Motion 59: NO
I vote NO on Motion 59: Section 14 "Level 3 description".
My reasons are as follows:
1. This is just too long and too complicated. I think it is important
that a standard strives for clarity and readability. For comparison, in
IEEE-754, the description of the four (!) levels is only half-a-page
long. Here the theoretical description of just level 3 is already more
than one-page long.
2. I disagree with the requirement that an implementation should compute
results which representations are independent of the representation of
the inputs. Even for the sake of reproducibility, this is too strong.
Reproducibility should only require that the outputs be the same when
given the same inputs, at a given level. There is no reason for
level-2-equal inputs to produce level-3-equal outputs (rather than just
level-2-equal outputs).
3. The purpose of an interchange format is to allow exchange of data
between two different implementations. So making an interchange format
almost completely implementation-defined is pointless.
Best regards,
Guillaume