larry@JPL-VLSI.ARPA (12/30/86)
GLOBAL DATA AND SW ENG The only reason I can see for Ada versions of ForTran COMMON or HAL/S COMPOOLS is for quick one-for-one translations of code from ForTran or HAL/S to Ada. If one needs global data there are better ways to do it. Norm Cohen discusses this at length (with appropriate cautions) in his _Ada_as_a_Second_Language_. See especially pages 293-7 and 370-3. There are times, of course, when global data is an optimal solution to a problem. It's the nature of any kind of engineering that trade-offs always have to be made. But the number of people with SW responsibilities who don't have any SW engineering knowledge is a national disgrace shared equally by education and industry. I find it sourly ironic that the military--supposedly the least foward-looking, "with-it" American institution--is the biggest supporter of SW engineering. BEGINNING TEXTS Cohen's book has two notably useful features. All chapters have a Summary section; some of the more difficult ones have a Details section which can be skipped by those who need a less complete discussion of the topic. Later one can make a second pass over the book, reading the details sections for a more complete understanding. Further, there's a lot of pragmatic advice about how to use language features. This makes for a book of 800+ pages, but the advice (and the examples and clear writing) actually makes it easier to read than shorter books, because it gives a context for the parts of Ada. ADA A SIMPLE LANGUAGE? Ed made a comment that reminded me of what I consider a myth: Ada's complexity. We always over-estimate the complexity of a familiar language and under-estimate the one(s) in which we're expert. (The fish-in-water syndrome.) And until the last year or two there weren't very many validated compilers around for people to use. So the viewpoint we got was that of the designers or implementors, which is very different from (what is becoming) the typical user. I suspect that in a few more years we'll be hearing a new cliche: that Ada is SIMPLER to use than previous languages. For, after internalizing the Ada mind-set and developing a reflexive mastery of its syntax, the user can forget about bits, bytes, subprograms, and data (at least until it becomes time to do trade-offs and optimizations). Instead, s/he'll write programs using elements in the problem domain. Larry @ jpl-vlsi.arpa
4526P@NAVPGS.BITNET ("LT Scott A. Norton, USN") (01/03/87)
----------------------------Original message---------------------------- > I find it sourly ironic that the > military--supposedly the least foward-looking, "with-it" American > institution--is the biggest supporter of SW engineering. It may be ironic, but it should not be suprising. My last tour of duty was on the 20 year old cruiser Harry E. Yarnell. This ship's tactical data system program was at least ten years old, and had components that were older than the ship. In ten years of maintenance, you pay for poorly engineered software over and over. Because the military can not live on the cutting edge of technology, but must use systems that last twenty years or more, we require better engineering in software, as well as in the supporting hardware. Lt Scott A. Norton, USN Naval Postgraduate School Monterey, CA 93950-5018 <4526P@NAVPGS.BITNET>