Minutes of the BCS Fortran Specialist Group Meeting
held at the Royal College of Midwives,
15 Mansfield Street, London on 26 October 1989
Present: Owen Brauley Redbridge Council
John Chelsom NAG
John Dyke HRC
Mike Geary NAG
Dave Griffiths SSF
Carol Hewlett LSE
Peter Holland SSL
Chris Lazou ULCC
Youngjo Lee IC
John Nelder IC
John Reid Harwell
Les Russell AWE
Tony Webster University of Salford
John Wilson University of Leicester
John Young PE-MOD
1. APOLOGIES FOR ABSENCE
Apologies for absence were received from Miles Ellis, E
Golton, Mike Gunn, Valerie Harmer, David Muxworthy, Mike Nunn and
Laurie Schonfelder.
The Vice-Chairman, Chris Lazou, opened the meeting in the
absence of the Chairman, John Wilson who arrived just after the
start. The meeting was being held next door to BCS HQ as the
rooms there had been double-booked. As this was the first time
a meeting had been held at this venue the various fire exits
were pointed out. It was announced that the morning session
would be Fortran business as usual and the afternoon meeting
would be two talks on Expert Systems.
John Wilson apologised for his late arrival due entirely to
the persistent late running of his train. He suggested that as
in the foreseeable future the Fortran business of the Group would
decline future meetings start at 11.00 am with coffee and
biscuits from 10.30 am.
2. MINUTES OF THE LAST MEETING [29 June 1989]
Under Financial Matters it had been stated that there was
likely to be an increase in the handling of VAT. The meeting
felt that this statement was ambiguous and that it was the work-
load that was likely to increase, not the amount or the value of
VAT.
In Appendix A on page 3 of 3 in the seventh line BSC should
read BCS.
3. MATTERS ARISING
The approval of an "honorarium" for a person, part time, to
look after the Group's VAT on a routine basis was postponed.
The meeting again discussed the form of the Group's 20th
Birthday Celebrations. Chris Lazou was asked to invite a keynote
speaker so as to give enough notice. John Wilson and Miles Ellis
were asked to consider various possibilities and to report back
to the next meeting.
The Secretary, John Young, had not received any summaries
of the talks on Fortran on PC's. He was asked to contact the
various speakers for further information.
The dates of the future meetings of the Group were
confirmed. The January meeting on Parallel Processing would be
organised and introduced by Chris Lazou. The details of the
April Meeting which will be the AGM and the 20th Birthday
Celebrations will be discussed at the next meeting. John Young's
suggestion of Members' talks for the June meeting was accepted
and discussed. With the closure of BCS HQ in London from 1 July
1990 it was decided to approach the Rutherford and Appleton
Laboratory for approval for a visit to the Atlas Building.
4. FORTRAN FORUM 89
The Chairman, John Wilson, reported to the meeting on the
recent Forum held in London and in Edinburgh. Attendance at both
venues was poor, in particular, at Edinburgh. It was felt that
the lack of publicity was a major factor but this was partly due
to the short notice of the public review period for the Fortran
8x draft. All members of the Group received adequate notice and
John Reid's personal note of the Vienna meeting also contained
details. Only Computer Weekly published the dates of the Forum
(but not the times) and there was a paragraph about it.
Regrettably, Computing did not even manage to publicise the event
in its Diary. There were other problems in Scotland in
publicising the Edinburgh Forum which contributed further to the
low turn-out.
The Forum went well with generally favourable comments. The
structure of the meeting was an overview of the new features
followed by the presentation of specific items and completed by
a general discussion. At the end of the Forum a series of straw
votes was taken. Of those present a large majority were in
favour.
David Muxworthy, as Chairman and convenor of the BSI Fortran
Panel, appealed for written comments. He also announced the
closing dates of the public review period for the Draft standard
as follows:- X3J3 24 November 1989
BSI 15 December 1989
ISO 1 February 1990
The BSI Fortran Panel is due to meet on the 3 January 1990 to
decide the British response to the Standard. Last time the Panel
voted NO and gave its reasons. This time the response will be
YES with or without comments.
It was then stressed, that copies of the Draft Fortran
standard were still available from BCS HQ as issue of the
document was not linked to the Forum 89 as had been the case for
Forum 87.
5. REPORTS FROM X3J3 AND WG5 REPRESENTATIVES
X3J3 Representative John Reid reported on the meeting in
Vienna by reviewing Appendix A of the last Minutes. The meeting
was relaxed and consisted mainly of editorial corrections. He
reported that there was still work needed on S8. There had been
some discussion on character sets which may be non-portable and
possibly not be included in the Standard.
WG5 Representative John Wilson reported on the meeting at
Ispra again by reviewing Appendix A of the last Minutes. The
meeting was also relaxed with only formal meetings. WG5 had
requested that X3J3 limit the changes made to S8 and not to
impede the progress of the standard. Time was spent on
discussing character sets and it was decided that no special
characters would be allowed in variables. The future of the link
between WG5 and X3J3 was also discussed.
John Reid had just returned from an X3J3 meeting at Estes
Park in Colorado and reported to the meeting (see Appendix A of
these minutes). This meeting was an extra, ad hoc meeting and
not a formal meeting. This had been postponed until January and
the February meeting had been cancelled. Hence, the meeting was
attended by only about half the X3J3 membership and they were not
able to pass any motions. However, a good working week was
achieved on the draft standard.
The big item of news was the acceptance by X3 (the parent
body of X3J3) of a proposal to retain Fortran 77 as a separate
ANSI standard. This meeting was during the same week as the
X3J3 meeting and, in spite of 3 suggestions for compromise from
X3J3 members, X3 determined not to discuss the proposal and
"bulldozed" the item through. This was a complete reversal of
an X3 decision taken a year ago. It was just possible that SC22
(the parent body of WG5) could make the same decision but was
thought unlikely. The X3 decision means that Fortran 8x will
have a new standards number. It is now possible to institute a
program of work to propose extensions to F77 which will not be
compatible with 8x. In reality, ANSI will have two separate
standards for Fortran.
John Reid felt that the US had taken this decision because
contracts from the DOD required a standardised language. Fortran
8x would not be standard for some time but F77 had been renewed
in the interim. The meeting asked whether it should make a
statement about this situation but it was felt that it was not
appropriate as it was an American problem. The only appropriate
bodies for the Group to make a statement to were BSI and X3J3.
The Vice-Chairman, Chris Lazou, pointed out that the Forum 89 has
unanimously voted for one standard and suggested that the Group
should go on record as supporting only one standard.
Then there was an appeal to individual members of the Group
to submit comments on the draft Fortran 8x standard.
John Reid then returned to the report of Estes Park meeting
and covered the following technical points. There appeared to
be a problem with copy-in and copy-out arguments which were
Pointers and Targets. The letter ballot on the draft standard
to X3J3 members would be formally passed by January. X3J3
members Jeanne Martin and Laurie Schonfelder had re-written the
rules of Structures and Commons. COMMON now allowed a mixture
of Data Types provided that all variables matched exactly.
Equivalences would also be allowed in this case.
There then followed a discussion on Pointers.
6. ANY OTHER BUSINESS
Closure of BCS HQ
The Chairman, John Wilson, reported to the meeting that the
British Computer Society is to move its headquarters to new
premises in Swindon in June 1990. The lease on 13 Mansfield
Street expires on 30 June 1990. It is proposed that a new
meetings room will be retained somewhere in Central London and
will be available from 1 September 1990. Hence no room bookings
can be taken for July and August 1990. There would be meetings
rooms in the new HQ in Swindon.
Specialist Group Management Committee
The following new Specialist Groups were approved:
Computer Conservation Society (in conjunction with the
Science Museum)
Software Testing Group
FOCUS - an Affiliated Group run by the Federation of Army
Wives
The Software Engineering Specialist Group which was "laid
down" in the summer has been "resurrected".
Technical Committee
This committee has been re-vamped with restricted membership
which does not include Fortran.
Mailing Lists
The use of the Group's mailing list by commercial
organisations can only be made with the approval of the Chairman.
Publicity
Press statements made by members of the Group which might
be interpreted as being on behalf of the Society will need BCS
approval.
Resolution
John Reid proposed the following resolution which was
carried unanimously.
"This Group unanimously abhors the X3 decision to have two
ANSI Fortran standards and urges BSI to oppose any
corresponding ISO proposal".
Endorsement
The Group unanimously endorsed the following votes taken at
Fortran Forum 89.
(i) Press for early ratification of Fortran 8x.
(ii) Want one world-wide standard.
(iii) Reject Fortran 77 as a separate standard.
7. DATE OF NEXT MEETING
The next meeting of the Group will be held on Thursday, 25
January, 1990 at BCS HQ with the afternoon session on Parallel
Processing.
John Young, Secretary
12 January 1990
To: Fortran Forum, BCS, NAG, etc.
From: John Reid
Date: 25th October 1989
Subject: X3J3 meeting in Estes Park
Note: This is a personal note on the meeting and in no sense does it
constitute an official record of it.
The November meeting of X3J3 has been postponed to January so that
the comments from the public on the second draft standard will be
available. The Estes Park meeting was called to work on the editorial
state of the document and fix holes and inconsistencies, on the
. assumption that the January meeting will be dominated by processing
comments from the public. It was an 'ad hoc' meeting that passed no
formal motions. The several hundred changes that were agreed will be
submitted to the full membership in a letter ballot.
I raised a problem that my colleague Philip Sweetenham noticed in
connection with procedure arguments having the attribute TARGET. Since
the Standard has always been phrased to allow implementations to
perform copy-in copy-out of arguments, a pointer that is associated on
entry with a target actual argument need not be associated with the
corresponding dummy argument and a pointer that is associated with the
dummy argument on return need not be associated with the actual
argument. This situation was appreciated and my wording accepted. I
was asked to add an example showing that if it is required to preserve
the associations on entry and return, the target should be replaced by
a pointer to it.
I participated in a significant clean-up of the description of storage
association, structures in COMMON and EQUIVALENCE, and pointers in
COMMON. We decided that in addition to numeric and character storage
units, there should be a different 'unspecified' storage unit for
nonpointers of each combination of type and type parameters not
present in Fortran 77 and for pointers of each combination of type,
type parameters, and rank. Common blocks are permitted to contain
mixtures of kinds of storage units, provided the sequences of kinds
match exactly in each case. Equivalences are allowed when the kinds
match.
A clarification of the rules for generic procedures was
accepted. No renaming of generic procedures will ever occur and two
interface blocks bearing the same name will be treated as one. This
will allow, for example, a module containing a definition of SIN for
interval arithmetic to be used at the same time as a module containing
a definition of SIN for very long precision.
A problem connected with optimization in the presence of pointers was
identified, but no solution adopted because it appears to need a
technical change to the standard. It is illustrated by the sequence of
statements
T = 1.
P = 2.
Z = T
where T is a target and P is a pointer. If it is known that P is never
associated with T, the value of T can be retained in a register for
the assignment to Z. We considered some kind of OVERLAP statement to
allow the programmer to specify overlap and non-overlap sets.
During the same week, there was a meeting of the parent committee X3,
which considered the proposal to retain the ANSI Fortran 77 standard
and change the Fortran 8x program of work to the production of a
separate standard. X3J3 was not consulted and a letter ballot showed
that more than 2/3 of the members disapproved. Jerry Wagener attended
the meeting on behalf of the Chairman. The proposal had been approved
by a comfortable majority in an X3 letter ballot and X3 agreed to
reconsider the matter with great reluctance. Jerry made a
presentation, but the discussion was truncated by a vote to 'call the
motion' and it passed comfortably. I see this as an insult to X3J3 and
the international community. The unwillingness to debate the issue
adequately and seek an alternative solution that would have wider
acceptance (Jerry went with several suggestions) strikes me as
thoroughly unprofessional. It was only a year ago that X3 gave a
directive to X3J3 that 'the US support a single Fortran standard
(domestic/international, i.e. one document)'.