date:    25 October 2006
To:      J3
From:    Bill Long
Subject: Issue 16, bits compatible term
Reference: J3/06-007r1, 06-214

---------------------

Discussion of issue 16 (page 311)

Issue 16 asks whether "bits compatible" be folded into a new concept of
"type and kind compatible"?

The term "bits compatible" is currently used in three contexts: bits
editing [269], argument association [304,311 (several places)], and
variable definition [500].  The usage on page 500 might be removed
pending the resolution of Issue 76 (see paper J3/06-326).

In all these contexts the issue of rank compatibility is independent
of whether one of the objects is of type bits, so the question is
effectively whether "bits compatible" can be folded into TKR
compatible [304:9-12].

It would seem irregular to involve TKR in the description of
input/output editing. In the context of argument association involving
intrinsic types, TKR compatible implies the same types and equal kind
parameter values. Unless both objects of are type bits (the
uninteresting case here) neither of these conditions applies to bits
compatible.  Rather, what are equal in the case of bits compatible are
the object's sizes expressed in bits. The term "bits compatible"
directly ties to this definition and is intuitively descriptive.
Rather than incorporate orthogonal concepts into a new meaning of TKR,
I believe it is simpler and clearer to keep the separate "bits
compatible" term.


Edits:
------
(none)