(SC22WG14.323) Re: reaction to Variable Length Array Proposal

[Anthony L. Kimball]

To me the rub is that you need to be able to write intercallable code.  I
do not feel that the reasons you describe are as significant to users cum
library writers.  C should allow it's users to generate the code they wish,
within reason.  That's why I'm very fond of RMS's proposal: It both retains
forward declaration order (avoiding semantic morasses such as mutally
recursive sizeof()) and allows the user to indicate a particular stack
order/register assignment where the underlying architecture supports it.

//alk (for alk)
Anthony L Kimball
Thinking Machines Corporation

   Date:     Thu, 11 Feb 93 11:13:00 EST
   From: Doug Gwyn (VLD/VMB) <gwynaBRL.MIL>

   As was pointed out to me off-line, one of the main reasons people
   are asking for "preservation of the Fortran calling order" (which
   by the way is backwards from most of the Fortran library functions
   I once designed) is that they expect automatic Fortran-to-C
   converters to be applied to existing Fortran code and these
   converters are not smart enough to remap calling sequences.
   Another reason offered is that existing Fortran library documentation
   could still be reasonably used by human C coders if the order were
   preserved.  (Of course it could also be used if the interface map
   were described by simple rules, e.g. all dimension arguments follow
   all other arguments but otherwise the order is the same as Fortran.)

   I don't much buy these arguments but out of fairness they should be
   understood.