[comp.software-eng] Boswell

michaelo@teklds.TEK.COM (Michael O'Hair) (05/03/88)

This has less to do with Sofware Engineering than it does project 
management, but read on if you like:

After surviving several real disasters, I'm of the opinion that a
"Boswell" should be assigned to every "major" project.  However, 
my vision of a historian's role is to gather data for use by 
management in preparing for future projects.  After a project is
completed, successful or not, the "project history" is collated 
and presented to management, who will, hopefully, read it and learn
what went right and what went wrong.

The danger is that the historian can become viewed as a spy.  It would
be possible to do the data gathering as an "oral history" with anonymity
guaranteed to those who desire it.  The interviewers would have to be
trained at investigative techniques (journalism, not police) to be able
to filter out, or at least earmark, self-serving statements, ad hominem
attacks, and the like.

Ideally, each team would do their own history and list what they did
wrong and what they did right, giving opinions/reasons for each.  It
would take very secure people, however, to be able to write down "We
made a bad choice in personnel/algorithm/hardware/etc and here's what
it cost."  The bright side would be to be able to say "We looked at
the problem carefully, came up with approach XYZ and reduced the time
to market by 10%."

That's my two cents worth.


		    Michael J. O'Hair
		    Design Test Integration Division
		    Tektronix, Inc.

Disclaimer:  "It's not my fault."

janssen@titan.SW.MCC.COM (Bill Janssen) (05/06/88)

In article <3058@teklds.TEK.COM>, michaelo@teklds.TEK.COM (Michael O'Hair) writes:
> ... After a project is
> completed, successful or not, the "project history" [should be] collated 
> and presented to management, who will, hopefully, read it and learn
> what went right and what went wrong.

Seems to me that that attitude right there, of hopeful historians presenting
something to "management", who may or may not read it, is its own problem.
This will only work if you have management *pulling* for those results,
asking for the data, eager to see how they can do better.

Bill