British Computer Society Fortran Specialist Group

Minutes of a meeting held on 7th June, 1976 in the Staff Common Room at the

Polytechnic of Central London, 115 New Cavendish Street, London W.1.

Present:

Mr. B. H. Shearing (Acting Chairman)   Alcock, Shearing and Partners

Mr. J. P. Arslett                      Shell International Petroleum

Mr. G. N. Ashley                       I.C.L.

Mr. J. D. Beasley                      Rothamsted Experimental Station

Dr. A. C. Day                          U.C.L.

Mr. M. R. Dolbear                      B.P.    

Mr. J. L. Dyke                         Huntingdon Research Centre

Mr. P. J. Hansom                       Honeywell I. S.

Mr. D. Hill                            C.C.F.

Dr. I. D. Hill                         M.R.C.

Mr. D. J. Holmes                       Rolls Royce (1971) Ltd.

Dr. J. Hutton                          SRC Rutherford Laboratory

Mr. R. C. Hutty                        Leicester Polytechnic

Mr. J. R. Jones                        Liverpool City Council

Mr. M. Lewis                           Imperial College Computer Centre

Mr. P. Loftus                          Marconi-E1liott Avionic Systems Ltd.

                                       Rochester.

Mr. G. N. Mullery                      Ferranti Ltd., Bracknell

Dr. J. Murchland                       Traffic Studies, U.C.L.

Mr. K. Normington                      Lanchester Polytechnic

Mr. T. D. Palmer                       Computer Power

Mr. R. W. S. Rodwell                   I.C.L.

Mr. P. A. Clarke (Secretary)           Rothamsted Experimental Station

Apologies for Absence:

Mr. D. Muxworthy (Chairman)            University of Edinburgh

Mr. G. Harding                         Leeds University

Mr. Shearing took the chair in the absence of Mr. Muxworthy who was unable to be

present.

l.    Approval of Minutes

The minutes of the 5th of April were approved subject to the following corrections:

(l) The name of Dr. D. Clifford of Digital Equipment Company should have

appeared among the list of those present. (2) Mr. Greenwood should have

been Dr. Greenwood. (3) On page 2 'complemented' should have been 'complimented'

and 'concensus' should have been 'consensus'.

2.    Matters arising from the Minutes

Item 2:     Pauline Walters of the BCS headquarters staff had checked with the

accountant and informed the secretary that subscriptions to specialist

Groups do not count for income tax relief.

3.    Election of vice-chairperson

No names were forthcoming for the position of vice-chairperson. Any person willing

to take on this duty should contact one of the officers. Mr. Shearing has agreed to

act as the groups representative on the BCS Specialist Groups Committee.

4.    Activities of other Fortran Groups

4.1   ANSI X3J3 Activities and discussion of public Comments sent to ANSI.

Mr. Shearing described the main points raised at the February l0-12 and

March 31 - April 2 Meetings of the ANSI X3J3 Committee. These included

changes to the syntax of the PARAMETER statement to permit constant expressions

and typing according to normal Fortran rules. And an extended Gw.dEe

Format edit descriptor which acts as Fw.d or Ew.d.Ee depending on the

magnitude of the datum.

The next X3J3 meetings are scheduled for l2 July 1976 - to discuss public

comments, and for September 1976 when, assuming all comments are resolved,

the new standard will be adopted. Mr. Shearing then asked Mr. Lewis

to describe comments received from the public.

Mr. Lewis said that he had received six letters and immediately prior

to the meeting had been given a large list of comments made by BSI. He

had not had time to study the latter but he would try to extract some of the

more important points. Dr. I.D. Hill who had chaired the BSI meeting gave

a brief description of it (see 4.4.5 below). The comments were then

considered individually and a vote taken by those present either on the

comment as it was or on a principle indicated by several comments. A vote

was taken on the proposed IF-THEN-ELSE construct. The resulting votes

(see Appendix A) were sent to X3J3.

Dr. Day said that he had been invited to a BCS Standards Committee meeting

(See 4.4.6 below) to describe the proposed ANSI Fortran Standard and the

comments made by the working party to review the standard.

Mr. Shearing said that the BCS Standards Committee had mounted a duplicate

exercise to that of the Fortran Specialist Group to obtain public comments on

the proposed standard. This was because, at our February meeting (Item 4.1)

no-one was prepared to take any action to form a consensus view on the

Standard - in the form of a summary of public comments. Mr. Muxworthy

had informed the Standards Committee of this, together with the warning that it

would take too long to produce such a document before the June 30th deadline

for public comments. This obviously had not been sufficient to deter the

Standards Committee from such an exercise.

Mr. Shearing said that he would try to investigate the relationship of

the Standards Committee to Specialist Groups at the next Specialist

Groups Committee.

Mr. Clarke said that several members of the group had been promoting the

draft standard and these included Dr. K. V. Roberts (of Culham Lab.) who

has given a talk on 'The New Fortran Standard' at a Symposium on Software

held in Belfast on 8th April, Mr. P. A. Clarke (Rothamsted) gave a talk

to Agricultural Research Council computer users on 'Fortran Developments'

at Rothamsted on 12th April, Mr. R. Hutty (Leicester Polytechnic) had

held a seminar at the Polytechnic to discuss the 'new draft standard'

during the first week in June, Dr. I. D. Hill and Mr. W. S. Hilder are

to give a seminar on 'The proposed new Fortran Standard' at Clinical

Research Centre of the MRC on 23rd June.

P.S. X3J3 have extended the public comment period to 28 September, 1976.

4.2 Fortran Development Committee

We have received the April edition of the FOR-WORD newsletter (Vol 2 No. 2

pages 11-16). This contains details of the IF-THEN-ELSE proposal,

details of the X3J3 meeting of March 31-April 2 and correspondence.

The calendar of events in the USA includes: A panel discussion on "Why

(or IF) Fortran will survive" to be held at the NCC, New York on 7-10

June, participants include P. Schneck, F. Engel, A. Ralston and L. Meissner.

A meeting on "Highlights of the New Fortran Standard" will be held at the

ACM 1976 Annual Conference, Houston on 20-22 October.

4.3 CODASYL FDBML Committee

The minutes of the 11th meeting held on 3-5 March l976 have been received.

The main items discussed include (i) adoption of standard Fortran rules for

comments, margins and continuation for DML statements (ii) Improvements to the

ALIAS statement (which can be used, for example, to map COBOL names to

Fortran names) in the DDL.

Good progress has been made on the Fortran database facility and the latest

draft (version 0.5) has been received from Dr. Chester M. Smith, the

chairman of the committee who indicates that comments on this on a point-

by-point basis would be welcomed.

It is the impression of the committee that the first version (l.0) of

their Journal of Development will be available and complete after meeting

number 13 in July 1976.

4.4 Other Groups

4.4.1 International Standards Organisation (ISO/TC97/SC5)

A letter has been sent from the ISO Secretariat to ANSI X3 indicating that

the ISO committee will meet on 16 July 1976 to discuss whether or not to

revise the ISO/R1539 Fortran Standard.

4.4.2 European Computer Manufacturers Association (ECMA)

ECMA had planned to meet on 4th June 1976 to make final comments on the

draft ANS Fortran Standard. No details have been received yet.

4.4.3  Purdue Workshop

Corresponence has been received from Mrs. Maxine Hands and Matthew R. Gordon-

C1ark together with copy of the latest revisions of Instrument Society of

America (ISA) standards: ISA S61.1-1975, procedures for executive

functions, process input/output and bit manipulation. Draft ISA S61.2 -

(2-11-76), procedures for file access and the control of file contention.

The work on ISA S61.3 on tasking is reported to be proceeding slowly.

4.4.4 Dutch Fortran Study Group

Correspondence has been received from Mr. Heyns. The Dutch group have

sent proposals to ANSI to delete the assigned GO TO statement and the

ASSIGN statement. They present some convincing arguments - such as the

irregular use of INTEGER variables to hold 'LABEL' information.  Those

present thought that the proposal was unlikely to succeed on the grounds

that the feature was fairly heavily used.

4.4.5 British Standards Institute (BSI DPS/13)

Dr. I. D. Hill reported that a meeting of BSI DPS/13 had been held on 27

May to consider comments on the draft proposed ANS Fortran revision. Those

present consisted of members of BSI and some representatives from other

groups namely Mr. A. M. Conybeare (BSI Technical Officer and Secretary

to DPS/13), Dr. I. D. Hill (acting Chairman, representing BCS)

Mr. A. Addyman (ALLC), Mr. H. Pitcher (NCC), M. D. Maisey (ECMA), Dr. A. C. Day

(UCL), Mr. B. Chapman (BETA) and Mr. P. A. Clarke (BCS Fortran S. G.).

About 90 comments were considered and after discussion, 102 comments

were forwarded to the BSI executive to forward co X3J3. The comments were

categorised as general comments, matters of substance and editorial detail.

BSI has a vote on the ISO committee and may insist on a few changes to

the standard.

4.4.6 BCS Standards Committee

Dr. Day said that Mr. F. Taylor had invited him to the next BCS Standards

Committee meeting on 17 June to discuss the draft proposed FORTRAN standard

and the working party review of it. This he had accepted because it would give

him the chance to complain about the content of the review paper being leaked

to the press prior to publication.

Mr. Muxworthy had received a note saying that the BCS Standards committee

were considering becoming the official distributors of the revised

U.S. Navy Fortran test suite. He had replied saying that on the basis

of our experiences with the early version, the suite itself should be

well tested before release. He also hoped that intervention by the BCS

would not increase the price to end users.

5.   Progress Reports

5.1  Promotion Working Party

Mr. Clarke said that the draft Fortran contact list had been revised and

would be distributed with the minutes of the meeting.

P. S. Please note that Dr. A. C. Days address should read '19 Gordon St,'

not 'Gower St.'

The Introductory document for new and prospective members of the group also

needed revising.

5.2   Other Reports

5.2.1 Preprocessor Working Party

Dr. Murchland said that the objectives of the working party were published

as an appendix to the minutes. So far, the main activities had been in

circulating relevant articles and comments. Suggestions had been circulated

on Fortran record convention, redefinition of comments, conventions for use

of comments to contain commands for code processors, standard numbering of

Fortran Statements, alternative continuation convention for Fortran and

mixed upper/lower case listings of Fortran.

It had been decided to investigate two structured preprocessors, MORTRAN

which employs macro principles and FLECS which is conventional. It was

hoped that these would be easy to obtain at little expense.

5.2.2  U.S. Navy Tests

The U.S. Navy tests that had been sent to Mr. Martyn Thomas at SWURCC

(Bath University) had been run on ICL 2900 Fortran F0 and Fl compilers by

Ann Rogers. A written report is expected shortly. The Fl compiler

is for use with the system K operating system. The F0 compiler is for

use with the system B operating system. Fl and F0 are supposed to be

compatible - currently there are some minor differences. With the tests,

the main cause of problems was the sequence ,/, in Formats (an extension

over the l966 ANSI standard) used in some of the tests (and permitted on

many or most compilers). Some 6H Hollerith constants caused problems too.

Some comment cards starting C*** were taken as end-of-file by the operating

system.

6.   Any Other Business

Dr. Clifford (the speaker at the last meeting) has sent a photocopy of the

slides describing DEC Fortran (for PDP8 and 11) and compiling techniques. This

is available on request from the secretary.

Dr. Murchland commented that the Fortran books in the BCS Library were mostly

elementary. None had the word 'advanced' in the title. The group decided to

inform the librarian that we recommend 'Fortran Techniques' by Dr. A. C. Day,

published by Cambridge University Press, 1972 (reprinted 1974) £1-40.

Mr. Clarke had noted some interesting articles on Fortran in the CERN

newsletter by Mr. M. Metcalf. These would be particularly useful to users of CDC

Fortran.

P.S. Concerning the efficiency aspects of (overlayed) programs on IBM 360 and

370 (see December 1975 Minutes), a modification to the IBM system software

has been received from the Stanford Linear Accelerator Laboratory.

This has the effect_of making a significant (20-30 percent) increase in

throughput. Copy available from the secretary.

The Annual report of the group is attached (Appendix B).

7.   Date of Next Meeting: Monday 4th October 1976

8.   Dr. I. D. Hill (MRC) gave a talk on "Language Standards and Algorithm

Editing" (see Appendix C for details).

  Appendix A

    M E M O R A N D U M

From: B.C.S. Fortran Specialist Group            To: Lloyd Campbell, Secretary,

                                                     ANSI X3J3

Date: l4th June, l976

Subject: Public Comments on the draft proposed Fortran Standard

The following straw ballots were taken at our meeting of 7th June 1976 and

relate to some British public comments on the draft proposed Fortran Standard.

Appendix B

BCS FORTRAN SPECIALIST GROUP ANNUAL REPORT l975/76

The original reason for the formation of the Group in January 1970

was to attempt to influence the revision of the American Standard for Fortran,

work on which was expected to take a couple of years. The publication of

the draft standard has been expected annually since 1972 but during this

year it finally appeared (on March 1 1976). The Group may reasonably claim

to have had some influence on the draft and it has well-established links

with the ANSI Fortran committee. The draft is not yet a standard: various

rounds of voting have to take place and it is not impossible that major

revisions may yet be made.  The group bought 50 copies of the draft standard

and distributed them at cost to members in the U.K.

Throughout the year the Group has closely followed developments

with the draft standard and it has made its opinions known to ANSI. Contacts

have been maintained and papers exchanged with the CODASYL Fortran DBLM

Committee, the Fortran Development Committee, the Fortran Committee of ECMA,

and the Dutch Fortran Study Group.

A working party, chaired by Dr A.C. Day, has compiled a report on

the draft American Standard. The report is to be sent to ANSI and is to be

published in the August 1976 Computer Journal. Another working party, under

the chairmanship of Dr J.D. Murchland, has been formed to investigate Fortran

preprocessors.

Following a member survey, the pattern of meetings has been changed

this year; they now occupy a full day with discussion sessions in the

morning and a more formal talk in the afternoon. The topics covered have

included: Compiling on mini-computers, Optimizing Compilers, Structured

Fortran and other preprocessors and there were two talks on particular

manufacturers' Fortran, viz. DEC and Univac.

There have been 6 meetings during the year with an average

attendance of 18; the mailing list has grown to over 150 and efforts are

being made to prune it and to introduce a subscription for non-BCS members.

The present chairman and secretary, D.T. Muxworthy and P.A. Clarke, were

re-elected for 1976-77; the vice-chairman, E.O. Bodger, has been relocated

by his employers and is now reporting on Fortran events in California for

the group. This would be an appropriate place to thank him, other past and

present officers and members of the steering committee and working parties,

staff of the Technical/Branches division at BCS HQ and our speakers during

the year for all their work for the Group.

D.T. Muxworthy

25 May 1976

Appendix C

  Summary of a talk by Dr. I. D. Hill on Language Standards and

   Algorithm Editing

Dr. Hill related some of his experiences that had derived from being

the Algorithms Section Editor for the Journal of Applied Statistics during the

last four years.

The initial problem in setting up an algorithms section was in the choice

of languages to be allowed. Earlier the CACM had chosen Algol 60, followed

later by Fortran. Appl. Stat. decided upon Algol 60 as defined in the

revised report, Fortran, to ASA 1966 standard and PL/1 as defined in the IBM

Manual.

The next problem was whether to permit any extensions and minor

differences to the standard. As there was no agreed set of extensions, it was

decided to follow the standard to the word. Despite this, some of the first

algorithms to be published did contain a few extensions, e.g. AS2 contains a DATA
statement specifying an array name.

This decision also means that some petty restrictions in the standard are

enforced - for example, a Hollerith string containing colon might be permitted

but a comment line containing colon would not.

Some common errors that have been found are: assuming a variable local

to a subroutine remains defined upon return, using assignment and IF statements

to move and compare Hollerith strings and referencing functions with a

Hollerith constant as an argument (this is permitted for subroutines).

In addition to enforcing programs to be standard conforming, the

editors expect a reasonable standard of programming, and also recommend some

conventions in situations where the language permits alternative methods of

expression (Ref: Journal of Royal Statistical Society, Series C (Applied

Statistics) Vol 24 No. 3, 1975 pp 366-373).

For example code such as that found in the article by D. J. Evans and

M. Hatzopoulos in the current issue (Vol 19, No. 2, May 76) of the Computer

Journal contains many examples of subscript expressions such as: A(J,N+P+1-

I-J-K)=A(J,N+P+1-I- J-K)+S*A(I,P+K) which as well as being non-standard, should

also have been evaluated in advance. The Algorithms section Editor would have

returned this as something to be 'reconsidered'.

Other examples of this type, that have been submitted include things like

D=-l.0*(B-C) instead of D=C-B and even W=B**C followed by W=ALOG(W)/ALOG(B)

instead of W=C. The code 1.0/SQRT(44.0/7.0) for 1/√(2π) should be reduced to

the constant 0.3989422804.

The conventions recommended by the algorithms editor include: using

labels in consecutive ascending order; for fault handling, the last argument of

the subprogram should be called IFAULT and should be set to zero if no faults

occur, or some other value if faults do occur; logical IF is preferred to

arithmetic IF where appropriate; comment lines should not duplicate the

introductory text; function names should follow the implicit typing rules so

that the user does not have to include type statements in the referencing

program; the constants 0.5 and l.0 are preferable to .5 and 1. ; All DO loops

should end with CONTINUE and Input/output should not appear in the algorithms

(except when they are specifically designed for input/output).

It is interesting to note that very few algorithms are accepted without some

change. To enhance the appearance of the published algorithms, a Fortran layout

program (FLOP) has been used.