Minutes of BCS Fortran Specialist Group Meeting held at
BCS HQ on 28 July 1988
Present: Martin Counihan - University of Southampton
Miles Ellis - Oxford University
Dave Griffiths - SSF
Mike Gunn - Surrey University
Andrew Harris - Z & S Consultants
Carol Hewlett - LSE
David Hill - MRC
Peter Holland - SS(E)L
Alan Jenyon - CAD Centre Ltd
Chris Lazou - ULCC
Brian Meek - Goldsmith's College London
Mike Nunn - CCTA
John Reid - Harwell
Lawrie Schonfelder - Liverpool University
Neil Smith - RAE Bedford
Julian Tilbury - University of Salford
Paul White - Met Office
John Wilson - Leicester University
John Young - MOD(PE)
l. APOLOGIES FOR ABSENCE
Apologies for absence were received from Les Russell, David Muxworthy
and Dave Vallance.
2. MINUTES OF THE PREVIOUS MEETING [19 April 1988]
The Secretary apologised for the late posting of the minutes due to
printing problems at BCS HQ. Because of this the change of date of this meeting
was not notified to members of the Group after the previous set date - it was
hoped nobody was inconvenienced.
The Chairman had received a letter from David Muxworthy criticising the
tone of 3(i) on the BSI Response to ANSI on Fortran 8x. It was agreed that the
word "misguided" be deleted from the last line but two and the phrase "BSI was
forced" a further 4 lines above be changed to "normal BSI policy in these
circumstances was".
The Treasurer pointed out that the interim statement of accounts referred
to in the minutes was, in fact, missing and Appendix A was the final statement
of accounts. The Secretary said both statements constituting Appendix A had
been sent to BCS HQ for mailing but the interim statement of accounts had been
removed. Members of the Group wishing to receive a copy should contact the
Secretary.
In item 4 the title Charter Engineer Status should read Chartered
Engineer Status.
The Secretary apologised to all the Non-BCS members of the Group for the
oversight of sending out last year's Membership Renewal form. Apologies were
also made to those BCS members who received renewal forms. These problems are
partly due to changes in the management of specialist groups at BCS HQ.
3. MATTERS ARISING
Chartered Engineer Status
The BCS had published a list of those members who had been given CE.
It contains nearly 2000 names. The Chairman said his application was still
being considered after being returned to BCS and then re-submitted by them.
BSI Response to ANSI
Lawrie Schonfelder said that the BSI (and the ISO) has serious disquiet
about ANSI not just in the field of Fortran 8x. ANSI seemed to favour
individual comments about standards and appeared to lack appreciation of
national comments which are not being taken seriously. The national public
comments on Fortran 8x was an example where X3J3 had apparently not taken
enough notice.
Closure of Deposit Account
Lawrie Schonfelder asked the Treasurer for the reason for closing the
Group's Deposit Account. Miles Ellis explained the Current Account was a
special Gold Account which would accrue interest like a Deposit Account.
4. BCS BUSINESS
Sponsoring Attendance of WG5 meetings
The Chairman had written to Roger Johnson of BCS requesting that the
Group sponsor a member to attend WG5 meetings once a year. The reply stated
that BCS Specialist Group money could not be used in this way. However, it may
be possible to fund attendance from other BCS funds.
Specialist Group Management Committee
The Chairman attended the meeting on 25 May and summarised the following
points from the minutes.
(i) VAT is to be paid on all meetings and conferences
(ii) The BCS logo is available for Apple computers
(iii) The Bulletin and Newsletter are to be merged
iv) There is a draft constitution for specialist groups
Evening Meetings
It was suggested that the Group use some of the profit made on the last
two Forums to set up evening meetings on Fortran.
Initially, any meetings should be held centrally in London but meetings
in other parts of the country could be organised jointly with local BCS
branches. Meetings should be free and should appeal to a different group of
people than that of the Group. It was suggested that somebody needed to
organise and publicise these meetings. Chris Lazou offered to set up an initial
meeting in London and was asked to bring a proposal to the next meeting on a
fixed program. Brian Meek proposed that Chris Lazou books a room in London and
then to let people know the date. He proposed that the next meeting decide the
content of the Evening meetings and suggested titles like "Fortran" and
"Converting to Fortran 8x".
A discussion followed on the format of the meeting. John Wilson suggested
that we agreed on a meeting at the end of the year which would be after X3J3
and WG5 meetings. He also suggested that two more dates are fixed in early/late
Spring 1989.
5. ITEMS RAISED BY MEMBERS
Fortran Compilers on PCs
Mike Gunn suggested that the Group might consider a scheme to evaluate
Fortran compilers for PCs. Some money could be spent on producing a sort of
"Which Report"on tested compilers. It was pointed out that Bath University was
already looking at compilers. Brian Meek thought the scheme would be labour
intensive and too expensive and that it was better to collect reports on
evaluation and provide information only.
In the discussion that followed it was suggested that an evaluation suite
of programs could be made and that there should be some form of bench-marking.
Brian Meek offered copies of reports and a checklist.
John Wilson proposed a sub-group which needed somebody with time and
money to set up an information centre and to make a simple set of programs. If
there was no volunteer the idea should be shelved.
In the further discussion that followed it was agreed that this was a
mammoth task and a long term project but it could start in a small way. Any
test deck of programs would be to do with quality rather than speed. It was
pointed out that there already were various groups comparing compilers
including the Salford undergraduate project on COBOL.
Contributions to the project could be invited from the Group and Carol
Hewlett asked for names of compilers with the appropriate computers from the
Group. This initial list is summarised in Appendix A.
6. FUTURE MEETINGS
The next meeting of the Group will be held on Thursday, 6 October 1988
when Dave Vallance would be invited to talk on his Portable Package Framework
and also to introduce his FTN77/386 - a PC Fortran Compiler.
The following meeting provisionally set for Thursday, 29 January 1989
would be on Fortran Pre-processors - Past and Future.
The AGM is due to be held on Thursday, 20 April 1989 at Oxford - please
note change of date.
7. ANY OTHER BUSINESS
Dave Vallance asked if his leaflet on FTN 77/386 could be included in
the next mailing to members. It was agreed that as there were several precedents
for this procedure which included covering the cost of mailing that this was
acceptable. It was to be made clear that the BCS and the Fortran Specialist
Group does not in any way endorse the product but was a useful way to inform
interested members as well as recovering the costs of mailing. The Secretary
was asked to produce a standard disclaimer which was to be included in all
future mailings.
John Young
Secretary, BCS Fortran Specialist Group
31 August 1988
[Appendix A was not attached to the minutes]
[this was not attached to the minutes but was circulated as a separate document
in September 1988; it records the debate held in the afternoon of the meeting]
Debate on the Future of Fortran 8X
1. Introduction
This report on the future of Fortran 8X by the Fortran Specialist
Group of the British Computer Society which took place on 28th July
1988 is a combined attempt by three members to summarise the major
points discussed at the meeting.
The debate was opened by Brian Meek who as Chairman of the ISO
Programming Languages Panel wanted the comments of the Group to send
to BSI.
The Chairman of the Group, John Wilson, then presented the background
to the current impasse on Fortran 8x.
2. Background
At the latest meeting of X3J3 in Urbana in May 1988, the committee
ended in deadlock. Having reviewed all 400 public review comments and
ISO country votes it identified a number of major technical tissues
and voted on them individually. It then drew up a package of
alterations to the 88 document based on these votes as follows:
- remove concept of Deprecation
- add Pointers
- add Bit Intrinsics
- remove RANGE & SET RANGE
- remove IDENTIFY & ALIAS
- simplify Generalized Precision
- add INCLUDE
- add DO WHILE
- add significant blanks in free source form
This was voted on as a whole: result (15-21-1) !
The committee was unable to come up with a scheme which obtained a 2/3
majority on straw votes. Therefore a number of policy proposals were
being prepared for the next meeting which will take place in Jackson
Hole in August (8-12).
3. ISO and the BSI
On the 20th June 1988 a meeting of the BSI Fortran Panel took place to
discuss the X3J3 impasse and to attempt to formulate comments to both
X3J3 and WG5.
Several of the new proposals were brief1y outlined and
discussed. However, at least three different proposals were supported
by those members of X3J3 present and it was not possible to come to an
overall consensus amongst the Panel members: in any case this was not
considered appropriate as none of the proposals had been properly
formulated and presented. Thus a rather bland statement has been sent
to X3J3 by David Muxworthy, convenor of the BSI Fortran Panel, along
the following lines:
- X3J3 should pay close attention to the ISO country comments and note
that even where the vote was "no" most countries were broadly in
favour of the proposals in 88.104.
- The package of alterations (referred to above) and voted on in
Urbana is very similar to the position advocated by both Canada and
the UK.
- The UK agree with the comments in a recent position paper from
Canada (June 1988) that there is a strong need for a standard
quickly.
- Individuals and countries are urged to be flexible in order to
produce something which will meet a substantial need now and provide
the basis for further evolution.
ISO WG5 meets in Paris in September (19-23) - it would be very helpful
to go to that meeting with something more substantial as a UK
position.
4. The Proposals
At the last count there were 11 separate proposals tab1ed for the
Jackson Hole meeting. It is not possible to describe them ell in
detail but they appear to fall broadly into 3 categories and it may be
possible for the Group to express a preference for one category. As
there were three representatives of X3J3 present, it was suggested
that each summarise one of the categories. As it turned out each X3J3
member broadly supported the category he was asked to present.
4.1 Category A: Simplified S8
John Reid is one of the people who has a proposal in this category to
be presented at the Jackson Hole meeting. The idea is to develop the
model voted on at Urbana retaining the basic structure of the 88
document. Pointers are added but the main aim is to reduce the size of
the language.
4.2 Category B: The Layered Model ("Onion Skin")
This is a joint proposal from Adams, Brainerd, Martin, B. Smith and
Wagener. The full language contains a "core" as a subset which
contains a "base" language as a subset. The Base is Fortran 77. The
Core contains features which do not require explicit procedure
interfaces. The full language adds Pointers and Bit Intrinsics as well
as significant blanks, all of which are based on existing
proposals. The essential difference between the core and full language
is that core features require only stack-based management whereas the
full language allows heap-based management. Lawrie Schonfelder
explained the basis of the "layered" or "onion-skin" approach.
4.3 Category C: 2 Languages
A joint proposal is being submitted by Phillips, Allison, Ellis,
Leonard, P. Smith, Weaver. Miles Ellis explained the thinking behind
this approach. The first language will be Fortran 77 plus those
features of S8 which are considered consistent with current Fortran
concepts emphasizing efficiency. The second language will be
incompatible with Fortran 77 and will be a modern language using the
new S8 features unshackled by the restraints of old Fortran (storage
association etc.).
5. The Debate
There then followed a lively discussion during which some of of the
points made were:
- Subsets had proved to be a disaster in the past: they must not be
introduced in the guise of a layered approach, as implementers will
probably base their implementations on, (on example» the "core"
model with selective extensions into the full language - with
disastrous effects on portability.
- Fortran 8x must look to the future: if it is accepted it will be the
basis of scientific programming well into the 21st Century and it
must, therefore, include features which will make it an efficient
language on the non-von Neumann architectures such as true parallel
machines, which will grow in importance.
- X3J3 is being too conservative: some items should be deleted in this
revision in order to establish the principle that features of the
language can be removed when they become obsolete and little used
(e.g. Assigned GOTO or PAUSE).
- If two new languages are defined instead of just one then only the
simpler one will be implemented in the short-to-medium term.
- Facilities in Fortran 8x should be chosen on the basis of their
benefit to the user, not for their ease df implementation; the
latter only has to be done once, the former by many users for a
number of years.
- A better approach to a two language Standard would be to keep
Fortran 77 as one language (possibly with the addition of MIL STD
1753) with the other one being a free-format, cleaner, slimmer
Fortran 8x with all redundant features of Fortran 77 removed.
6. Straw Votes
At the end of the meeting after several members had already left, two
votes were taken (with X3J3 members not allowed to vote!):
1) Which of the following approaches would be acceptable?
(Multiple voting allowed)
S8 as it is: 11
Simplified S8: 12
"Onion skin": 1
Two languages: 1
2) Which of the following approaches would be acceptable?
S8 as it is: 4
Simplified S8: 9
Undecided: 1
7. Summary
The Chairman made the following points:
- There was support for the Canadian plea for a spirit of compromise
and flexibility in order to get an agreement, and that this must
happen quickly if Fortran 8x as a whole is not to lose credibility.
- It was noted that WG5 and the Group had been pretty consistent over
the last two or three years in being generally happy with the SB
document rather than the radical departures from it as, for example,
the 1986 "compromise attempt". This position was reflected in the
straw votes. The "onion skin" model was seen as a re-hash of
core-p1us-modules and was not liked for the reasons given.
- Some form of language evolution is considered to be essential,
although it was agreed that the concept of deprecation in S8 was
unworkable. (This could be construed as supporting the 2 languages
proposals although the votes indicate that there was only one
supporter. If it is accepted then language 1 WILL be Fortran 77 and
language 2 WILL be Ada.)
8. Conclusions
It is very difficult to summarise an open debate in a way which is
likely to be useful to X3J3 and WG5. We have picked out a number of
points made, but of course in some cases a point was only made by 1
person and was probably not shared by the others present. The votes
however should give some indication of general feelings.
John Wilson
Miles Ellis
John Young
September 7, 1988