My view on John's paper
Folks,
Since John has stated he is going to pursue a motion on "red and blue"
sticky, I'd just like to mention a few reasons why I don't think this is
suitable for a 1788 standard:
I believe the C++ community has already set a satisfactory precedent with
bool_set, and its probably likely that bool_set may become part of a future
version of the C++ standard. The definitions for tetrits presented in my
paper are compatible with this methodology.
P1788 of course has no obligation to conform to other standards, and I don't
suggest on this merit alone that my paper should be accepted. However, to
not ride on the coat-tails of other people's success -- when possible -- I
think misses out on a big opportunity. So this is why I mention it.
More to the point, John's ND and SuD bits can really only be viewed as a
trits worth of information (this is illustrated nicely by John in the tables
on p. 7), and so he tries to rectify this problem by adding a fourth
"ill-formed" bit. IMO, all this accomplishes is to take the bool_set concept
that would otherwise likely be familliar to an increasingly larger and
larger audience of engineers and turn it into something unfamilliar. Plus it
requires 3 bits to do what bool_set accomplishes in 2 bits.
Also, I don't see that anyone (inluding John) shows a plausible real-world
example of another attribute (other than "domain") that might require
anything other than a bit of information. So I think the theoretical
arguments John tries to make in this direction with regards to tetrits are a
bit too abstract.
I also find it extreemely disturbing that with John's propagation mechanism,
a lengthy computation involving a function that is "nowhere defined" and
"nowhere undefined" may yield a final decoration that is "somewhere
undefined". In my view, this is a violation of the concept of structural
induction in the sense that the final result should represent the worst
exceptional condition encountered while evaluating the DAG.
Sincerely,
Nate