The first meeting of the IEEE 754R revision group was held Monday 8 January 2001 at 1:00 pm at Network Appliance, 495 E Java Dr, chair and host Bob Davis, firstname.lastname@example.org. These notes are intended to give an informal flavor of the discussion. Attending were Bob Davis, Dick Delp, Rick James, Dan Zuras, David Scott, Peter Tang, and David Hough.
The next meetings were scheduled for 12 Feb, 12 Mar, 9 Apr, 9 July, 8 Oct, 14 Jan. A venue for 12 Feb is to be determined; the other meetings have been scheduled for the Craftsman conference room, building 2 at Network Appliance, 1-5pm.
The plan is that a draft for circulation will be ready after 9 Apr, which can be amended on 9 Jul, and passed to the parent committee by the end of the year. The IEEE/CS Microprocessor Standards Committee would then conduct its review, a process requiring about a year. This kind of fast-track scheduling depends on few, minor, upward-compatible changes. By the way, an acceptable outcome from the IEEE's point of view would be to ratify the existing standard unchanged, as has happened at the previous reviews circa 1990 and 1995.
A mailing list has been established for this work. Send a message "subscribe stds-754" to email@example.com to join. Knowledgeable persons with a desire to contribute positively toward an upward-compatible revision are encouraged to subscribe and participate.
The bulk of the meeting was devoted to listing some plausible revisions. Items which are not quite upward-compatible, strictly speaking, would presumably be "should," and perhaps "shall" for new implementations, in order to grandfather in existing implementations. The following is a list of ideas from various sources. Being in the same list here does not imply the same level of support or probability of ultimate success.
One instance in which an inexact exception could be helpful would be in detecting inadvertent precision-losing implicit casts in languages like Fortran which permit them; but the cases in which that inexact exception would not be lost among normal roundoff must be fairly uncommon.