Subject: Re: Epistemology of FS
From: Ian Lance Taylor <>
Date: 01 Feb 2005 15:19:41 -0500

DV Henkel-Wallace <> writes:

> 1 - how much effort is expended on free software?  Who knows?  This is
> almost impossible to figure out, because it's impossible to find out
> directly for software in general.  Metrics such as KLOC etc are
> notoriously _irrelevant_.  For proprietary software you can at least
> measure revenues from the big public companies (MSFT, ORAC etc) and
> the DoL wage aggregates for the various employment codes and get
> _some_ idea of what's directly spent -- but even that's muddy.  For FS
> it's much harder.  Checking the # of Sourceforge projects doesn't tell
> you any more than looking at KLOC.

There are two different kinds of free software efforts these days.
One is the commercial kind, from companies like Red Hat and IBM.  That
kind can be measured, albeit muddily as you say.  The other is the
hobbyist effort, combined with the "scratch an itch" effort of people
who do not work directly on free software but contribute to it as they
work on other things.  That kind most likely can not be measured
reliably, any more than it is possible to reliably measure the effort
put into neighborhood soccer leagues.

You didn't mention another reason that it is very hard to measure the
effort expended on software, which is that the field is still immature
in many ways.  Different programmers have radically different degrees
of productivity, an order of magnitude or more, although salaries do
not range nearly that widely.  Measuring salaries is even more
approximate than it is in most engineering fields.

[ Re: lack of convincing evidence that it has significant economic
  impact ]

> 2 - Is even harder, even for the full IT market.  Figuring out the
> impact of IT has been a notorious problem that's kept legions busy for
> decades.  In the FS case it's worse.  Total money spent _purchasing_
> Free Software is probably less than $250MM.  All the other spend is
> buried.  The market (pseudo) research firms like gartner show a tiny
> share for Linux (and nothing for the BSDs which are everywhere)
> because most of those dollars aren't measurable.  The 1U server I just
> bought from Dell is $0 for Microsoft == but $0 for anyone else.
> The general form of this problem (economic value of declining costs)
> is a rich source research for macroeconomists trying to fiddle the
> inflation adjustors, but the specific case of FS is lost in the noise.

I think the economic impact on the IT market is profound, but as you
say it is extremely difficult to actually measure.

It's hard to determine whether IT is really a productivity enhancer,
particularly given the bogus productivity numbers produced by the
U.S. government.  But it's easy to determine that there is substantial
economic activity in the general area of IT.

The effect of free software on the IT market is similar to the effect
of cheap third-world labor forces on big industry: it undercuts
prices, and forces them down to the cost of labor plus shipping.  In
fact, the effect is stronger in the case of free software because 1)
the capital investment required is very low, 2) hobbyists work for
less than even third-world wages, and 3) shipping costs are in any
case near zero.  I've had a couple of conversations with people who
think that free software is destroying the U.S. IT industry in the
same way that some people say Chinese labor prices are destroying the
U.S. manufacturing industry.

Whether free software is improving the economy, through lower prices,
increased efficiency, etc., is an open question.  Free software is
changing the IT industry much as cheap data transfer is changing the
music and movie industries.  What the end result will be can not yet
be reliably predicted.