In my work area, I have numerous workstations of different flavors -- PC's
running Linux, NT and Windoze, Sun, HP and SGI. I don't think we have any
IBM or DEC. I have 126 file systems in my NFS "automount" map.
If I write a Fortran unformatted file on my PC, I can't read it on my HP
workstation. I think the only problem is endian-ness -- I think both my
PC compiler vendor and HP put four-byte record-length fields before and
after each logical record, they both use IEEE floating point, they both
use 2's complement for negative numbers, ASCII, .... Maybe there's a
problem with the representation of LOGICAL, too.
In order to facilitate interchange of unformatted Fortran files, I suggest
that vendors extend the meaning of FORM= in an OPEN statement. The extended
values would all indicate variations on unformatted (binary) data. They
could be names of the compilers of code expected to read unformatted data
being written, or the names of the compilers of code that wrote the data
being read. I think in most cases the differences, if any, are a matter of
byte order. I don't think it's appropriate to standardize this, but it
would be best if vendors informally agreed on each other's names, and
published, or at least exchanged, their unformatted data format. Of course,
no vendor need supply anything more than the "formatted" and "unformatted"
mandated by the standard.
Do vendors need anti-trust protection, by way of a formal standards
organization in order to do this?
What do you think of this idea?
Best regards,
Van Snyder
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
|