Below is a copy of the cover message and most recent draft of
the IEEE-FP Floating-Point proposal posted to comp.lang.forth,
August, 28, 2009.

Because some typos have been fixed, the version number here has
been changed to 0.5.4a instead of 0.5.4.

Discussion has been going on for quite some time in
comp.lang.forth.

http://tinyurl.com/RfD-IEEE-FP-050-051
http://tinyurl.com/RfD-IEEE-FP-052
http://tinyurl.com/RfD-IEEE-FP-053
http://tinyurl.com/RfD-IEEE-FP-054

--------------------------------------------------------------
Here's draft version 0.5.4a.  It mostly aims to clean up a few
things in anticipation of the September meeting.

The proposal is of course in no way ready for submission to the
standards committee.  It's not ready for a CfV, and before
reaching that point its organization needs to be revamped, and
reference implementations need to be supplied.

There are still unresolved issues in the c.l.f. discussion, but
my hope is that the proposal is far enough along to give
committee members a sense of where it's headed, and perhaps
generate feedback from them.

Issues still under or needing discussion:
-----------------------------------------

* Exception handling for fp number interpretation.

* Accuracy of decimal/binary conversion.

* >FLOAT vs. >IEEEFLOAT, REPRESENT.

* Whether current rounding with roundTiesToEven as default is
  okay as a replacement for DPANS94 round to nearest and
  implementation-defined rounding.

* Whether GET-FSTATUS and SET-FSTATUS are okay.

* F~, others?

Missing:
--------

* Reference implementations for decimal/binary conversion,
  maybe others.

* Glossary for managing rounding modes.

* Glossary for alternative exception handlers.

Summary of changes:
-------------------

* Reverted to one word set, with fp status flag and rounding
  mode words in an extensions glossary.  Revised introduction
  accordingly.

* Made correct rounding a "shall" for binary/decimal conversion
  and arithmetic, and a "should" for IEEE recommended, correctly
  rounded functions.  Removed "nominally shall" terminology.

* Removed MAX-FLOAT-DIGITS environmental query.  Renamed
  IEEE-FP-ENV query as IEEE-FP-EXT.

* Revised REPRESENT to use correct rounding and signaling, and
  to require reproducing round-trip conversion; and expanded
  rationale.

* Accuracy of conversion no longer implementation-dependent.
  Accuracy of calculation remains so.

* Default signal handling for text interpretation, in particular,
  +/-infinity and no stopping on overflow.

* Redrafted D>F and rationale.

* Redrafted F>D and rationale, added suggestion by mhx for
  ambiguous condition behavior.

* In >IEEEFLOAT, "nominally shall" replaced by "shall" for
  current rounding.

* For F., FS., FE., added rationale statement on change to
  current rounding.

* Noncore math functions:  "Nominally shall" replaced by "shall"
  for current rounding.

-- David