Re: Programmers -> Engineers; Engineers -> Programmers

[ryer@harp.camb.inmet.com (Mike Ryer)]

There are many first-rate programmers who have a general scientific education
and the interest and quickness to learn a lot about radar -- just don't let
them design your antennas.  There are many first-rate electronic engineers
who have the patience and flexibility to learn programming -- just don't let
them code your flight control system.

An electrical engineer can learn on his/her own how to program, but it takes
working in a team within an established software engineering culture to really
learn about maintainability, robustness, configuration management, defensive
programming, realtime constraints, etc.  (A succession of hard knocks will
also work, though this is tough on the first few projects).

A good programmer or software engineer should be able to find ambiguities and 
contradictions in the requirements document written by a radar designer, think
of boundary conditions and unusual scenarios, design for efficiency on 
processing the kinds on input that actually occur in practice, etc.  In 
other words, if they're good, they will have insight into the physics, 
electronics, and mechanics of the system.

A successful project wants software engineers who understand the underlying
science, and radar designers who understand software.  At least a few of each.
Maybe you'll find a superman who does quantum mechanics in his head *and*
writes code that's easy for other people to understand.  (I haven't).

To the extent that you don't have the excellent electronic and software
engineers who learn enough about each other's fields, you need communicators,
reviewers, requirements testers as well as code testers (note Arianne 5
story), and a lot of extra time.

So the summary is:  
  Q: Is it better to take engineers/scientists who understand the system
     and teach them how to program?  Or is it better to take programmers
     and teach them about the radar system?
  A: No.

Mike Ryer, speaking for himself, employed by Intermetrics.