YES P1788/M0029.02:Level3-InterfaceOnly, *BUT*
Yes, *BUT*.
I believe the exclusion of overflows of the same sign is unnecessary
for the interchange format.
Exemplar: Given M = F.max for some 754 format F, then given lo=0.8*M
and hi=0.9*M the interval I_one=[lo hi] is a valid value. But the
interval I_two = 2 * I_one is not a valid interchange value. That
inconsistency is unnecessary and should be eliminated from the
interchange definition.
In fact I would go further and require that all pairs of values drawn
from F be acceptable for the interchange format. One might compromise
that generality by insisting that the first element of each such pair
be strictly* <= the second element of the respective pair. However,
that restriction is also unnecessary and an obstructive constraint on
further development because it eliminates the possibility of using the
interchange format between implementations that support "improper"
intervals. The interchange format should be as generally useful as
possible.
*{some attention must be paid to the canonical ordering of 754 values
because [-underflow +underflow] is not the same interval as
[+underflow -underflow] even though 754 specifies that -underflow and
+underflow compare as equal.}
Such standard-imposed constraints on the interchange format buy
nothing. The acceptability of any interchange data set will need to
be verified by any implementation importing it whether there are
standard-imposed constraints upon the valid values or not. So
demanding such constraints does not make the implementation any
simpler, faster, or easier. However such constraints do impose
unnecessary burdens on implementations and their future evolution.
Lee Winter
NP Engineering
Nashua, New Hampshire
United States of America