A Random Paper on Software

In article ,
says...
Putting the application aside, this is a good paper on software.

It's a classic. Parnas spent many years doing the basic research on what
we now call "software architecture". He was one of the first to deal
seriously with the issues of designing big, complex software systems,
finding ways to keep them intellectually tractable.

Last summer, I read Parnas' Collected Papers. It was a bit embarrassing
to reread papers I had tried to read when they were first published. He
was presenting solutions to things I didn't know were problems.

Unlike most researchers, he put his money where is mouth was, leading a
team that built the flight software for a variant of the Navy's A-7
fighter. Just defining the requirements was a significant
accomplishment. Making it work in the memory- and cpu- constrained,
multi-input, time-critical environment of a flight computer pretty much
silenced the early complaints that Parnas-style designs couldn't deal
with the realities of real-time programming.


Parnas got a lot of flack about this particular paper. His conclusion
was, in some circles, politically unacceptable, so he just had to be
wrong. These "critics" didn't understand the paper, as Parnas wasn't
talking about the desirability of SDI, nor of the geopolitical
implications of SDI. He was just saying the software was beyond the
state of the art. The saddest example of this was when I heard Admiral
Grace Hopper explain how the SDI code could be tested. I have too much
respect for Hopper's straight-talking style to think she was trying to
deceive us, so I had to conclude that she just didn't understand the
differences between a COBOL compiler and a complex multi-threading,
multi-tasking, multi-processing, distributed, real-time, absolutely-
*has*-to-work-right-the-first-and-only-time-it-is-ever-used software
required for SDI.



It is a sad reflection on the commercial software industry that much of
Parnas' techniques from 20 years ago are still beyond the typical state
of the practice.


And for
those who believe that designing digital logic is designing software (many do
believe that since design entry now has a superficial resemblance to software
with Ada-like or C-like syntaxes)

Because conventionally software is sooo much more reliable than
contemporary hardware??
--
Kevin Willoughby lid

The loss of the American system of checks and balances
is more of a security danger than any terrorist risk.
-- Bruce Schneier