FSG minutes for 25 January 1990

Minutes of the BCS Fortran Specialist Group Meeting

held at BCS HQ, 13 Mansfield Street, London

on 25 January 1990

Present: Robert Allan Daresbury Laboratory (SERC)

Martin Counihan Southampton University

E Golton RAL

Dave Griffiths SSF

Carol Hewlett LSE

Peter Hollan SSL

David Holmes Rolls Royce plc

Chris Lazou ULCC

Mike Nunn CCTA

Les Russell AWE

N R Saville Private Consultant

John Taylor Meiko Scientific Ltd

John Wilson Leicester University

John Young PE-MOD

1. APOLOGIES FOR ABSENCE

Apologies for absence were received from Miles Ellis

Valerie Harmer, David Muxworthy and Lawrie Schonfelder.

The Secretary, John Young, apologised for the short notice

of the meeting and thanked the BCS HQ staff for the prompt

despatch of the minutes. He pointed out that meetings are now

planned at least six months ahead and that it was hoped that any

changes would be notified to members well in advance.

[Secretary's Note: Brave words! Please note that the AGM has

been postponed from 5 April 1990 to 10 May 1990].

2. MINUTES OF THE LAST MEETING [26 October 1989]

The Secretary apologised for not contacting the speakers of

the talks on Fortran on PC's for their summaries. He had not yet

contacted the Rutherford and Appleton Laboratory concerning a

visit to the Atlas Building. It was hoped to arrange this visit

in October and the Secretary would liaise with E Golton.

On Page 3, 2nd paragraph, after some discussion the sentence

starting "Time was spent on discussing character sets and it was

decided that ..." should read "Time was spent on discussing

character sets and it was agreed to press X3J3 to remove user-

defined character sets from identifiers".

3. MATTERS ARISING

Fortran Forum

The Chairman, John Wilson, reported to the meeting on the

final outcome of Fortran Forum 89. The main item of expenditure

had been the printing of the Draft Fortran Standard Manuals which

had cost considerably more than the original quotation. So far

only the money for 33 manuals had been collected but it was

thought that a similar number had been sold by the Scottish Sub-

Group. All the speakers' expenses had been paid. With the Forum

costing over £1,000 the Group's reserves were now depleted.

It was agreed that more manuals had to be sold in order to

re-coup some of the costs and that they would be advertised in

the next newsletter.

It was then pointed out that the manuals would very soon be

out-of-date. If it was possible to produce a list of the changes

made to the Draft Fortran Standard then the Manuals would remain

a useful reference document.

Second Public Review

E Golton commented that he had received no acknowledgement

for his comments on the Draft Fortran Standard sent to X3J3

during the second public review period. Indeed, he had only just

received a reply to his comments sent during the first public

review period.

In the general discussion that followed there were a number

of complaints about the letters sent out mainly that the replies

were "off the point". Apparently, there was no procedure within

X3J3 to comment on the comments.

In defence of X3J3 it was difficult to reply totally

individually to 400+ comments some of which were long and

detailed.

X3J3 had received about 30 letters complaining about the

lateness in replying to comments from the first public review

period.

4. REPORTS FROM X3J3 AND BSI REPRESENTATIVES

Report from X3J3

The Chairman, John Wilson, read John Reid's report of the

recent X3J3 meeting which was available at the meeting and is

included as Appendix A. Only the following technical changes are

reproduced here for completeness:-

(i) Delete user-specified characters in identifiers

(ii) Adopt the ES edit descriptor

(iii) Remove the requirement that precision be monotonic with KIND.

The meeting noted that X3J3 had adopted the informal name

"Fortran 90" for the new standard.

The meeting also noted that this was John Reid's last

report. The meeting agreed that these reports had been the

"backbone" of the Group's information on Fortran 8x and the

Secretary was asked to write to John thanking him for his

contribution. The flow of information had been so regular and

concise that nobody in the meeting could remember when they had

started. The Chairman announced that Miles Ellis would produce

reports for the Group in future.

The Draft Fortran Standard was now complete with only

editorial work to do.

Report on the BSI and ISO Vote

John Wilson reported that the BSI Fortran Panel had met on

3rd January 1990 to formulate the comments to accompany the UK

vote on the new Fortran Standard. There was no discussion on the

adoption of Fortran 77 as a separate standard by ANSI as the

matter is now irrevocable.

The official UK vote is to be simple and short as possible;

"YES" with the following comments:-

Comment one: We appreciate that in almost all cases our

reasons for voting "NO" in the ballot on the first DP 1539

have resulted in changes to the proposals in accordance with

our wishes.

Comment two: We consider early adoption of the revised

Fortran standard to be of overriding importance and

therefore urge that the draft be processed with no further

changes other than correction of errors, clarification of

ambiguities, improvement of specification and rectification

of editorial and typographical mistakes.

John Wilson then read the report from the IST/5/5 convenor,

David Muxworthy, concerning the rationale behind the UK vote on

the second DP 1539. The panel received nine written comments,

eight of which were generally in favour. After much discussion

the panel decided that early publication of the new standard was

of paramount importance and would pursue the request for

technical changes through WG5 and X3J3 meetings if possible.

It was reported that the US vote was similar to the UK vote.

It was announced that the next WG5 would be in London from

26 February - 2 March 1990. There followed some discussion on

who was organising the meeting and its venue.

5. AGM AND 20TH BIRTHDAY CELEBRATIONS

Annual General Meeting

The Chairman, John Wilson, announced to the meeting that

Miles Ellis would not be standing again as Treasurer of the Group

due to other commitments. The other three members of the

Committee were prepared to stand again to maintain the continuity

of the Group. John Wilson then appealed for names of people

prepared to stand for the next Committee. With none forthcoming

the Secretary, John Young, was asked to prepare Nomination Papers

for the four posts of Chairman, Vice-Chairman, Secretary and

Treasurer. The date of the AGM was announced to the meeting as

5 April 1990 [but this has been changed to 10 May 1990].

20th Birthday Celebrations

As the Treasurer, Miles Ellis had been unable to arrange

this event and as the Group's finances were at a rather low level

it was decided to postpone the Group's celebrations for a year

and make them 21st Birthday Celebrations. John Wilson would

invite a well-known speaker who would be asked to talk on some

aspect of the history of Fortran. Mike Nunn offered to form a

Sub-Group for the 21st Birthday of the Group so that arrangements

could be made well in advance. It was agreed that comments and

suggestions for the event could be invited from members of the

Group possibly on this year's AGM nomination form.

[Secretary's Note: Since the cancellation of the 20th Birthday

Celebrations and the postponement of the AGM to 10 May 1990 David

Muxworthy has agreed to talk on the "Fortran Specialist Group:

The first two decades" after the AGM].

6. BCS AND OTHER BUSINESS

Specialist Groups Management Committee

The Chairman, John Wilson, did not have much to report on

the recent meeting. The following items may be of interest to

the Group.

(a) Specialist Groups Reference Manual

This manual contains information about the BCS and

Specialist Groups, suggestions how to run a Group, any legal

requirements and a current list of all the Groups and their

officers. The Handbook is distributed to Chairmen, Secretaries,

and Treasurers; Chairmen were asked to report any errors.

(b) BCS pages in Computing and Computer Weekly

There had been many complaints about the entries in the BCS

pages in Computing and Computer Weekly. Various suggestions for

improvement were made, including a pro-forma to be used by Group

Secretaries. However, it seems the BCS has little control over

what does or does not get published.

A standing invitation has been made to any BCS member

passing through Hong Kong to speak to the Branch.

Scottish Sub-Group

The Scottish Sub-Group convenor, John Bruce, had organised

a meeting in Glasgow for 5 February 1990 on Expert Systems. The

Sub-Group intends to "mirror" the Group's topics but will only

hold afternoon meetings as they have little business to discuss.

BSI IST/5 Meeting

Following the IST/5 meeting in November the Chairman, John

Wilson, was asked for the Group's views. The resolution and

endorsement taken at the meeting of 26 October 1989 were passed

on via the CCTA representative.

Letter from the Technical Board

John Wilson had received a letter from Roger Johnson

requesting members to identify missing activities within BCS with

a view to forming new specialist groups.

Survey on Fortran Programmers

John Wilson had received a request for information on how

many Fortran Programmers there were in the UK. The meeting was

unable to answer this question but it was pointed out that a

survey on use of Fortran had been undertaken recently. The

question appeared to have been some sort of market research for

the launch of a new software product.

DISC - Delivering Information Standards to Customers

John Wilson had received a questionnaire concerning this

government sponsored body to take over BSI's Information

Technology standards including Fortran. He also had received a

comment on DISC from Lawrie Schonfelder.

Parallel Processing Specialist Group

The Vice-Chairman, Chris Lazou, sought collaboration of the

above group with the Fortran Group. The Parallel Processing

Group's evening meetings are free. There is to be an important

meeting on 14 March 1990 starting at 6.30 pm at the Fleming

Lecture Theatre, University College, Gower Street, in London.

A visiting Russian Professor Alexander Shaferenko will be giving

a talk on "EVAL - The Language". Approval for a flier for this

meeting to be included in the next mailing was given.

7. DATE OF NEXT MEETING

The next meeting of the Group will be the Annual General

Meeting to be held on 10 May 1990 at BCS HQ, 13 Mansfield Street,

London from 10.30 am. The afternoon speaker will be David

Muxworthy who will talk on the "Fortran Specialist Group: The

first two decades". Please note that this is a change of date

from the previously announced date.

John Young, Secretary

28 February 1990

Appendix A

To: Fortran Forum, BCS, NAG, etc.

From: John Reid

Date: 19th January 1990

Subject: X3J3 meeting in Richardson, Texas

Note: This is a personal note on the meeting and in no sense does it

constitute an official record of it.

1. Summary

Jeanne Adams was unwell, so the meeting was chaired by Jerry

Wagener. The second set of public comments letters was to hand and the

Committee considered how to respond technically, what to recommend for

the US ISO ballot in the light of this commentary, and what

administrative machinery to use to reply to the individuals.

A significant majority of those who commented on the language as a

whole were supportive and many requested that few if any further

changes be made in order to expedite processing. There was a broad

consensus in the Committee that few technical changes be made and

there were no procedural arguments. Everyone worked very hard

correcting errors, making editorial improvements, and considering the

comment letters. Only the following technical changes (other than bug

fixes) were adopted:-

(i) Delete user-specified characters in identifiers (Section 3)

(ii) Adopt the ES edit descriptor (like lPE, but with no

effect on subsequent edit descriptors) (Section 4).

(iii) Remove the requirement that precision be monotonic with

KIND (Section 5).

The US members recommended that the US ISO vote be NO, changing to YES

with the adoption of these three changes and the corrections

accumulated in the document S18. This is nearly a YES vote since

other countries will find it easy to accept these requirements.

Furthermore, the informal name 'Fortran 90' was unanimously adopted,

an earnest of the intention to finish this year.

There was no further discussion of the retention of Fortran 77 as an

ANSI standard because the decision has been made by X3 and is final.

The ISO Fortran Working Group (WG5) meets in London, 26 February to 2

March, to consider the ISO country votes and will have in front of it

a revised version of the draft standard including all the S18

changes. Certainly, it will need more editorial polishing, but it

looks as if the next stage is near, a Draft International Standard

(DIS), perhaps in August when WG5 meets in the Netherlands and X3J3

meets in Oxford. Once it is a DIS, no technical changes are permitted.

2. Second set of public comments

The second set of public comments is not as voluminous as the

first. It consists of 153 letters that arrived in time and 10 letters

that arrived a few days late, a pile of paper about two inches

thick. In addition, about 30 letters have been received objecting to

the X3J3 replies to letters received in the first comment period. This

time, the overall tone is supportive. By my count, 74 letters are

supportive of the language and 38 are opposed. Including the protest

letters (but not counting anyone twice) increases my counts to 75-43.

Many edits have been made in response to the comments (though not as

many as last time). In addition, the subgroups reviewed the recurring

themes of requests for technical changes and made recommendations to

the Committee. To have acted on many of them would have gone against

the many requests to stop making changes and get the Standard out

soon. Straw votes identified the following items as having significant

support in the Committee:-

(i) Change '<>' to '/='.

(ii) Delete user-specified characters in identifiers.

(iii) Adopt the ES edit descriptor (like lPE, but with no

effect on subsequent edit descriptors).

(iv) Remove the requirement that precision be monotonic with KIND.

The chairman ruled that the first was an error because a comment letter

pointed out a conflict with an IEEE standard. This was therefore

adopted as part of S18. The second and third were adopted with text

and will be mentioned in the US vote. The fourth does not have agreed

text, but will also be mentioned in the US vote.

3. User-specfied characters in identifiers

The changes made last May in respect of user-specified characters were

done hurriedly in response to a perceived international

requirement. They have not been welcomed by WG5 and now appear to be

premature in the light of ongoing standards work in connection with

character sets. It was at first proposed that all the May changes be

undone. However, in a near unanimous vote (30-2), the proposal was

amended to retain the exclusion of control characters from the set of

'representable characters' that can appear in the free source

form. The amended motion was adopted (27-5).

4. Input-output

The ES (scientific) edit descriptor is like 1PE, but has no effect on

subsequent edit descriptors. It was adopted by 27-6 and will obviate

the danger of fixed-point output being unexpectedly scaled by the

factor 10.

In addition, the following minor changes were adopted (as part of Sl8):-

(i) Permit preconnected files to have the BLANK='ZERO' property,

for conformance with Fortran 77 (intended originally for

card readers).

(ii) Disallow undelimited character NAMELIST input.

(iii) Require all the objects in control information lists to

have default KIND (I think that this is an unnecessary

restriction that would be a nuisance to the programmer,

but I was in a small minority).

(iv) Delete the NULLS= specifier (it does not return anything

useful).

(v) Require that an expression in an input list must not be

affected by the input of any array element by that item

(for instance, the section I( :I(1)) is not permitted,

where I is a rank-one integer array).

A proposal to allow n to be optional in an nX edit descriptor, with

the default 1, failed (8-17).

5. Data entities

It was decided by the US members (23-1) that there should be no

requirement that the precision of real and complex entities be

monotonic with KIND. The reason is that an implementation may be

unable to offer a further KIND because of lack of room between

existing KIND values. I spoke against this because it makes the rules

for the interpretation of mixed KIND expressions more complicated,

which runs counter to the simplifications made in this area following

the first public comment period.

6. Generic calls

I worked quite hard in collaboration with the data subgroup to tighten

up the wording on generic calls and how ambiguities are resolved. The

revised wording was adopted unanimously. It means the retention of the

requirements of Section 14 of S8 on the differences between two

procedures if they are to have the same generic name (wording

elsewhere suggested that it was only necessary for no ambiguous call

to be possible, which may be hard for the processor to verify). Also

the rules for extending an intrinsic and extending a generic accessed

from a module are now set out (it is important that both these

facilities be allowed).

7. Future trip reports

My job is changing and I am not expecting to attend any more X3J3

meetings in the US, though I hope to attend the Oxford meeting in

August. Therefore, this will be the last time that I write a

trip-report for AEA Technology and adapt it for you.

Enjoy using Fortran 90!