Path: utzoo!utgpu!news-server.csri.toronto.edu!cs.utexas.edu!asuvax!ncar!ncar.ucar.EDU!ftower
From: ftower@ncar.ucar.EDU (Francis Tower)
Newsgroups: comp.software-eng
Subject: Re: Modifiability
Message-ID: <11714@ncar.ucar.edu>
Date: 6 Jun 91 21:10:39 GMT
References: <2192@Terra.cc.brunel.ac.uk> <1991Jun5.201807.13286@netcom.COM> <62683@siemens.siemens.com>
Sender: news@ncar.ucar.edu
Reply-To: ftower@ncar.ucar.EDU (Francis Tower)
Distribution: comp.software-eng
Organization: Climate and Global Dynamics Division, NCAR
Lines: 30

In general, the traditional waterfall life-cycle fails when maintenance is
concerned because those fine, fat documents which tell managers that a phase of
the development is finished are left to gather dust.  In particular the
software specification is often never revised  and the product begins to
diverge from the specification.  The result is that the source itself is
maintained and both old and new optimization knowledge is spread into the
source with details lost.

Balzer, Cheatham and Green in their classic paper 'Software Technology in the
1990's: Using a New Paradigm' (IEEE Comp. Nov 1983)  proposed that a program's
specification be formalized and that the implementation be mostly automated. 
The problem then of maintainability would be pushed into the specification
phase.

With greater acceptence (more in Europe and Canada) specification languages are
being used on software products (Specifying a Real-Time Kernal, IEEE Software
Sept 1990; 'A Formal Specification of an Oscilloscope' IEEE Software, Sept
1990) and they are not only making the maintenance function more tractable,
they are also detecting flaws in the specification as well as providing a tool
for exploring the effects of changes.

If you are not using some form of formal specification then the longer and more
frequently a program is modified, the more calcified it becomes.  Decisions
made long ago on data structures and optimizations act as ruts to force
maintanance into a narrow track.  My suggestions in this case are:
  1. A concise guide to programming style and practices,
  2. Choice of a minimal set of allowable structures
  3. Elimination of program optimization,
  4. High degree of modularity with as much generic code as possible,
  5. Effictive QA and CM functions with management involvement on projects.