Fortran vs. C, and noalias: old postings, new data

[Henry Spencer]

In article <TMB.90Dec7234544 at bambleweenie57.ai.mit.edu> tmb at bambleweenie57.ai.mit.edu (Thomas M. Breuel) writes:
>   --> X3J11 seems to be reluctant to include a noalias declaration
>       in C

Not really right.  X3J11 believed that there was a fairly strong need for it.
The trouble was the near-total lack of any *experience* with such things in
C.  There is no substitute for actual experience with proposed language
features:  "expert intuition" is wrong more often than it's right.

X3J11 did indeed attempt to put a `noalias' feature into C, but persistent
technical problems (people knew what was wanted, or thought they did, but
nobody seemed to be able to come up with words that said it satisfactorily)
and some singularly bad timing (they waited until too late) provoked a
firestorm of protest and it was removed.

(Incidentally, note the past tense.  ANSI C is a standard now, and has been
for some time.  X3J11 is no longer actively considering language changes,
and won't be until the standard comes up for review several years hence.)

My own opinion is that we don't really know how to make this work in C yet.
We need actual working experience with some tentative solutions.  Standards
committees -- the ones that do their job right -- are supposed to standardize
proven existing practice, not invent their own untried solutions.  Contrary
to the popular misconceptions in certain quarters, the right thing for a
standards committee to do when faced with a demand for a standard in some
poorly-understood area with no existing practice is to *refuse*.  Bad
standards (actively bad ones, I mean, not just suboptimal ones) really are
worse than no standards, and committees have no way to magically produce
good standards when nobody understands the subject area well enough.
-- 
"The average pointer, statistically,    |Henry Spencer at U of Toronto Zoology
points somewhere in X." -Hugh Redelmeier| henry at zoo.toronto.edu   utzoo!henry