[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
Issue: FORMAT-ROUNDING (Version 1)
- To: KMP@STONY-BROOK.SCRC.Symbolics.COM
- Subject: Issue: FORMAT-ROUNDING (Version 1)
- From: Jon L White <email@example.com>
- Date: Sun, 13 Nov 88 20:22:28 PST
- Cc: CL-Cleanup@SAIL.Stanford.EDU
- In-reply-to: Kent M Pitman's message of Sun, 13 Nov 88 18:33 EST <881113183353.8.KMP@BOBOLINK.SCRC.Symbolics.COM>
I strongly oppose this change.
The choice to "Round Up" is counter to the choice for the ROUND function
on CLtL p.216, where "ties" are specified to round to the nearest even
integer. It is counter to the general direction that CL should follow
IEEE conventions wherever possible. [Note that having "ties" do "Round
Up" is a unique property of VAX hardware -- there isn't even any IEEE
rounding mode that corresponds to it.]
According to CLtL, p390, the allegedly "unpredictable" results are in
fact very predictable -- precisely one of two strings will be output,
both of which are guaranteed to represent the number in question;
and there is no other possibility. Narrowing the choice down to just
one of these two strings serves no useful purpose, and may even cause
some (very minimal) implementational difficulties for some machines.
But if we must choose between one of the two, I suggest we pick an
algorithm that follows the IEEE style rather than the VAX style.
-- JonL --