Re: mixed decimal floating and binary floating arithmetic

[Keith Bierman <khbkhb@xxxxxxxxx>]

On Fri, Apr 17, 2009 at 10:28 AM, Steven Hobbs
<Steven.Hobbs@xxxxxxxxxxxxxxxx> wrote:

> I take it that American National Standard X3.9-1966 USA Standard FORTRAN
> (and ANS X3.10-1966 USA Standard Basic FORTRAN) were not Fortran

That it has been permitted for several rounds of standardization since
is sufficient to make changing it now impractical. '66 was replaced by
'77 which in turn begat Fortran 90, Fortran 95, and Fortran 2003.

Also, note that many '66 era compilers permitted it as an extension.
That the standard prohibited it meant that users could not count on it
working, not that a compiler was required to diagnose it (in modern
versions of Fortran standardization those things the compiler must
diagnose are *constraints* not user level prohibitions) much less take
any action to prohibit it.

...>
> I do not think the answers to these questions are easy.
When starting from scratch, they are hard. Once on accepts the notion
that any $standard compliant program should also be a $(standard+1)
compliant program it is hard to take away features ... even if they
are now recognized as ill-advised the problems are hard, but in
different ways (viz. it's no longer a question of if, but how).


-- 
Keith Bierman
khbkhb@xxxxxxxxx
kbiermank AIM