cdb@hpclcdb.HP.COM (Carl Burch) (12/08/88)
The below is Len Moss's (of Stanford Linear Accelerator Center) summary
of the November Fortran standards committee (X3J3) meeting. It is posted
here with his permission.
- Carl Burch
- - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -
Subject: Summary of 110th X3J3 Meeting, 13-18 Nov 1988
X3J3 met in Cambridge, Massachusetts from 13 through 18 Nov 1988. At
the beginning of this meeting X3J3 had not yet agreed to a plan for
revising the draft in response to public comments. Three different
plans had emerged from the previous meeting. The plans differed
considerably in the extent to which they reduced the overall size of
Fortran 8x; however, most members agreed that each plan represented a
net simplification. Unfortunately, none of the plans seemed close to
achieving the necessary 2/3 majority support.
In the meantime, fearing that X3J3 was deadlocked, WG5 (X3J3's
international counterpart) adopted its own plan. The WG5 plan
accepted some (though not all) of the simplifications which were
common to the X3J3 plans, but also added some features (e.g., pointers
and bits) which significantly increased the complexity (some of the
X3J3 plans had accepted these features as well, but offset them with
additional cuts).
In the opinion of some X3J3 members the WG5 plan represented little,
if any, net decrease in complexity. Nevertheless, in the interest of
achieving a single, world-wide standard, the committee agreed after a
brief debate to adopt the WG5 plan, at least as a starting point.
Most of the rest of the meeting was devoted to discussing, in some
cases modifying, and then voting on each of the items in the WG5 plan.
Nearly all the items were passed, at least in principle, but a number
were considerably modified and a few rejected. The major features of
the resulting draft are as follows.
* Remove:
_ Concept of deprecation
_ RANGE/SET RANGE
_ ALIAS/IDENTIFY
_ Square brackets
_ User-defined elemental functions
_ New form of DATA statement
_ Derived type parameters
* Change:
_ "Specified" REAL/COMPLEX precision with passed-on precision
("REAL(*,*)") to "selected" precision (i.e., one KIND=
parameter, no passed-on kind)
_ Interface blocks, to allow user-defined generic procedures
_ Array constructor syntax, to use implied-do for repetitions
* Add:
_ Pointers (but in a substantially different form from the WG5
plan)
_ MIL-STD 1753 bit intrinsic functions
_ Significant blanks in free form source
_ KIND= parameters to INTEGER and CHARACTER types
_ INCLUDE
_ Binary, octal, and hexadecimal constants (but, unlike WG5,
only in DATA statements)
* Still open: adding stream I/O (straw votes show substantial
agreement on adding this feature, but the committee has not yet
resolved what approach to take).
* Major differences between X3J3 and WG5 plans:
_ WG5 added a KIND= parameter to LOGICAL (intended to provide a
bit data type as well as other sizes of LOGICALs)
_ WG5 deleted internal procedures, mainly because this change
was common to all three X3J3 plans (presumably some X3J3
members had changed their minds on this issue in the interim).
_ WG5 changed "Host association" to "use association".
In response to pressure from WG5, from WG5's parent committee, SC22,
and from X3J3's parent committee, X3, a very tight schedule for
revising the draft was adopted. If X3J3 succeeds in meeting this
schedule, a new draft will be forwarded for the next round of public
review by 1 June 1989.