To retain backward compatility for umpteen lines of code the assignment of 5
and 6 to standard input and output will have to be retained at least as a
compiler option if not in the standard. I'd prefer to see a complete
departure from the numerical assignments to symbolic assignments such as
input, output and error for the standard streams and for the number of
different symbolic stream names to be limited only by system resources,
perhaps configurable as a compiler option to restrain programmers.
Whether or not you believe in end-of-file markers they do exist for all
files as something has to indicate where the data stream ends: it may be a
special record but that's still something to mark the end of file and , yes,
editors do write end of file records.
The ENDFILE functionality may be rarely required nowadays but can still be
useful in certain circumstances. My own preferred approach would be to copy
the data from the problem file to a new file which would then be properly
terminated but this may be impractical for large datasets if there's
insufficient free space.
My personal feelings about the whole development of FORTRAN over the years
is that all standard functionaly should continue to supported through the
various stages: IV->77->90->2000. The language is being developed for the
use of the programmers and not for the compiler writers. It may be a
nuisance for compiler writers to continue to support "obsolete" features but
it's an even greater hassle for programmers to have to rewrite software
which has been working well for years.
John
Name: John B. Macallister E-mail: [log in to unmask]
Post: Nuclear and Astrophysics Laboratory, Keble Road, Oxford OX1 3RH,UK
Phone: +44-1865-273388 (direct) 273333 (reception) 273418 (Fax)
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
|