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=-1.9 required=5.0 tests=BAYES_00 autolearn=ham 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: 1996/04/18 Message-ID: <31764781.6534@lmtas.lmco.com> X-Deja-AN: 148234938 references: <00001a73+00002c20@msn.com> <828038680.5631@assen.demon.co.uk> <828127251.85@assen.demon.co.uk> <315FD5C9.342F@lfwc.lockheed.com> <31729038.20BF@lfwc.lockheed.com> <31735952.42B@lmtas.lmco.com> 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: 1996-04-18T00:00:00+00:00 List-Id: Robert Dewar wrote: > > The policy has been not to go beyond conformance testing in the minimum > requirement imposed by centralized policy on validated compilers. I think > that is a sound decision, since it leaves much more flexibility for the > project officers. "Project officers?" Sounds like you are discussing the use of Ada in a particular domain. I'm not. > It is up to you to specify an Ada compiler > that meets your requirements. One component over which you have no > choice is that it must meet ACVC requirements. Explain why I have no choice in ACVC... if I'm using Ada for a non-DoD project. For that matter, I always have a choice with any DoD requirement: I can ask for a waiver. > it would be a mistake > to mandate that ALL users of Ada in the DoD for ALL purposes have no > choice but to require all these features. And yet, it's _not_ a mistake to mandate ACVC for non-DoD users? Why? > This would increase tool > cost to no purpose for applicatoin areas in which some or all of the > above criteria are irrelevant. But there's an important assumption buried in this statement, which I have tried (apparently in vain) to exhume. Why is there only _one_ criterion appropriate for all application areas -- ACVC compliance -- particularly when there are users who are _not_ mandated to have an ACVC certificate? Furthermore, if for some reason we can't have more than one criterion, why should it be ACVC (as it is defined today)? As to the former question (more than one criterion), what is there about such things as the SEI CMM or ISO 9001 that only make it appropriate for one set of Ada compiler users? Particularly when I see a lot of Ada propaganda targeted to users who want high quality products, why are quality standards of this type so inappropriate? As to the latter question (ACVC is the best single criterion), why? What information do we have that ACVC is the best way to achieve the goals which most users share? Because it gives us compilers of higher quality than the average Pascal compiler? > Yes, I can see how you would like to spread your costs, but the fact of > the matter is that a non-critical accounting application written in > Ada does NOT need this level of testing. Really? Speaking for myself, I would be very upset (or my employer would be upset, depending upon the error :) if my paycheck were miscalculated due to a bug in the compiler. Nonetheless, you mention "non-critical" acocunting applications. Is the average Ada user developing non-critical system (using "critical" in the "mission critical" sense described in the Ada requirements and glossy brochures)? Is this the target user base? If so, what does the ACVC provide for emsthese users? Are there better (cheaper?) ways to provide what the ACVC provides? If your claim is that the average Ada user is choosing Ada for reasons other than to build high-quality systems (lowest cost, perhaps?), then I'll admit I'm on the wrong tack. Is there a language that is focused on supporting the development of high-quality systems, if not Ada? > On the contrary, such an > application might have other requirements, e.g. to pass the ADAR > decimal arithmetic tests, which for you would be irrelevant. Sounds like a domain-specific test. I was thinking more of a general-purpose measure of quality, that could be applied across multiple domains. > That's really the issue here -- how much to REQUIRE of all vendors in > all fields. Absolutely. For example, it appears to be common practice in Europe to require ISO 9001 certification of vendors across all fields. This is due to a recognition that poor quality products, although cheaper in price, are rarely cheaper in use. In the U.S., the DoD is moving toward SEI III certification for vendors as a prerequisite. Except for compiler vendors, of course. > The decision to go no further than ACVC testing as being > universally mandated is very deliberate, but it is assumed that > application domains will specify whatever they need. Could you direct me to the documentation of this decision, and the deliberations that led to this decision? Perhaps it could answer my questions. -- LMTAS - "Our Brand Means Quality"