From joseogl@microsoft.com Thu Feb 24 03:10:16 1994 Received: from netmail2.microsoft.com by dkuug.dk with SMTP id AA10422 (5.65c8/IDA-1.4.4j for ); Thu, 24 Feb 1994 03:10:16 +0100 Received: by netmail2.microsoft.com (5.65/25-eef) id AA07539; Wed, 23 Feb 94 18:11:19 -0800 Message-Id: <9402240211.AA07539@netmail2.microsoft.com> Received: by netmail2 using fxenixd 1.0 Wed, 23 Feb 94 18:11:19 PST X-Msmail-Message-Id: BEED6A3B X-Msmail-Conversation-Id: BEED6A3B From: Jose Oglesby To: daemon@newton.ncsa.uiuc.edu, SC22WG5@dkuug.dk Date: Wed, 23 Feb 94 18:04:49 TZ Subject: RE: (SC22WG5.506) Defect Item 125 X-Charset: ASCII X-Char-Esc: 29 Considering existing implementations is a legitimate concern of standard bodies. There are existing Fortran 90 implementations with new ones coming along. What do these implementations do? If most of them do not implement the TARGET attribute by copy in/copy out then making this change may be harmless. On the other hand, if most of them use copy in/copy out there is a problem. If the standard was underspecified and internally inconsistent, it seems wrong to now penalize the people who implemented early. If you really want to refine the specification please consider adding a new attribute in the next version of the Standard. In time the old TARGET attribute can be made obsolescent. -- jose ---------- | From: Jeanne T Martin | To: | Subject: (SC22WG5.506) Defect Item 125 | Date: Wednesday, February 23, 1994 10:27AM | | Received: from newton.ncsa.uiuc.edu by netmail.microsoft.com with SMTP (5.65/25-eef) | id AA06181; Wed, 23 Feb 94 12:04:52 -0800 | Received: by newton.ncsa.uiuc.edu id AA20759 | (5.65a/IDA-1.4.2 for joseogl@microsoft.com); Wed, 23 Feb 94 13:57:39 -0600 | Message-Id: <199402231831.AA24759@dkuug.dk> | Errors-To: netmail!x3j3-request@ncsa.uiuc.edu | X-Sequence: SC22WG5@dkuug.dk 506 | X-Charset: ASCII | X-Char-Esc: 29 | | | I agree with Lawrie that forcing copy in/copy out semantics for an argument | with the TARGET attribute is wrong. This is not what users expect. It | provides no utility, in fact the reverse. If the semantics are going to | be pinned down, they should be pinned down in such a way as to enhance the | utility of the language - not detract from it. | | I realize the copy in/copy out semantics may be easier to implement, but | that should not be the overriding concern of a standards body. | | Jeanne Martin | email: jtm@llnl.gov postal code: L-300 | phone: (510) 422-3753 fax: (510) 423-8704 | Lawrence Livermore National Laboratory | Livermore, CA 94550 | |