[J3] J3/WG5 <-> POSIX?

[Robert Corbett]

The IEEE standards webpage says that IEEE Std 1003.9-1992 was withdrawn on 6 February 2003.  Its status is listed as "active - superseded", whatever that might mean.  No hint is given as to what superseded it.

For a time, Sun Fortran included an implementation of IEEE Std 1003.9.  It was a bother to maintain because the standard extended the semantics of POSIX I/O to include some "safe" I/O principles which were popular at the time the standard was written.  In particular, if an I/O system call returned a status of E_INTR, it had to be resumed.  My experience was that users seldom wanted such behavior.

Sun's implementation of IEEE Std 1003.9-1992 was used. Although Sun Fortran included its own operating system interface library, some users thought it was better to use a library API defined by an independent standards body.

Robert Corbett

> On Mar 5, 2018, at 10:02 AM, Keith Bierman via J3 <j3 at mailman.j3-fortran.org> wrote:
> 
> I'd have thought that object oriented storage would suit many interesting classes of programs. Others might benefit from direct use of mmap. Not sure if there's a good replacement for direct streaming I/O like we see in seismic processing.
> 
> Keith Bierman
> khbkhb at gmail.com
> 303 997 2749
> 
>> On Mon, Mar 5, 2018 at 10:57 AM, Anton Shterenlikht via J3 <j3 at mailman.j3-fortran.org> wrote:
>> >From: Bill Long <longb at cray.com>
>> >
>> > If there are still PXFxxx calls in Fortran programs that are essential and have no intrinsic replacement, we might look at creating a replacement.
>> 
>> Yes.
>> Or, if an HPC replacement for posix IO will emerge
>> in future, I'd like to revisit coarray IO.
>> 
>> Anton
> 
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://mailman.j3-fortran.org/pipermail/j3/attachments/20180305/1776ca27/attachment.html>