Am Do., 29. Aug. 2024 um 08:49 Uhr schrieb Michael Sperber <
[email protected]>:

>
> On Wed, Aug 28 2024, "Will Clinger" <[email protected]> wrote:
>
> > You appear to be assuming there is some objective "actual meaning
> > of x|p".  In my opinion, the purpose and specification of that notation
> > has never been defined so well as to create an objectively actual
> > meaning.
>
> To provide some historical context, the notation's immediate ancestor is
> this paper:
>
> https://www.deinprogramm.de/sperber/papers/numerical-towe of the r.pdf
> <https://www.deinprogramm.de/sperber/papers/numerical-tower.pdf>
>
> See section 6.6 "External Representations".
>
> The paper proposes a greater revision of the numerical tower in light of
> its inconsistencies and practical problems.  In that context, 0.7|52
> denotes an exact (and thus well-specified) number.  Within the greater
> debate, some minor suggestions from that paper were adopted for R6RS -
> such as the mantissa-width tagged format - but (sadly IMHO) the greater
> proposal was not, leading to the imprecisions in the R6RS wording.
>

Thank you for pointing out that paper and clarifying the origin of the x|p
notation.  The difference to R6RS in this context seems to be that what is
0.7|52 in the paper is #e0.7|52 in R6RS mainly because of R6RS's rules for
inexact number literals (and because of dropping the R5RS distinction of
exactness/inexactness in the paper).

In the context of the paper, the meaning of the R6RS text becomes even
clearer, IMHO, although we all agree that it is not optimally clear.

I would like to add that the R6RS text actually gives an example that
should make the meaning clear: "For example, 1.1|53 is a representation of
the best approximation of 1.1 in IEEE double precision."  Notably, this is
consistent with what the paper proposes.  Independently, it clarifies that
the meaning of 1.1|53 as a number (not a number object!) is independent of
any machine representation.  The other paragraphs describe the relationship
of the number object with this number, which it has to represent.

Reply via email to