Portability and long long

[Doug Gwyn (VLD/VMB)]

What all sane developers of portable software do when such data size
issues are considered important (more the exception than the rule, in
my experience) is to simply typedef some appropriate types in the
tailorable configuration header that you're going to need anyway in
order to accommodate variations among systems.  For example, the one
in the BRL CAD package defines a "fastf_t" type for the fastest
floating type, and we tailor the definition when installing the
package on a new system.

Also, power-of-two sizes are unnecessarily restrictive; you should
allow int48, int36, etc. if you're going to go this route at all.

In my opinion, cluttering the compiler with support for this kind of
stuff is inappropriate unless it is designed into the language from
the outset.  NCEG should not adopt the mindset that they are designing
a new language; if they were it really ought not to closely resemble
C!  The computing community surely collectively has learned enough to
design a language free of C's numerous quirks and deficiencies.  But
that is not NCEG's job..