NetNews Usenet Archive 1992 #16/NN_1992_16.iso/spool/comp/software/2977

Internet Message Format | 1992-07-27 | 2.7 KB

Path: sparky!uunet!mcsun!corton!cenaath.cena.dgac.fr!geant!vutien
From: [e-mail] (Vu Tien)
Newsgroups: comp.software-eng
Subject: Re: The origin of 'software-eng'
Message-ID: <[e-mail]>
Date: 28 Jul 92 08:38:03 GMT
Sender: [e-mail]
Organization: Centre d'Etudes de la Navigation Aerienne
Lines: 48
Originator: vutien@coque


For me, the process of software engineering is (ideally) something similar to
algebra manipulation in mechanical engineering, before committing to real production.
What I mean is that, *ideally* :

1) we start with some logic predicates that describes what the system properties are,
   along with properties that the outside world provides,

2) we develop the predicates, or factorise them, or simplifies some in the process.
   If there is some difficulties in solving, we may add some restrictions, as in
   algebra.

   The way each engineer factorises, develops terms, depends on the goal that is seeked,
   again as in algebra. What makes a good design, a bad design, or a transcendantal
   design, lays in how the terms are grouped. A good designer may not be a good inventor,
   by analogy, some theorems proven by Karl Friedrich Gauss (the king of maths) were
   quite botched up when proven, and later some body found a more elegant way to
   prove them again.

   The way terms are grouped and developed may also facilite reuse.

3) the ultimate goal of design is that the final step of the logical equations are
   ready for programming. By analogy, when the details of maths of a structure are
   known, then we can build a diesel engine, a turbine, a jet plane, etc..

Now all this is for an ideal world, because :

a) software is *very* complex, not at all like Copper Mining (cf SEI maturity model
    talks) and developing a problem may reveal itself an NP tesk, taking years
   and centuries of CPU time for designing large systems,
b) the incompletude theorem of Goedel says that it is impossible to prove a set of
   axioms inside its own frame. We need external asssertions. So human common sense is
   still needed.

However The closer we got to that ideal, the better.

To conclude, the French word "logiciel" (logics) for software may have influenced me,
but I believe that the software industry has tried too much to align itself to a
factory model, while software is more a design process that a production process.
This means that we need to investigate a software model similar to mathematical
development.

Smiley : Don't be scared of the word mathematical. All other engineering trades use it :
hydodynamics, physics, micro-electronics, bio-chemistry, so why don't we ?

---
[e-mail]
[e-mail]