From mboxrd@z Thu Jan 1 00:00:00 1970 X-Spam-Checker-Version: SpamAssassin 3.4.4 (2020-01-24) on polar.synack.me X-Spam-Level: X-Spam-Status: No, score=-1.3 required=5.0 tests=BAYES_00,INVALID_MSGID autolearn=no autolearn_force=no version=3.4.4 X-Google-Language: ENGLISH,ASCII-7-bit X-Google-Thread: 103376,42427d0d1bf647b1 X-Google-Attributes: gid103376,public From: Ken Garlington Subject: Re: Ada Core Technologies and Ada95 Standards Date: 1996/04/25 Message-ID: <317FB33E.550F@lmtas.lmco.com>#1/1 X-Deja-AN: 151456986 references: <9604241816.AA07770@most> content-type: text/plain; charset=us-ascii organization: Lockheed Martin Tactical Aircraft Systems mime-version: 1.0 newsgroups: comp.lang.ada x-mailer: Mozilla 2.01 (Macintosh; I; 68K) Date: 1996-04-25T00:00:00+00:00 List-Id: W. Wesley Groleau (Wes) wrote: > > Point One: > But I'm certainly against the United States government meddling in things > any more than it already does! Especially in something that affects many > people in other countries. OK. Then you would agree that ACVC testing should be reduced or eliminated, in favor of something else. I can accept this. What (if anything) would you recommend be done in lieu of DoD "meddling" to improve compiler quality, or would reducing/killing ACVC by itself improve compiler quality? > Point Two: > We (Adaphiles in general, no one in particular) complain that Ada is not > getting the respect it deserves from the commercial world. We won't > change that situation by mandating that all vendors explode the cost of > their tools by expensive testing. Many of these potential commercial > customers don't test their own product. Why would they want to pay to > test the compiler? I think the better question is: "How much would these potential commercial customers be willing to pay for a high-quality toolset?" If the potential commercial Ada user is a college student, not much. If it's someone building "critical" systems (in the broad sense of "critical") probably more. How about this question, also: "How likely is it for Ada to penetrate the commercial world if the tools are not of significantly higher quality than tools for other languages?" > Point Three: > I think KG has missed an important point that RKBD has failed to mention-- > the DoD's goal in mandating validation has nothing to do with quality. > Validation assures the DoD that they are using "ONE language" instead of > 123* very similar languages. Actually, RKBD _has_ mentioned this point on more than one occasion, (although in other posts there has been suggestion that ACVC has _some_ relationship to quality) and I've agreed with it every time. I just don't see it as important to my argument. To me, it's more important to ask: Which is more important: language conformance, or (other) factors that relate to compiler quality? If we're spending a lot of effort on conformance, and that's not justified by the role of conformance to user needs,(and in particular if that emphasis causes other factors to be neglected), we're being silly. To fix this, we can either change the ACVC such that it does relate to these other factors, we can do other activities in addition to the ACVC, or we can do other activities instead of the ACVC. My personal feeling: If you tell me that I can either have a compiler that has very few bugs, good performance, etc. but some language extensions; or a compiler with no language extensions, but a lot of bugs, poor performance, etc., I'll choose the former any time. I can tell my users not to use an extension, if they want their code to be portable. If the compiler is of low quality, the code oftentimes isn't portable regardless of whether or not the compiler passes the ACVC! So, given that ACVC has to be supplemented/replaced by other measures, do we have these measures done solely by individual users on individual vendors, or is it _also_ smart to identify activities that _all_ (or at least a good majority) of users want done for all/most vendors, and do those in a coordinated (DoD/AVO? ARA? some kind of de facto standard?) fashion? Note that we don't even necessarily have to mandate a particular level for these measures. Different users may be happy with different levels. However, if every user has to use those measures, it will be more efficient to take them once for each compiler release, and share the results among all users. -- LMTAS - "Our Brand Means Quality"