Bill; Fairchild is right.

Ideas for new instructions often begin with the notion that something
ugly ought to be avoidable; but that notion is not enough, even when
it has obvious merit.

To take such a notion further it should first be implemented as a
macro, perhaps in several ways, complete with diagnostic machinery
(and instrumentation that can be disabled).  Then, after you  have
used that macro for a time, and probably after making some iterative
improvements in it, you will be in a better position to propose its
implementation and to make a case for its usefulness.

You will need to be able to make a very good case, one that cannot be
shot down because your proposed instruction has some disagreeable side
effect.  Moreover, the idea that some alternative sequence of several
instructions is error-prone, lends itself to abuse, will probably not
be persuasive.  All instruction sequences lend themselves to misuse by
clots.

New instruction proposals abound.  Many of them are ill-conceived; and
few of them get far; most, it would appear, are not even implemented
experimentally in multicode.

John Gilmore, Ashland, MA 01721 - USA

Reply via email to