From owner-sc22wg5@dkuug.dk Fri Aug 8 19:04:24 2003 Received: (from majordom@localhost) by dkuug.dk (8.12.8p1/8.9.2) id h78H4OvV031548 for sc22wg5-domo; Fri, 8 Aug 2003 19:04:24 +0200 (CEST) (envelope-from owner-sc22wg5@dkuug.dk) X-Authentication-Warning: ptah.dkuug.dk: majordom set sender to owner-sc22wg5@dkuug.dk using -f Received: from mailhub.dfrc.nasa.gov (mailhub.dfrc.nasa.gov [130.134.81.12]) by dkuug.dk (8.12.8p1/8.9.2) with ESMTP id h78H4JEc031540 for ; Fri, 8 Aug 2003 19:04:20 +0200 (CEST) (envelope-from maine@altair.dfrc.nasa.gov) Received: from mail.dfrc.nasa.gov by mailhub.dfrc.nasa.gov with ESMTP for sc22wg5@dkuug.dk; Fri, 8 Aug 2003 10:01:17 -0700 Received: from altair.dfrc.nasa.gov ([130.134.20.211]) by mail.dfrc.nasa.gov (Post.Office MTA v3.5.3 release 223 ID# 0-71686U2500L200S0V35) with ESMTP id gov for ; Fri, 8 Aug 2003 10:04:13 -0700 Received: by altair.dfrc.nasa.gov (Postfix, from userid 201) id B0AF0357ED; Fri, 8 Aug 2003 10:04:08 -0700 (PDT) From: Richard Maine MIME-Version: 1.0 Content-Type: text/plain; charset=us-ascii Content-Transfer-Encoding: 7bit Message-Id: <16179.55304.619910.179217@altair.dfrc.nasa.gov> Date: Fri, 8 Aug 2003 10:04:08 -0700 To: sc22wg5@dkuug.dk Subject: (SC22WG5.2926) clarification of technical intent in N1553 In-Reply-To: <200308081603.h78G3X9w031243@dkuug.dk> References: <200308072303.h77N34rh026353@dkuug.dk> <200308081249.h78CnM0a030247@dkuug.dk> <200308081330.h78DUv8x030439@dkuug.dk> <200308081507.h78F7FHt030927@dkuug.dk> <200308081603.h78G3X9w031243@dkuug.dk> X-Mailer: VM 7.07 under 21.4 (patch 12) "Portable Code" XEmacs Lucid Sender: owner-sc22wg5@dkuug.dk Precedence: bulk John Reid writes: > No, I think we meant exactly what we said. There may be some useful > information that can be provided for a quiet NaN. Why disallow providing it? Ok. I can see some sense to that. And in any case, I'm not trying to argue the technical decisions now (too late for that), just to make sure that I understand what they are. That part sounds pretty clear. > > Anyway, if we really meant exactly what we said for input (specifying > > that we get a quiet NaN when there are no optional characters, but > > allowing the form with optional characters to give either a quiet > > or signalling NaN), then the above edit for output isn't symmetric. > > Why not? I intended it to be. On input, a NaN with no parentheses is a > quiet NaN. So we need to disallow a signaling NaN to be output this way. That sounds sensible, but when I put input and output together, I end up with the following. 1. If we start with a quiet NaN: Write it out. We don't specify whether it has the optional info. Read the output back in. Because it could be in either form, we could get either a signaling or a quiet NaN. 2. If we start with a signaling NaN: Write it out. Your proposal specifies that we include the optional parens. Read the output back in. This form could give us either a signaling or a quiet NaN. So it seems like, starting with either kind of NaN can end up with either kind; doesn't seem like we have achieved much. Seems to me that if the only thing we specify on input is that the form without parens gives a quiet NaN, that the useful corresponding restriction on output would be that a quiet NaN would give that same form. Then you'd at least be able to guarantee that starting with a quiet NaN, you'd end up with one. But that output restriction would violate the spirit you mention of allowing possibly useful information about quiet NaNs to be expressed. So I guess I'm now coming to the conclusion that if we meant exactly what we said about input, there isn't a corresponding restriction on output that fits the same spirit and also portably assures any useful end-to-end result in this matter (I regard ending up with the same kind of NaN as you started to be a useful result here). So perhaps that end-to-end matching isn't part of the inferred requirements. (I didn't go back to check whether we actually wrote down a formal set of requirements.) In that case, perhaps my point 4 would be best rejected, as its purpose was to achieve such matching; doesn't look to me like we will get it, so perhaps the best answer is to do exactly the edits that WG5 passed and no more. Even if WG5 didn't reason it through this way, perhaps they got the best answer by accident anyway? -- Richard Maine | Good judgment comes from experience; Richard.Maine@nasa.gov | experience comes from bad judgment. | -- Mark Twain