From bill@hcx2.ssd.csd.harris.com Fri Nov 15 19:41:04 1991 Received: from travis.csd.harris.com by dkuug.dk via EUnet with SMTP (5.64+/8+bit/IDA-1.2.8) id AA00616; Fri, 15 Nov 91 19:41:04 +0100 Received: from hcx2.ssd.csd.harris.com by travis.csd.harris.com (5.61/harris-5.1) id AA18489; Fri, 15 Nov 91 13:41:06 -0500 Received: by hcx2.ssd.csd.harris.com (5.61/CX/UX-5.0) id AA03389; Fri, 15 Nov 91 13:40:59 -0500 Date: Fri, 15 Nov 91 13:40:59 -0500 From: bill@hcx2.ssd.csd.harris.com (Bill Leonard) Message-Id: <9111151840.AA03389@hcx2.ssd.csd.harris.com> To: Z3000TT@AWITUW01.BITNET Cc: SC22WG5@dkuug.dk In-Reply-To: Z3000TT@AWITUW01.BITNET's message of 15 NOV 91 19:06:28 <$468927632S0404D19911115T180628.0001.Mail-VE> Subject: (SC22WG5.16) Attn: ( - from AWITUW01) X-Charset: ASCII X-Char-Esc: 29 > From: Z3000TT@AWITUW01.BITNET > X-Sequence: SC22WG5@dkuug.dk 16 > From: "Wolfgang Walter" > To: SC22WG5@AWITUW01.BITNET > Subject: Re: (SC22WG5.16) WG5 Ballot on String Module Mr. Walter's reasons for voting for this proposal struck me as slightly illogical, so I'd like to respond. > My vote is YES. > 1. The new work item ISO 1539-1 for varying-length character > strings was deliberately opened (more than 2 years ago) to > provide some very fundamental functionality in the area of > character string and text handling which could have been, but > is not provided by the F90 standard itself. The current > proposal undeniably provides this functionality and is > desperately needed as soon as possible - especially in view of > the availability of at least one F90 compiler on the market. Well, the module is available already. There is nothing to prevent you, or anyone else, from using it. That does *not* justify its adoption as a standard, however. How about we standardize one particular algorithm as THE way to solve linear systems of equations? Certainly there are many applications that need such a facility. Why isn't there a standard for one? By Mr. Walter's reasoning, there should be one today! > The later this module is standardized, the more different > versions of similar string handling facilities will be > implemented by individuals, thus destroying compatibility and > portability of these features. (We are already writing F90 > programs which use and need the string module. These are > NUMERICAL applications, not word processing or others!!!) I fail to understand why this needs to be a standard for you, or anyone else, to use it, or why the lack of standardization will make other people's implementations non-portable. If there is one thing that this string module has demonstrated, it is that such a facility can be implemented completely within the bounds of standard Fortran 90. No doubt there are other possible implementations, also adhering to standard Fortran 90. Why would those other implementations be non-portable? One of the rationale for separating the string facility from standard Fortran 90 was, as I recall, that it would be better to allow some period of competing implementations, so that we get the best we can. I am not saying that the string module does not have merit. Indeed, it is the existence of a portable implementation that argues, to me, for delay in standardization. Lack of standardization will not prevent anyone from using this module, nor from implementing their own version if they so choose. However, delaying standardization will give users, and vendors, a chance to evaluate this and other implementations to see what the advantages and disadvantages of each one are. For example, we have little or no data on the performance of this string module. We have very little data on how appropriate the interfaces are to a wide variety of applications. We also have a dearth of data about how compiler vendors might implement this facility as an intrinsic module -- that is, something the compiler recognizes and perhaps does special optimizations for. Standardizing now means freezing, for a very long time, a facility that could prove to be much less than optimal. If, on the other hand, this implementation proves itself, delaying standardization will cost nothing, as far as I can see. We have everything to gain and nothing to lose. So, my advice to Mr. Walter is this: Go ahead and use the string module. Let us know how it works, and what problems, if any, you encounter with it. Record any ideas you have for how the interface might be improved. Hopefully, others will do the same, perhaps even being so bold as to create a competing facility with which to compare. Then, when there is an adequate pool of experience, we can standardize. Bill Leonard Harris Computer Systems Division 2101 W. Cypress Creek Road Fort Lauderdale, FL 33309 bill@ssd.csd.harris.com --------------------------------------------------------------------------- As I was going up the stair, I met a man who wasn't there. He wasn't there again today. I wish, I wish he'd stay away. -- Hughes Mearns ---------------------------------------------------------------------------