Subject: A new(?) account for why free generates quality
From: "Jonathan S. Shapiro" <>
Date: Thu, 23 Apr 1998 10:21:39 -0400

> If you're a bad programmer, the proprietary-
> software-development route, as an employee anyway, is a much
> better way to earn a reliable income.  That, to me, is a sure
> indicator that the proprietary-software development *model* is
> not optimal, not even from the viewpoint of market economics.

I agree that the market economics is suboptimal.

It's reported in the US papers that the software market is chronically
undersupplied with programming talent.  If this is actually the case,
anyone who can warm a chair and has the required educational
credentials (certifications?)  should be able to generate a reliable
income, up to and including my dogs. [Hmm.  Perhaps *that* is how I
shall pay the rent. :-)] The test of whether the market is functioning
is whether an individual's income converges on their effectiveness.

In practice, the real market isn't a free market of competitors, and
hasn't been since the introduction of "middle management" in
approximately 1840 (Recommended: Alfred Chandler, "The Visible Hand").
In large companies, a significant component of "effectiveness" is
inherited from being on the right projects rather than from actually
doing anything.  This is part of why large companies find it so hard
to reward good people.

This suggests (to me) a reason for the quality of free software that I 
haven't seen suggested before:

1. Low margins forces such organizations to run lean in the
   engineering area.

2. Lean organizations more readily reward real value than political

3. Good producers therefore bubble to the top more quickly.

Large companies can hire good people, but in Silicon Valley I'm told
the average turnover is running less than 18 months.  In that amount
of time, a large company can't *identify* the good players, much less
move actively to retain them.

If this were true, I would expect free software companies doing
complex stuff to experience lower *variance* in engineer productivity, 
and therefore better schedulability.  Does anyone have any data that
would support or refute this?