Minutes of BCS Fortran Specialist Group Meeting held at
the University of Oxford's Computing Teaching Centre,
59 George Street, Oxford on 20 April 1989
Present: John Appleyard Polyhedron Software
Martin Counihan Southampton University
John Dyke Huntingdon Research Centre
Miles Ellis Oxford University
E Golton RAL
Dave Griffiths SSF
Les Hatton Programming Research Ltd
Carol Hewlett LSE
Peter Holland SSL
Mike Nunn CCTA
Les Russell AWE
Paul White Met Office
John Wilson Leicester University
John Young PE-MOD
1. APOLOGIES FOR ABSENCE
Apologies for absence were received from:
Mike Geary NAG
Anne Gooding Air Products plc
Mike Gunn University of Surrey
Valerie Harmer University of Surrey
Chris Lazou ULCC
R I McLeod NEL
Brian Meek Goldsmiths College
David Muxworthy University of Edinburgh
John Reid Harwell
Lawrie Schonfelder Liverpool University
Dave Vallance Salford University
2. WELCOME TO THE COMPUTING TEACHING CENTRE
The Treasurer, Miles Ellis, who is also the Director of the
CTC welcomed the Group to the Centre. He first presented the
Fire Drill regulations, then announced that coffee was available
free, courtesy of the Centre, and described the luncheon
arrangements. He also invited members to look around the
building if they so wished.
The Chairman, John Wilson, then reviewed the plans for the
afternoon programme. These were a demonstration of FLINT by Les
Hatton who had yet to arrive, an informal talk by the Secretary,
John Young, and a conducted tour of CTC by Miles Ellis. John
Wilson then asked the Group if there was any other business to
be discussed at the end of the meeting.
It was noted that a number of London members of the Group
were attending the meeting in spite of the strike on the London
Underground.
3. MINUTES OF THE PREVIOUS MEETING [19 January 1989]
The Chairman, John Wilson, commented on the Programming
Research Ltd flier on FLINT. He said that the Group was prepared
to support products like FLINT providing that the advertising was
not overtly commercial. The FLINT flier (and the SALFORD 386
Fortran compiler flier) for which the companies paid the Group's
mailing costs seemed to be about right in presentation and the
companies seemed to be happy with the arrangements. The flier
for the Fortran Questionnaire for Diane Meehan of Liverpool
University was included free of charge.
4. MATTERS ARISING
Fortran Compilers on PC's
The Sub-Group convener, Mike Gunn, was unable to be present
at the meeting due to a prior commitment in another part of
Oxford. He had hoped to prepare a short summary of the
information he now holds but sent apologies to the meeting via
Carol Hewlett for not having had the time. He was able to report
that there are about 8 reviews, a couple of manuals, the list of
verified compilers, and some compiler manufacturer's literature.
The Chairman, John Wilson, suggested that Mike might be able to
report his work at the next meeting. (Secretary's note: Mike
Gunn is prepared to present his summary to the Group at the June
meeting).
The IUSC and CHEST evaluation of Fortran Compilers had been
completed without any single compiler fulfilling all the
requirements. John Wilson had received a message from Lawrie
Burbridge at Bristol University giving details of three compilers
that came closest but for one reason or another could not be
recommended as the sole Fortran compiler for University use. The
three compilers are the Ryan McFarland (RM), the Prospero (PRO),
and the Salford 386 Fortran Compilers. John Wilson stressed that
these recommendations were for the Academic community only.
There followed a general discussion on Fortran Compilers on
PC's with some time spent on the problems caused by a lack of
space on PC's.
Scottish Sub-Group
The Secretary, John Young, reported that he had formally
written to John Bruce. He had no further news to report.
Fortran Forum
The Chairman, John Wilson, led a discussion on when and
where this year's Forum should be held. It was agreed that it
would have to be held within the final review period of the
Fortran 88 Standard which was likely to be for two months. This
Forum would be the last for the current standard and its main
task would be to seek approval for Fortran 88. It is also hoped
that there could.be some time for discussion of the new standard.
It was decided that there would be two venues for the Forum
- one in London and one in Scotland. It would be useful to hold
them within days of each other and to publicise them jointly.
If the Forum was held in early to late September a University
could be an appropriate venue. The Group would organise the
venue for the London meeting and it was hoped that the Scottish
Sub-Group could organise the meeting in Scotland.
Miles Ellis, being the sole representative of the X3J3
members, reported on the meeting in Stanford. He updated John
Reid's report which was circulated to members in March as
Appendix A to the last set of minutes. He described how the
procedures of X3J3 had been completely changed at the last two
meetings. X3J3 had laid aside their normal way of working in the
face of pressure from WG5 (and X3) to get the new draft of
Fortran 8x ready in the timescale wanted by WG5. X3J3 had felt
there was too little time to get the draft ready but had adapted
to the circumstances. The latest draft of S8 now had a two
colour, red and green, cover to symbolise the unity of X3J3 and
WG5 on the new standard.
Commenting on John Reid's report and item 4 Miles Ellis said
that nonadvancing input/output was X3J3's way of describing
stream i/o. This feature needed to be defined in a watertight
way. He also pointed out that item 5 Structures in COMMON and
EQUIVALENCE the derived data types in COMMON equivalencing would
be a "non-portable" feature.
Miles Ellis then went on to report the result of the X3J3
informal letter ballot on forwarding the latest draft of S8, and
also the result of the WG5 informal ballot. In the X3J3 vote
there were 31 yes, 11 no, and 2 abstentions with computer vendors
split about 50/50. He also reported that the IBM representative
had done a complete "U-turn" and voted yes. In fact, IBM had
decided to support the new standard because of its concern over
the possibility of multiple Fortran standards worldwide and had
requested X3J3 to circulate its letter explaining the reasons for
its yes vote. The X3J3 vote gives the necessary two-thirds
majority for acceptance of the current draft to be the basis of
the new standard. In the WG5 vote, 9 countries presented votes
which were all yes.
Miles Ellis then said that at the next meeting of X3J3 in
May there would be a final ballot for the approval of the final
draft. This then passes to X3 for approval to be followed by,
probably, a two month public review period. (See Appendix C).
There followed a general discussion of various points
raised. The meeting felt that the form of POINTERS needed to be
decided. There was worry about unbounded pointers, in particular
that they cannot be optimised. E Golton said that the X3J3's
proposal on pointers was clever but felt that a pointer should
always point to a specific object. He felt that the main use of
pointers would be as a dynamic arrays facility. The meeting was
told that array constructor symbols square brackets [,] would
be an alternative to (/,/) which would now be the standard. The
problems with character handling in more than one byte of
computer storage had yet to be resolved. The meeting noted that
X3J3 had not yet accepted the new standard as Fortran 88 but
still uses 8x. It was also noted that deprecated features had
a lifetime of 10 years.
The meeting then speculated on the publication of the first
Fortran 88 working compiler. It was felt that the first
compilers would, of necessity, be Fortran 77 compilers with
Fortran 88 as an enhancement and would probably be on betatest
by mid-1990. It was felt that it could be several years before
the first true Fortran 88 compiler would be written.
6. BCS BUSINESS
John Wilson reported that he had been unable to attend the
recent Specialist Group's Management Committee meetings and
therefore had nothing to report. He also reported that the Vice
Chairman, Chris Lazou, had attended a recent meeting of the
Technical Committee which was considering winding up.
The Secretary, John Young, reported that he has received two
memos from BCS concerning the Branches and Specialists Groups.
It was noted that advertising space in Computing and Computer
Weekly is free on a first-come first-served basis.
(It was decided to break for lunch at this point and to
reconvene the meeting in the afternoon for the AGM).
7. ANNUAL GENERAL MEETING
Chairman's Address
The Chairman, John Wilson, reviewed the Group's past year.
There had been four meetings all held at BCS HQ with an average
attendance of 18. The topics covered included ADA, the Fortran
88 Debate, Toolpack, the Salford 386 compiler, and Post-
processors. The significant development of the year was the
formation of the Fortran on PC's and Scottish sub-groups.
Looking forward he said the Group needed to work towards the
Fortran Forum and to set the dates of the meetings for the coming
year with subjects. Looking back he said David Muxworthy had
informed him that the Group was formed in 1970 and had sent a
copy of the initial meeting's minutes. These were read out to
the meeting and discussed. (Secretary's note: it was discovered
after the meeting that Peter Holland was at this first meeting).
Turning to this year's programme it was decided that next
year's AGM would be followed by an informal celebration of 20
years of the Group with, perhaps, a speaker from the original
Group. It was also likely that it would be time to celebrate the
completion of the new Standard Fortran F88.
Treasurer's Report
The Treasurer, Miles Ellis, presented his interim statement
of accounts. He explained that the Group has an Internal and
External Account: the Internal Account is used to cover costs
incurred within the BCS and the External Account was money that
the Group used for normal income and expenditure. The interim
statement of accounts is Appendix A1 and the final statement of
accounts is Appendix A2.
Election of Officers
There being no nominations for the four posts of Chairman,
Vice-Chairman, Secretary and Treasurer and the present Committee
being prepared to serve again they were duly elected.
Chairman - John Wilson
Vice-Chairman - Chris Lazou
Secretary - John Young
Treasurer - Miles Ellis
Any Other Business
It was decided that the Group's membership of X3J3 which had
lapsed would now be terminated.
The Netherlands Fortran Group which regularly receives the
Group's minutes had written to ask whether the Group would like
to receive their minutes. The meeting decided that they would.
Demonstration of FLINT
The Managing Director of Programming Research Ltd., Les
Hatton, had set up a demonstration of FLINT on a SUN computer in
the University of Oxford's Computing Department. He was able to
demonstrate the capabilities of the FLINT package for the static
analysis of Fortran programs. He showed that "bugs" which are
undetectable by normal compilation methods could be found in any
amount of Fortran code.
Secretary's Informal Talk
This talk was cancelled due to the over-running of the AGM
and the time taken to walk from the CTC to the Computing
Department for the FLINT demonstration.
Conducted Tour of the CTC
Miles Ellis showed two members of the Group around the CTC
and described the main functions of the Centre.
John Young, Secretary
16 June 1989
Appendix A1
British Computer Society Fortran Specialist Group
Provisional Accounts 1988-89
A) Internal Account
600 Budget Allocation for 1988-89 600.00
Expenditure
Minutes (July) 60.31
(Sept) 87.99
(Jan) 52.55
214 200.85
43 Catering (July) 12.50
Speakers'
Expenses (Sept) 16.40
(Jan) 182.30
26 198.70
300 412.05 (412.05)
300 Surplus returned to HQ 187.95
B) Current Account
Balance at 30/4/88 2014.11
Income
10 Subscriptions 110.00
80 Advertising 200.00
Via HQ 22.25
Transfer from Gold 100.00
432.25
Expenditure
Transfer to Gold A/C 2000.00
Sec's Expenses 24.30
WG5 rep. to X3J3
(contribution) 100.00
2124.30
Nett deficit 1692.05 (1692.05)
Provisional Balance at 30/4/89 322.06
C) Gold Account
Balance at 30/4/89 NIL
Income
Transfer from
Current A/C 2000.00
Expenditure
Transfer to
Current A/C 100.00
Nett Surplus 1900.00 1900.00
Provisional Balance at 30/4/89 1900.00
2014 PROVISIONAL TOTAL ASSETS AT 3O/4/89 2222.06
T. M. R. Ellis
Treasurer
18th April 1989
Appendix A2
British Computer Society
Fortran Specialist Group
Final Accounts 1988-89
A) Internal Account
600 Budget Allocation for 1988-89 600.00
Catering
(July) 12.50
(Oct) 25.00
(Jan) 25.00
43 62.50
Newsletter/Minutes
(July) 60.31
(Sept) 87.99
(Jan) 52.55
(Mar) 115.99
(Apr) 42.63
214 359.47
Speakers' Expenses
(Sept) 16.40
(Jan) 182.30
26 198.70
300 620.67
300 Surplus returned to HQ/(Deficit) (20.67)
B) Current Account
373 Balance at 30/4/88 2014.11
Income
10 Subs. 110.00
Via HQ 22.25
1028 From Gold a/c 100.00
1668 232.25
Expenditure
To Gold a/c 2000.00
27 Secy's Expenses 28.65
WG5 rep. to X3J3
(contribution) 100.00
2128.65
1641 Excess of income over expenditure (1896.40)
2014 Balance at 30/4/89 117.71
C) Gold Account
962 Balance at 30/4/89 NIL
Income
From curr. a/c 2000.00
66 Interest 180.50
2180.50
Expenditure
1028 To current a/c 100.00
(962) Excess of income over expenditure 2080.50
0 Balance at 30/4/89 2080.50
2014 TOTAL GROUP ASSETS AT 30/4/89 2198.21
Note that comparative figures for 1987/88 refer to Lloyds bank deposit
account and not to Gold account. The deposit account was closed at the
end of 1987/88 and the balance transferred to the current account. At
the beginning of 1988/89 £2000 was placed in the (high-interest) Gold
account.
T. M. R. Ellis
Hon. Treasurer
11th June, 1989.
[There is no Appendix B]
To: Fortran Forum, BCS, NAG, etc.
From: John Reid
Date: 24th May 1989
Subject: X3J3 meeting in New York
Note: This is a personal note on the meeting and in no sense does it
constitute an official record of it.
1. Summary
This meeting followed a letter ballot of all X3J3 members on
forwarding the draft standard to X3, which passed 31-1 1-2 (see
Section 2 for further details). It was a gruelling meeting; on two
days, we had additional evening (8-10 p.m.) sessions in full
committee, several subgroups met after the evening party at our host's
house, and many individuals wrote up proposals in the small hours. The
Chairman set the objective of processing all the ballots and
pre-meeting documents (with priority to those voting NO, in the
hope of achieving closer consensus), and agreeing responses to the 462
letters received from the public comment period. Vast numbers of
editorial changes and bug fixes were approved. I summarize the more
significant ones in section 3. The Chairman's ruling on limiting
discussion to such items held well; the only exception was in
connection with character handling (Section 4), for which there were
extenuating circumstances. A roll-call vote near the end of the
meeting approved (33-12) forwarding the revised document.
Because of the NO votes and because a straw vote of X3J3 favoured a
4-month review period (because the changes are so extensive), a vote
of X3 will be needed before preceding to public comment. It was
suggested that this might be taken at the June 20-21 meeting of X3,
rather than by letter ballot which is the usual procedure. The comment
period should commence a few weeks after this.
2. X3J3 letter ballot
A letter ballot on forwarding S8.111 was held prior to the
meeting. There were 31 YES votes, 11 NO votes, and 2 did not
vote. The following voted NO:- Allison (Microsoft), Katz (DOD),
Lammers (Cray), Leonard (Harris), Marusak (Los Alamos), O'Gara
(Supercomputer Systems), Philips (Boeing), Rolison (Unisys), P. Smith
(Convex), Whitlock (DEC), and Yan (Data General). In most cases, the
predominant reason given was the size and complexity of the
language. There was also criticism of the state of S8 following the
large number of recent changes, the use of % for structure
qualification, lack of a BIT data type, and the presence of module
procedures. This time, Weaver (IBM) voted YES, because of overriding
concern that there be one new worldwide Fortran standard.
All those voting NO in the letter ballot also voted NO in the
roll-call vote at the end of the meeting; in addition, Lahey (Lahey)
voted NO instead of not voting. All those voting YES in the letter
ballot also voted YES in the roll-call vote, except for one absentee;
in addition, two new members voted YES as did one member whose letter
ballot was mislaid.
3. Bug fixes
The following bug fixes are of note:-
1. Undefined allocatable arrays. Because of implementation
difficulties, it was decided that automatic deallocation of an
allocatable array (on return from the scoping unit in which it
declared or on a return that leaves no executing procedure accessing
the module in which it was declared) be replaced by its allocation
status becoming undefined. An undefined allocatable array must not be
allocated, deallocated, referenced, or defined. My personal view is
that this fix is abhorrent. A far better solution (safer and simpler)
would be for module data always to be saved, but the Chairman viewed
this as breaking her guidelines.
2. Pointers in COMMON and EQUIVALENCE. It was confirmed that pointers
are permitted in COMMON but not in EQUIVALENCE.
3. Non-default character constants. It was decided that nondefault
character constants should have their kind parameters at the
front. This is inconsistent with the other types, but having them at
the rear leads to severe parsing problems.
4. Recursive entries. It was decided to regularize procedures defined
by ENTRY statements. The syntax has been extended to allow a RESULT
clause and they are now permitted to be called recursively directly as
well as indirectly.
5. IMPLICIT NONE. It was decided to make it more apparent that
IMPLICIT NONE must precede PARAMETER statements.
6. ALLOCATABLE, POINTER, and TARGET statements have been added to
complete the set of statements for specifying attributes.
7. Zero-sized objects. Zero-sized objects are now permitted to be
initialized, so that they can appear throughout the language.
8. Pointer functions. Function results, like dummy arguments, may be
pointers.
4. Character handling
Dick Weaver and Miles Ellis both attended a special ISO meeting on
character handling. This was called because the ISO world has become
increasingly concerned with the problems of multiple character sets in
programming languages, especially where one of these requires more
than one byte per character. The ISO meeting came to the firm
recommendation that programming languages should address these
problems. X3J3 therefore made the following changes to S8:-
1. The set of letters available for identifiers is 'user-defined',
that is, the processor must provide a means for the specification to
be supplied and must accept it. The default set of letters is A-Z and
any user-specified set must include the default set.
2. The processor-defined character set, available for use in character
contexts, consists of 'graphic' characters and 'control'
characters. In free source form, control characters are not allowed in
character contexts; in fixed source form, the processor may restrict
the use of some or all of the control characters.
5. Responses to public comments
Responding to all the public comments was a massive undertaking. 6713
items were identified in 462 letters. Each commentor will receive a
cover letter, a summary of the recurring themes (as seen by the
subgroup heads), a summary of the major changes, an annotated copy of
his or her letter, and listings of the relevant replies from the
Committee's database.
6. Address by Bill Rinehuls
Bill Rinehuls, Chairman of SPARC (Standards Planning and Requirements
Committee), attended for two days of the meeting. He explained to the
Committee that, as a result of the letter ballot, consensus had been
reached and that the decision at the end of the meeting would be
between forwarding S8 as it was at the beginning of the meeting or
forwarding the version with all the changes made during the
meeting. This interpretation did not seem to be acceptable to those
voting NO in the letter ballot; they all continued to vote NO, and I
really cannot believe that they preferred to earlier version of S8. He
also said that X3J3 can continue to make changes after forwarding
the document, and make their incorporation a condition of the US ISO
vote, so the need for further editorial polishing was not a reason for
delay.
He explained that the extent of complaints that X3 has received about
X3J3 and the detailed management that has been imposed are both
unprecedented. X3 has required detailed reports from each meeting, has
instructed that processing be completed by this meeting and that there
be one national and international standard.
7. Next meeting of WG5
The next meeting of WG5 will be in Ispra, 10-14 July.
8. Next meeting of X3J3
The next meeting of X3J3 will be in Vienna, 17-21 July. The deadline
for the pre-meeting distribution is 12 June.
BCS FORTRAN SPECIALIST GROUP - MEMBERSHIP RENEWAL FORM
Your subscription fee-for membership of the BCS Fortran Specialist
Group is now due for the year 1989-90. Please complete as below :
* I am now a BCS member. My membership number is :
* I wish to continue my membership of the BCS Fortran
Specialist Group and enclose a cheque or postal order
for £5 made payable to "BCS Fortran Specialist Group".
* I no longer wish to be a member of the BCS Fortran
Specialist Group.
* Delete as applicable.
Please send this completed renewal form (and subscription
if appropriate) to :
Karen McPhi1emy,
Specialist Groups Executive,
British Computer Society,
13 Mansfield Street,
LONDON, WlM 0BP.
Telephone 01-637-0471