J3/09-106
To:        J3
From:      Van Snyder
Subject:   Relationship between scale factor and rounding mode
Reference: 09-007, 09-105
Date: 2009 January 12

1.  The problem

Shouldn't scaling be done as in 10.8.5p3 no matter what I/O rounding
mode is in effect?

The caveat "If UP,..." appears not to have a reason, other than perhaps
to avoid imposing work on a vendor whose processor would be affected.
Needing to handle the new rounding modes, however, it's almost certain
that either the processor already handles all rounding modes according
to 10.8.5p3, or some code could be deleted, resulting in reduced ongoing
maintenance cost.

1.  Edits

[263:27-31: 10.8.5p3]---------------------------------------------------

Editor: Delete "If ... effect." from 10.8.5p3.

Make the list items independent paragraphs, properly capitalized and
punctuated, deleting ", and" from the end of the first one.

This isn't really a technical change since it replaces processor
dependent behavior with specified behavior.  It could be announced as a
tightening of specification in the compatibility subclause, but there
are no other such announcements there (currently 1.5.2) now.

If the "Alternative more extensive edits" in 09-105 are done, these
paragraphs should be combined with the "On input" and "On output"
paragraphs resulting therefrom.

2.  Alternative edit

[469: A.2]--------------------------------------------------------------

After "when the sign mode ... (10.8.4);" insert a list item

"o whether a scale factor is applied before or after conversion when the
   current I/O rounding mode is not UP, DOWN, ZERO or NEAREST (10.8.5);"