02-311r1

    Feature Changes

To:         J3
From:       Craig Dedo
Date:       November 13, 2002
Subject:    Feature Changes

Date:     28 Oct 2002
To:       J3
From:     Richard Maine
Subject:  Feature changes

The following are proposed public comments relating to features
to be included or omitted from f2k.

1. I propose that the enum feature be deleted.

   As currently constituted, it adds almost no functionality to
   the language.  Except for the BIND(C) case, it is nothing but
   another syntax for functionality that we already have.  The
   only new functionality of the feature is the kind
   determination of BIND(C), which doesn't appear to be much of
   an issue in practice since apparently most C compilers use C int
   for all enums.

   If a future revision ever does add a strongly typed enum
   feature, the overlap between that and the current one
   will contribute to the language seeming bloated and
   haphazard.  I believe this potential negative impact to
   outweigh the small benefits of the current feature.

   Adding a strongly typed enum feature now is not reasonable in
   my opinion.  It would have too many potential interactions
   with other features and would consequently cause substantial
   delay in the standard.

   Deletion of this feature has little impact on the rest of the
   standard.  The enum feature is currently nothing other than an
   alternate syntax for defining a type alias and a set of named
   constants; no other features depend on whether this or some
   other syntax for the same thing is used.

J3 Response: We recommend that we leave enums in the standard but require
the use of the BIND attribute.

2. I propose that an intrinsic to execute a system command be
   added.  The CD adds intrinsics to obtain information about the
   command used to execute the Fortran program.  The ability to
   in turn execute some other program is the other side of the
   same coin, often asked about in almost the same breath.
   Almost all processors have such a facility.  It begs for
   standardization.  Yes, the interpretation of the command
   depends on the processor.  Nonetheless, this is a very useful
   and widely used facility.  It is also trivial to add to the
   standard, being just an additional intrinsic with no special
   interactions.

   This functionality is indirectly available via C interop,
   but I think it merits direct support, just like the intrinsics
   to obtain command-line information have direct support in
   the CD.

J3 Response:  We will suggest that it be added to the list of features for
consideration for the next revision of the Fortran standard.

References
02-007r3, Fortran 2000 Committee Draft

[End of J3 / 02-311r1]