scope and extent of NCEG proposals

[Thomas M. Breuel]

I know little about the mandate or procedures of NCEG, but following
the discussions on the NCEG mailing list and going through the stack of
papers I received via US-mail (thanks), I have become a little
concerned about the scope of the changes to C being considered by NCEG.
Everybody seems to try to seize the opportunity and get their favorite
numerical feature into C.

The most important issues that I see with C for numerical programming
are variable-sized multidimensional arrays and a "restricted pointer"
declaration.

There are certainly many features that "might be nice to have", such as
variable-sized arrays inside structures, generic functions, operators
for IEEE support, precision specification for integers and floating
point numbers, vector operations, etc. 

But such features are not essential for numerical programming (FORTRAN
77 lacks most of them), and their implications for the programming
language and their interaction are poorly understood. By wasting so
much time on their discussion, getting the essential problems of C
fixed seems to get delayed more and more.

Personally, I wish NCEG would set right now a definitive and simple
standard on variable sized arrays and restricted pointers so that
people who want to use C for day-to-day numerical programming can get
their work done.

					Thomas.