(j3.2006) paper J3/14-174 (again)

[Lionel, Steve]

Bob writes:

"I still do not understand the intent of paper J3/14-174.
 The paper proposes not to have the next standard forbid a
standard-conforming program from connecting a single file to more than one
unit at a time.  Is the intent of the paper that the results of all I/O
operations on different units connected to the same file are processor
dependent?"

That's my interpretation, but I share Bob's puzzlement at what the benefit
of this paper would be. If I read it correctly, it would change a program
that opens the same file on two units from being non-standard to being
processor-dependent. How is this an improvement? Many implementations
already allow such connections as an extension, with results that depend on
the vagaries of the underlying file system. Since the paper does not propose
assigning an interpretation of this action, I don't see that it helps the
programmer at all.  It's not as if this was a syntax rule that is required
to be diagnosed.

The purpose of a standard is to ensure that standard-conforming applications
have given interpretations. Yes, the standard identifies certain things as
processor-dependent, such as the order in which components of a derived type
get finalized, but that's more to alert the programmer that they can't
depend on a particular order - it isn't processor-dependent as to whether
you can have more than one finalizable component at all.

If the sole purpose of the paper is to allow a program that opens the same
file on multiple units to be labeled as standard-conforming, that's not a
compelling argument to me.

Steve Lionel
Intel Developer Support
Merrimack, NH


-------------- next part --------------
A non-text attachment was scrubbed...
Name: smime.p7s
Type: application/pkcs7-signature
Size: 6616 bytes
Desc: not available
Url : http://mailman.j3-fortran.org/pipermail/j3/attachments/20140616/7fcc4eaa/attachment.bin