IBM's comments on 92-001 (FPCE)

[Doug Gwyn (ACISD/MCSB)]

Concerning the prototype parameter names:  As you note this is no
different for the Floating-Point C Extensions than for the base C
Standard.  I'm not convinced that it needs "fixing", as the
requirements on a conforming implementation clearly imply that the
actual prototypes put into effect by #include-ing standard headers
cannot be literally taken from the text of the standard.  (Indeed,
the prototypes might not even appear in textual form at all, but
could simply be a chunk of symbol table that is enabled when the
standard header is #include-ed.)  My personal practice in writing
standard header files is to omit the parameter names from the
prototypes, but if they are to be included then as you note they
need to be spelled in a conforming way, and double-underscore
prefixes is the most obvious method.  I think that adding this
syntactic clutter to the standards would slightly reduce their
readability and perhaps add confusion about whether the standard
prototype parameter names have to be used by implementations.