Subject: Re: Value returned from free(d) software
From: ian@cygnus.com
Date: Thu, 4 Dec 1997 00:45:25 -0500 (EST)

In fsb bb@wv.com (Brian Bartholomew) writes:

>The moral forces that constrain human behavior in the libre-ish world
>are particularly strong.  Libre software users are unusually far from
>what economists call "rational".  However, I'd like to see a free
>software economic model that makes a positive money return in a purely
>game-theory analysis.  It should thrive in the better communication of
>a "perfect free market", and not be based on artificial scarcity.

In my opinion, applying game theory to economics is like the old joke
of the drunk crawling around under a lamp.  ``What are you doing?''
``I'm looking for my quarter.''  ``Did you drop it around here?'' ``Oh
no, I dropped it down the street.''  ``Then why are you looking
here?''  ``This is where the light is.''

That is, using game theory in economics is switching from the complex
problem which is difficult to solve to a simpler one which can be
solved.  It's a different problem, though.

That said, one game theory model of libre software is simply the GPL.
Consider a market in which all software happens to be under the GPL.
Can that world be invaded by a non-GPL program?

I would say no.  A small proprietary improvement will be emulated by
other people in the GPL code, because using the proprietary version
will impose an ongoing cost as the proprietary version falls behind
changes in the GPL version.  A large proprietary improvement will not
happen, because it would have to significantly better--substantially a
new program--and would have to be developed while the GPL version is
being improved.  Few will be willing to invest in such a chancy
proposition (remember, we're assuming rational game players here).

This type of economics arguably already applies in certain small
fields dominated by GPL software: Unix editors, cross platform
compilers.

The positive money return comes from consulting and support.

I'm definitely waving my hands here, but I think that the case could
be made.

You're probably more interested in a case in which libre software can
invade a market dominated by proprietary software.  I can see at least
one case where that can happen.  1) The total revenue in the market M
is X.  2) There are several companies with operating budgets much
larger than X.  3) These companies have customers who, in order to
purchase the products made by these companies, must also purchase
products in M.  If all these conditions are true, then the large
companies can rationally jointly invest in libre software which they
can provide to their customers.  Of course, it is also possible that
one of the large companies will attempt to enter market M themselves.
However, this may lead them out of their area of expertise, so
producing the libre software may be a safer bet.

A market that may potentially enter this situation is web advertising.
You can't advertise on the web without a web server.  Companies that
wish to advertise on the web have budgets much larger than the
existing market for web servers, but they may not wish to enter the
web server market themselves.  The argument here is that these
companies can rationally act to ensure that there are always free web
servers.

Since in fact there are already free web servers, and in fact they
already dominate the market, the conclusion is that Netscape's and
Microsoft's web server revenues, such as they are, may be endangered.

I don't know much about Russ Nelson's market, but it's possible that
it could be described by this model.

To bring up the ``tornado'' model again, this is similar to the
process described by that model by which late adopters anoint a single
market leader.  I'm arguing that in certain cases--those cases where
the late adopters have money to burn relative to the size of the
market--it can be rational for the late adopters to join together to
create a libre anointed market leader themselves.

>Here's my favorite example.  Suppose aliens landed tomorrow and handed
>us a Star Trek food duplicator.  This would copy objects like digital
>copies work today: perfect copies and nearly free.  Would we say,
>"sorry, take it away, it will break our economy", or would we use it
>to eliminate world hunger?

In reality, we humans don't act rationally, and we don't act in
unision.  I expect that some entrepeneur would acquire the plans
somehow, start selling the duplicators, and would make a killing.
Then the economy would collapse, but it wouldn't matter as much since
the duplicators would be widely available and prevent the worst
side-effects.  Then the environment would collapse in heavily
populated areas, probably followed by large scale migration
accompanied by ethnic and religious warfare fueled by cheap weapons.
Fortunately, this is fantasy, not even science fiction (although The
Diamond Age by Neal Stephenson is science fiction, and has some of the
same elements--he skipped over the collapse of the economy, but think
about how society must have changed from today to the one he
describes).

>My impression for some time has been that Cygnus is making money in
>direct proportion to how much artificial scarcity they create.  Why
>else would they advertise their stuff as 'better than free software
>off the net'?  If it really is better, and it really is libre, then
>put it up for ftp and eliminate the differential.  My impression is it
>probably is better -- release engineering and integration issues makes
>a huge difference in usability -- but it's as minimally libre as
>legally necessary and getting less so all the time.  For example,
>witness the license change on the POSIX environment for Windows.

The cygwin32 (POSIX environment for Windows) license essentially
permits people to buy out of the GPL, so I don't think it could be
fairly described as creating an artifical scarcity, nor as minimally
libre.  The code is still under the GPL for anybody who chooses to use
that version.

You suggest that Cygnus should put the GNUPro release up for FTP, but
why should Cygnus do that?  It is a cost to Cygnus for no benefit.
Besides the obvious reasons that it is a cost, it is a cost because it
means that people will come to Cygnus and get software, and will then
ask Cygnus for help.  When Cygnus refuses the help, as it must, Cygnus
will seem mean-spirited and the reputation of Cygnus will go down
(and, arguably, Cygnus will really be mean-spirited, but one of the
implications of being in the support business is that you don't
provide free support; if you do, what are your customers paying for?).

This is not a far fetched scenario; similar cases have already
happened with the Cygnus release included in the O'Reilly Programming
with GNU Software book.

As somebody else pointed out, one reason for Cygnus to distinguish its
releases from net releases is to brand the Cygnus release.  It's
better for Cygnus if prospective customers, who may not know anything
about GNU, consider the Cygnus toolchain in the same way they do
competitor's toolchains, rather than considering Cygnus as being in
the different business of providing support for free software.

So while it is undeniably true that Cygnus is moving away from libre
software in certain cases, I don't think your analysis of the reasons
is correct.

In these comments about Cygnus, I stress that although Cygnus is my
employer, these interpretations of Cygnus's actions are entirely my
own, and are certainly incorrect at least in part.

Ian