Subject: Re: Learning Curve; Software learning percentage.
From: "Stephen J. Turnbull" <stephen@xemacs.org>
Date: Wed, 05 May 2004 15:16:20 +0900

>>>>> "Taran" == Taran Rampersad <cnd@knowprose.com> writes:

    Taran> Right. This is the problem, actually. I'm trying to flip
    Taran> this all on it's head - where the focus is the learning
    Taran> than the cost of manufacturing.

But cost of manufacturing is exactly what the "learning curve" _is_
about, and the content of the learning exactly what it is _not_ about.

Just call it something else, and don't hope to find a magic number in
a book somewhere, because what you're looking for is a function of
engineering effort, architecture, and market structure.

    Taran> Consider a proprietary operating system such as
    Taran> Windows. Microsoft reuses code - objects - within it's
    Taran> operating system and it's applications. For example,
    Taran> Microsoft Office makes extensive use of the Microsoft
    Taran> Foundation Classes, which are found within the operating
    Taran> system. Thus, they do not need to write them. At an
    Taran> organizational level, they are 'mass producing' in a
    Taran> respect;

But they are _not_ "mass producing" applications using the FC: each
new application is a custom product.  The point of the learning curve
is that once the design of the product is _fixed_ there is a "magical"
effect that lowers cost as you produce ever more exact copies.

If you want industrial precursors for reusable software modules, look
at both the history of the IBM 360 family of hardware, and the
practice in the automobile industry of designing "platforms" (I think
that's what they're called by automobile engineers).  Ie, a chassis
that a number of variant engine, body, and interior designs can be
"bolted on".  I believe both of these cases have been intensively
studied by the operations research/industrial engineering guys.

They probably have a term for the effect that you can use.

    Taran> An abstract hypothetical learning curve could be derived
    Taran> from that, maybe. What do you think?

I think it's important and interesting and you should call it another
name.

The learning curve is about straightforward insights for mass
production: variety is bad, short production runs are bad, monopoly is
productively efficient, experience is an enduring advantage.

The world you want to talk about is much more complex.  It's not even
monotonic in experience (cf Fred Brooks: "A perceptive student
remarked that _The Mythical Man-Month_ recommended a recipe for
disaster: Plan to deliver the second version of any new system
(Ch. 11), which Chapter 5 characterizes as the most dangerous system
one ever designs."---p. 260 of the 20th Anniversary Edition).  And you
want market structure to matter (closed source vs. FLOSS), right?

The learning curve _can_ apply in the world you want to talk about,
but its insights are independent of the ones that interest you.


-- 
Institute of Policy and Planning Sciences     http://turnbull.sk.tsukuba.ac.jp
University of Tsukuba                    Tennodai 1-1-1 Tsukuba 305-8573 JAPAN
               Ask not how you can "do" free software business;
              ask what your business can "do for" free software.