numeric-only

[David G. Hough on validgh]

I have set up a separate sublist which I call numeric-only for people who
only want to receive messages addressed to numeric-interestavalidgh.com
and not any addressed to ncegacray.com (where NCEG messages go) or
numeric-nonsunavalidgh.com (where selected Sun product announcements go).   

If you want to transfer from
numeric-nonsun (the sublist of non-Sun employees) to numeric-only, or
make any other change, send mail to me at numeric-interest-requestavalidgh.com.

However I don't recommend making the change.   The principal challenge for
numerical computation is to get the instruction set architecture, 
hardware implementation, operating system, languages, compilers, 
standard libraries, and mathematical software
all working in the same direction or at least out of each other's way.
To do this effectively we must know something about the other guys' issues,
or risk being ignored as we have in the past.

The NCEG mailing list is an adequate though imperfect mechanism toward that
end.   Like any mailing list it can get sidetracked by repetitive or
flamey postings, though.

In particular I think the discussion of C prototypes is very relevant,
on the whole.
C prototypes are where to find any interface information that is going to
be automatically conveyed.   As I have indicated in the past, I think that
might well include additional assertions about function parameters beyond
transmission order and type.   Yet C prototypes are tricky to implement in
full generality, and the details of syntax and semantics for extensions such
as variable-length arrays are very important in determining whether they will
ever be implemented at all.

	The machine designer who has put the extra bits in may now
	be amused to discover that the language designer has made it
	difficult to use the extra bits in higher languages...
	Seemingly we must design the hardware, the language, the 
	compiler, and the operating system (to handle overflow, etc.)
	together from the ground up!   We will have to leave the 
	reader confronted by this grim prospect.

		-- W. Kahan, Lecture Notes on Implementation of Algorithms,
		   February 1972