Support for Fortran Standards Development

Report September 2010

Introduction
Since January 2003 the BCS Membership Board (and its predecessors for specialist groups) have generously provided means to help members of the Fortran Specialist Group support the UK contribution to the development of international Fortran standards. Initially this was undertaken as a project under the Specialist Groups Development Fund; from 2006-2007 funding has been provided as part of the normal Fortran Specialist Group annual allocation.

The initial case from 2002 is at www.fortran.bcs.org/2003/devrep03.htm and subsequent annual reports are linked from www.fortran.bcs.org/standards/stanhome.php.

Activity 2009-2010
With the change in the BCS financial year, this report covers the period June 2009 to September 2010. There has been one ISO Fortran (WG5) meeting during this time, in Las Vegas in February 2010. There were in all four participants from the UK, one of whom was supported by the BCS, while the total attendance was 15 from six different countries. The convenor of WG5 is from the UK, as is the editor for the revision project; the secretary of each meeting and the editor of the Technical Corrigenda are generally appointed from amongst the UK members. Most detailed technical development work continues to be undertaken by the US INCITS committee PL22.3 (still generally known by its former name, J3), with worldwide contributions by email between WG5 meetings. One member of the British group is an active member of J3.

The Final Committee Draft of Fortran 2008 was issued for ballot in September 2009. The UK again voted 'no' (v. the report for 2008-2009). The overall vote was 9 approvals, 3 approvals with comments, 1 disapproval and 6 abstentions.

A prime concern for the UK was the confused state of vendors' implementations of the current standard which are largely incomplete but in some cases already contain selected features from the new draft. Thus each supplier is essentially defining its own subset of Fortran 2008. To ameliorate the situation we suggested a new project to define a portable subset. However at the February 2010 meeting it quickly became clear, apart from the fact that a true common subset would be a moving target, that there were multiple views on the content and purpose of a subset and that it would take an impracticably long time to define. The proposal was thus not accepted. Fortran is not alone in this situation. To varying extents, standards for other languages, including C, C++ and Cobol, are also being developed faster than vendors' ability to implement them, thus devaluing the prime raison d'être of a standard.

The other main concern, of the majority but not all of the panel, was that the new parallel features are deemed to be mandatory rather than optional. This point was also made in the UK vote in the previous international ballot and was rejected by WG5. This is a strongly held view so it was repeated but again rejected. Japan, also recognizing this to be a lost cause, took an alternative approach: it ignored the subject in its ballot but stated the same opinion as the UK in its national activity report.

We recommended deletion of the arithmetic IF statement since with IEEE arithmetic it is possible that none of the three branches is taken; moreover the statement has been deemed 'obsolescent' for nearly twenty years. This was resisted on the grounds that it would make some existing standard conforming programs non-standard conforming. This runs counter to the policy for language evolution established in 1990, which is probably now dead, that redundant features should be marked obsolescent as a warning that they may be deleted from a succeeding standard. WG5 decided to correct the definition of the statement rather than to delete it.

A number of clarifications and editorial corrections were also proposed by the UK; these were all accepted.

After incorporation of the edits agreed at the WG5 meeting, which essentially set the document in stone according to ISO rules, the Final Draft International Standard (document N1830 on the WG5 website) appeared for ballot in July 2010. The opinions of BSI panel members had not shifted significantly and there was a proposal that the UK now abstain in order to show continuing lack of approval. However at this stage of development ISO does not allow actively participating countries to abstain, leaving a straight choice of approval or disapproval. The former was chosen with some reluctance.

In September 2010, some five and a half years after member bodies started formulating their requirements, the result of the FDIS ballot was announced: the standard was approved by 18 votes to none, with 15 abstentions. Formal publication of the standard is expected in November 2010.

WG5 has various ancillary projects. Technical Corrigendum 4 for Fortran 2003 was issued during the period covered by this report. A Technical Corrigendum 5 was prepared but not published due to the imminent completion of Fortran 2008. A composite of all five corrigenda is in document N1823 on the WG5 website. Work is already starting on processing interpretation requests on the draft of Fortran 2008.

UK members have also contributed significantly to the Technical Report on Further Interoperability of Fortran with C. This project had been a cause for concern both with regard to the technical content and to the operation of the development group. New objectives and a new modus operandi were agreed at the February 2010 WG5 meeting, largely as a result of UK input. The report is scheduled for publication in November 2011.

In response to concerns expressed by the UK and others, in 2008 WG5 removed some coarray features from the draft standard with the intention that they be incorporated in a Technical Report on Enhanced Parallel Computing Facilities. A specification for the content of the report was produced at that time but development was not pursued. Work has now restarted with target dates for a first draft of the report in February 2011 and for final publication in November 2012.

UK members have also contributed to the cross-language Technical Report on Guidance to Avoiding Vulnerabilities in Programming Languages through Language Selection and Use. Language-specific annexes, including one for Fortran, are being developed for the second edition. A draft of the technical report, but not the final published version, is available as document N0238 on the SC22/WG23 website: http://www.aitcnet.org/isai/ and a working draft of the Fortran annex is in N0220.

All Fortran development papers are open to all interested worldwide; minutes of the most recent WG5 meeting are in N1810. WG5 documents are available via its website www.nag.co.uk/sc22wg5/. J3 papers are obtainable from www.j3-fortran.org/.

The Group is extremely grateful for the continuing support of the BCS.

David Muxworthy
BSI Fortran Convenor
13th September 2010


Valid XHTML 1.0! Comments on this or any other of the Group's pages should be sent by email to the FSG Web Editor.

Last modified: Wed 2 Nov 2011 09:30:35