From owner-sc22wg5@dkuug.dk Mon Apr 12 05:16:46 2004 Received: (from majordom@localhost) by dkuug.dk (8.12.10/8.9.2) id i3C3Gk1j078658 for sc22wg5-domo; Mon, 12 Apr 2004 05:16:46 +0200 (CEST) (envelope-from owner-sc22wg5@dkuug.dk) X-Authentication-Warning: ptah.dkuug.dk: majordom set sender to owner-sc22wg5@dkuug.dk using -f Received: from imo-m22.mx.aol.com (imo-m22.mx.aol.com [64.12.137.3]) by dkuug.dk (8.12.10/8.9.2) with ESMTP id i3C3GcE7078645 for ; Mon, 12 Apr 2004 05:16:42 +0200 (CEST) (envelope-from Wclodius@aol.com) Received: from Wclodius@aol.com by imo-m22.mx.aol.com (mail_out_v37_r1.2.) id 9.f7.390aceb3 (4214) for ; Sun, 11 Apr 2004 23:16:47 -0400 (EDT) From: Wclodius@aol.com Message-ID: Date: Sun, 11 Apr 2004 23:16:45 EDT Subject: Re: (j3.2004) (SC22WG5.3099) New entries in repository To: sc22wg5@dkuug.dk MIME-Version: 1.0 Content-Type: text/plain; charset="US-ASCII" Content-Transfer-Encoding: 7bit X-Mailer: AOL 5.0 for Mac sub 39 X-Spam-Score: 0.339 () NO_REAL_NAME Sender: owner-sc22wg5@dkuug.dk Precedence: bulk In a message dated 4/8/04 6:20:55 AM, d.muxworthy@tiscali.co.uk writes: >The BSI Fortran Panel has submitted four new entries for the WG5 >repository of requirements. The titles are: > >Co-array Fortran for parallel programming >Decimal floating point arithmetic >Conformance to IEEE 754R >KIND environment specification > FWIW my opinion as to what I would prefer to see as J3/SC22WG5's priorities: First consider whether it is feasible to have more than one more revision in the Fortran standard. With the gradual falloff in participation only one more revision may be possible. If only one more revision is possible consider making it a moderate revision that takes six or seven years to publication, otherwise aim for a minor revision in five years time. Second, the number of requests for clarification of the F2003 standard will increase first from vendors as their F2003 compilers near completion, then from users as they become familiar with the language. The next couple of meetings will allow a fair amount of time for addressing enhancements. Use the time to greatly trim the number enhancements on the table, so that you are well focussed when interpretations occupy a substantial amount of time. Third, the number of major enhancements that are proposed is overwhelming. My view of them at this time Generic programming - This is the most useful generalization of the language, but probably also has the most widespread impact on the language text. Make this the first priority, and realize this may be the only major enhancement in the next standard if you decide on a minor revision. Its inclusion may delay the standard past the five year goal of a minor revision. Co-arrays- As an existing implementation exists, this should be the most easily incorporated major enhancement. However after the relative failure of HDF I am reluctant to put a strong emphasis on parallelization features. If a minor revision is planned, only include this if it is not expected to complicate the inclusion of generics. IEEE 754R - Useful but wide impact, and not as useful as generics or exceptions. Might be included in a moderate revision. Exceptions - This is comparable in usefulness to generics, but, while its impact on the text is liable to be smaller, it has great subtlety that can make it even more difficult to reach a consensus. It has a particularly large impact on implementations. It's interactions with parallelization are a particular concern. If co-arrays are to be included, wait until their impact on the standard is well understood before beginning detailed work. Probably requires a major revision. Units - This is most easily described using generics, although generic implementations will be inefficient and will allow only runtime error reporting. Put it off until you have a good specification for generics. Save for a major revision if at all. Decimal floating point arithmetic -- useful for the numerically unsophisticated, but I don't consider them the main user base for Fortran. This probably has a larger impact on the implementers than on the standard text. Save for a major revision if at all. Finally as to moderate enhancements making it easier to specify the minimum bit size of REALs and INTEGERs, so that codes are more easily ported to 64 bit machines would have my highest priority.