George Corliss wrote:
My extrapolation of Nate's argument below sounds just a little bit like
Bill Walster's, "Containment is REQUIRED; all else are quality of
implementation issues." Or simply, "Thou shalt not lie."
A standard requiring (almost) only containment would allow InfSup and
MidRad, its exception-handling could be quite simple (but not trivial),
and there probably would be (shudder) NO common underlying levels or
theory.
Would that be any "standard" at all?
I don't agree there would be no underlying levels or theory, but otherwise,
yes, I think what you describe actually sounds pretty good! :-)
On the counter to your argument, though, I don't expect anyone will stop
using fast matrix multiply in Intlab just because 1788 (apparently) is
going
to decree the software is non-conforming. However, the question
lingering in
my mind then becomes: has P1788 really accomplished anything relevant
except
creating a bunch of rules no one likes to follow?