P.J. Plaugher is a heavyweight dude; he was president and founder of Whitesmiths, a compiler firm which was recently acquired by Intermetrics. His title there now is "Chief Engineer", which is something like "IBM Fellow". He has done a number of different C compilers, and is a Secretary to ANSI. He has spent the last six years serving on the X3.J11 committee trying to forge a standard for the C language.

The original C definition was Kernighan and Ritchie's textbook on C. As such, it was rather vague and incomplete. Though C requires a standard library, K&R does not specify it very well. Appendix "A" in the book contained a BNF grammar for the language, which was as close as many compiler writers got to a formal definition.

X3.J11 has done its best to rectify that. They have gone to great lengths to ensure that the new ANSI standard does not render existing programs incompatible. Plaugher noted that some standards committees like to redefine the language every couple of years - the FORTRAN committee is a famous example. X3.J11 has added a number of new features in deference to the international community (Far Eastern character sets), and for numerical analysts. They have also included some of the nicer features of C++, a second-generation C language devised at Bell Labs.

X3.J11 is almost done; they've voted to finalize the current draft, and send it to ANSI for review. ANSI C is the base document under discussion as an ISO standard in the international arena as well (Denmark and U.K. want to make a couple of changes, but this is a political problem more so than technical).

ANSI C gives you the ability to assist the compiler more than K&R C does. New compilers can do additional type checking and coercion when presented with ANSI code. New keywords have been added to help the compiler determine what variables to keep in registers, and which NOT to store in registers. The current state of the art in compiler technology allows compiler writers to be really ambitious; the objective these days is to generate code that is MORE EFFICIENT than assembly language. They can realistically do this because of the sophisticated flow analysis algorithms that are available these days.