From jls@uxb.liv.ac.uk Mon Jun 14 12:47:44 1993 Received: from mail.liv.ac.uk by dkuug.dk with SMTP id AA01853 (5.65c8/IDA-1.4.4j for ); Mon, 14 Jun 1993 12:47:44 +0200 Received: from uxb.liverpool.ac.uk by mailhub.liverpool.ac.uk with SMTP (PP) id <25575-0@mailhub.liverpool.ac.uk>; Mon, 14 Jun 1993 11:47:50 +0100 From: jls Message-Id: <8018.9306141047@uxb.liv.ac.uk> Subject: Observations on item 13 of X3J3/93-006r (N894) To: SC22WG5@dkuug.dk (SC22/WG5 members) Date: Mon, 14 Jun 93 11:47:25 BST X-Mailer: ELM [version 2.3 PL11] X-Charset: ASCII X-Char-Esc: 29 I feel I must comment on this document. Firstly, item 4 of this document is factually incorrect. The inheritance of implicit mappings into interface blocks was very definitely not an oversite. It was done deliberately. The US may believe this to have been a mistake but it was not an accident. It can not therefore under ISO rules be changed by a corrigendum. These are by definition allowed to correct defects inadvertently introduced during processing into a published standard. To make would recommended change to the standard therefore requires a revision or an ammendment not simply a corrigendum. Secondly, I for one remain singularly unconvinced by the technical arguments. In cases where the user has access to the source of the external procedure, and he/she wishes to produce the interface by simple copying of the specifications these will include the IMPLICIT statements that apply. If as is common when using libraries of external procedures the user does not have access to the source, the copy operation is not possible. In this case the user does not know the implicit mapping used in the external, all he/she knows is the characteristics of the procedure interface as given in the documentation. He/she also knows the implicit mapping he/she has set up in the program being written. It is therefore entirely logical for this to be the default mapping used in the interface block. The case that requires the user to be careful is the one where the external uses the I-N default mapping but the user code does not and the copy specification model is used. It is a moot point and very much a matter of opinion as to whether it is more surprising to have the host mapping apply by default or to have the default mapping reinstate itself automatically in each interface body declaration. At the time the standard was written those of us who looked at this issue carefully considered that the former was the route for least surprise and would produce least user error. Thirdly, there are many more irregularities in the treatment of interface blocks in F90 as defined. Most of these disappear, at least for the purposes of teaching the concept, if it an interface body is considered to be an explicit declaration of a procedure. Viewed in this light and with all characteristics explicitly declared most of the problems disappear. Finally, not only am I opposed to the proposed correction on technical grounds, but also to make such a change as a corrigendum is simply out of order under ISO rules. -- Dr.J.L.Schonfelder Director, Computing Services Dept. University of Liverpool, UK Phone: +44(51)794 3716 FAX : +44(51)794 3759 email: jls@liv.ac.uk