M0047 to amend Motion45: NO
I vote NO on the motion to require an Exact Dot Product based on CA in P1788.
I would vote YES if P1788 included a complete standard for Complete Arithmetic,
along the lines of the initial 29 Sept 2008 draft (Kearfott, Pryce & Revol),
written at a time when some of us still thought that P1788 covered "Interval
and Complete Arithmetic" as means toward "dependable computing". (That is
what had motivated my YES vote for Motion 9.) (This may be the same paper
as VanSnyderP1788.pdf in our position papers, or it may be older.)
In fact the IEEE "PAR" for P1788 only mentions Interval Arithmetic, so I
think Ulrich is out of luck.
I do believe that Complete Arithmetic is a valuable tool, and deserves
a standard of its own. The problem with hardware-flavoured standards
however is that computer technology changes so rapidly that methods that
were outstandingly useful at one point in time may become cumbersome at
a later stage of evolution. So a standard for Complete Arithmetic should
probably stress the logical aspects more than those tied to a physical
realisation -- except with regard to interchange formats.
Michel.
---Sent: 2013-08-26 19:24:53 UTC