Subject: Re: EROS license
From: "Stephen J. Turnbull" <>
Date: Mon, 28 Jun 1999 16:59:06 +0900 (JST)

>>>>> "Ian" == Ian Lance Taylor <> writes:

    Ian> What difference does it make if my feelings aren't rational?

Your feelings are irrational (in an economic sense) only if they lead
to contradictions.  I don't think it's worth accepting this burden,
even for the sake of argument, and even in the "usual" sense of
"against common sense."

The contradiction Russ has in mind is "Ian wants to increase the total 
amound of free software available to users and developers.  Dual
licenses incent developers and lead to more free software.  Ian's
dislike of dual licenses is therefore irrational."

But rather than say "So what if I'm irrational?", I think you should
simply deny the second premise.  As you proceed to do:

    Ian> If many developers belong to that class, and if the goal of
    Ian> the source license is to encourage contributions, then it is
    Ian> best to avoid dual licensing.

Or educate the developers who think so into "correct" thinking.  :-)

    Ian> Whether it is rational for me to belong to that class is not
    Ian> terribly relevant.

Only if you assume that education is impossible.  That could result
because you, say, hate the waste of paper involved in writing two
separate licenses.  But whether dual licenses and more resources
result in more free software or not is still under discussion.  If it
turns out to be true, then you have a contradiction between love of
free software and dislike of dual licenses.  So I would think
education on this point is possible.

You still can argue you are not irrational, but you have to make a
much more subtle (and selfish) point---that you are specifically
worried that your code will not be free, even if the total amount of
free code is greatly increased (I don't care how you measure it; you
just need to be able to say "more free code" in some relevant sense).

    Ian> When I explicitly reserve the right to gather up everybody's
    Ian> contributions and use them in ways that nobody else can, it's
    Ian> pretty clear that some of us are more equal than others.  The
    Ian> concept of the team is lost.

Of course it isn't.  Not everybody can be quarterback, and as Jackson
Brown put it, "the drummer came up and swept that girl away."  Roadies 
can hope, but they know in the end the stars get the girls.  Doesn't
mean they aren't part of the team.

    Ian> I argue that a dual license inhibits it.

Notwithstanding, in this case it may be true.

However, aren't you implicitly assuming in "dual license" that the
proprietor maintains rights in the software that you don't?  In the
case of the Aladdin license, for one, it's simply not true.  If you
make it plain that you retain the copyright on your patch and license
it to Peter Deutch on his own terms, he can't (legally) incorporate it 
in proprietary versions.  True, doing so is annoying to him (because
of the accounting), but I bet he'd be a sport about it and incorporate 
it in Aladdin Ghostscript.  I don't know how he'd handle proprietary
versions; I assume he would write to you (the author) about it.

According to Peter, this simply isn't a problem in practice.  Most
whole-file contributions come with "Copyright XXXX, Aladdin
Enterprises" already in the header (I don't know what legal standing
that implicit assignment has).

    Ian> A key privilege in any free software project is the ability
    Ian> to fork.  While a dual license need not prohibit a fork, it
    Ian> certainly gives it a weird conceptual complexity, another
    Ian> cause for concern, albeit arguably irrational.

I don't see how.  The code under the free license can be forked
without problem.

For example, there is a forked version of Ghostscript for Japanese.
In fact, two: one for GNU Ghostscript, and one for Aladdin
Ghostscript.  And there's actually a third fork, now defunct (the
current forks use different technology).  Peter doesn't like this (or
didn't several years ago), but that was because he wanted the Japanese
developers to submit a patch he could integrate into the main line,
and they didn't want to meet his technical requirements.  (What else
is new?  It took more than 10 years to merge Japanese GNU Emacs with
the main stream.)

Saying that it's conceptually complex means that you are maintaining
some sort of claim on the proprietary code.  That's irrational IMHO.

    Ian> I believe that we should all be free to fix problems in our
    Ian> software, which among other things means having the source
    Ian> code.  I believe that people who withhold the source code
    Ian> from us are acting against the best interests of society.

I am certainly willing to stipulate this.

    Ian> Therefore, I believe it is in my best interests, as a member
    Ian> of society, to encourage people to always provide source code
    Ian> to everybody.


    Ian> In particular, I should use my abilities as a developer to do
    Ian> so.


    Ian> Therefore, it is in my best interests to only contribute to
    Ian> projects which promise to always provide source code to
    Ian> everybody, while simultaneously encouraging other projects to
    Ian> make that promise.

Sounds like Aladdin Ghostscript to me.  Depends on how self-defeating
you want them to be about encouraging other projects, but lots of
drivers have been contributed to Ghostscript by commercial outfits.

    Ian> The belief that people make economic decisions rationally is
    Ian> the fundamental flaw in most economic analyses.

Of course I disagree.  The fundamental flaw in most economic analyses
is to assume "profit maximization", which is an excellent approximation
to "rational decision-making," without checking its accuracy in the
given case.  (Most cases where the decision-making process is of
interest fall into this category; where profit maximization is
accurate we can often fall back on even more approximate definitions.)

    Ian> People do not make rational economic decisions, except under
    Ian> weird definitions of rationality

Here's the one that economists agree on:

1) Given appropriately described alternatives A and B, the decision-
maker (DM) can say whether he likes A better, B better, or is
indifferent between the two.

2) Any alternative A is indifferent to itself according to the DM.

3) Preferences are transitive: if the DM likes A better than B, and B
better than C, the DM likes A better than C.  Boundary cases involving 
indifference are treated as you would expect.

4) Based on available knowledge about alternatives, the DM actually
chooses an alternative A such that the DM does not know of any
feasible alternative B which is strictly preferable to A.

Note the hedges "appropriately described" in (1) and "available
knowledge" in (2).  The former means that all variables that might
affect decision-making must be included; it does not prescribe any
particular form of description (such as "measurement in money terms",
or indeed even in physical terms).  You can base preference on love or
money---or any emotion you happen to feel.  The latter means that
various kinds of uncertainty do not preclude rationality in any way.

Do you have an objection to rationality defined that way?

For future reference, that's what _I_ mean by rational decision-making.

    Ian> (ones which try to approximate the way that people actually
    Ian> think, namely irrationally).

The place where the definition above has problems is that in order to
get the standard "marginal results", which link rational decision-
making to market prices, we need people to have enough information to
compute derivatives (we don't need them to compute derivatives!), and
that seems unlikely.  However, all the approaches I know of to the
problem show that to the extent that people approximate the necessary
information, rational decision-making will approximate the full-
information solution.  And there are strong incentives to improve

This last is also a problem; I don't know of any analysis of rational
decision-making about how to improve the process of improving ... the
process of decision-making, but this seems to be pretty second-order.

So rational decision-making does not require people to _think_
rationally, only that they come to rational conclusions.

    Ian> People only approach rationality in the aggregate--even then
    Ian> they rarely achieve it--and since we are talking about
    Ian> software developers here the pool of people involved is not
    Ian> currently large enough to make economic aggregate arguments.

Laboratory experiments with as few as three buyers and three sellers
result in extremely rational behavior, with correctly designed
institutions.  I think there are more than six developers on this
list.  As for the attempt to rig the game by specifying correctly
designed institutions," that's exactly what we're doing here---
designing institutions.

    Ian>    I view dual-licensing as critical to the success of many
    Ian> free software businesses.

    Ian> I think that businesses that rely on third party
    Ian> contributions would be best advised to avoid dual licensing.

I think that dual-licensing is essential, and that businesses that
wish to encourage third party contributions will have to carefully
construct policies that make it easy to present themselves as good
guys.  (PR is _not_ a substitute for good policies, nor for adhering
to them when it hurts; it is going to be necessary in this world full
of FUD.)

While I personally have no problem with the Aladdin license, I
wouldn't be surprised if such a policy didn't involve the GPL as the
public license, along with something like a public commitment (and a
history of performance) to releasing all internally developed and
contributed as free enhancements as soon as they are integrated into
the code base, possibly subject to some sort of "some of our contracts 
involve NDAs, but we insist on a time limit or we don't agree" caveat.

    Ian> I guess we'll find out which of us is correct over time.

I think you both are.  In particular, I think it will be fatal for
many FSBs to make public blunders in which they abuse the right to
take their own code proprietary, or they use licenses which make it
clear that they intend to do that with contributed code where

But (to make the kind of dynamic argument you made against my
proportional hackerishness assumption), I suspect that most of the
people who feel as you do are already doing free software, as much as
they can.  New recruits will feel less strongly about it than you do.

    Ian> sold by Cygnus's competitors, via an extreme form of
    Ian> price-dumping.

Shudder.  Please don't classify any behavior by an FSB as "dumping."

The only real worry is that having established a monopoly via dumping, 
the firm (or cartel) could use that to prevent emergence of
technically superior alternatives.  This can't possibly be true for an 
open source product!

It is essentially impossible to "dump" software, by any reasonable
economic definition of dumping.  They all involve distribution below
marginal cost of production, which is impossible for an information

In the real world, dumping (by the economic definition) invariably
involves (a) direct or indirect government subsidies (eg, California
rice), or (b) a protected domestic market (anything Japan exports),
and in either case it's probably a mistake.  It remains controversial
in economics whether purely strategic dumping can actually work in the
sense of establishing a monopoly that makes more profit over the
predation cycle than it would by using pricing policies that do not
depend on driving rivals out of the market.  (Japanese and Korean
industrial firms generally do not maximize profit, so they are a
different case.)

It is certainly true that it is tried, as evidence in any number of
anit-trust cases makes clear.  And certainly government subsidies to
dumpers add to their muscle.  But in general these should be
considered as transfers from stockholders in would-be monopolists (or
their government patrons) to customers.  (And this is true in the long
run for the Japanese firms mentioned above.)  This may be fatal to
some rival firms; that's a tough issue, but in general I think
protecting firms, even against such unfair competition, is a bad idea.
(Protecting their workers, not from loss of jobs, but rather from loss
of income, at least to some extent, is a better idea.)

Accusations of dumping in the real world usually come from firms which 
either have made mistakes that make them non-competitive, or which are 
incapable of convincing their labor unions of that fact.

University of Tsukuba                Tennodai 1-1-1 Tsukuba 305-8573 JAPAN
Institute of Policy and Planning Sciences       Tel/fax: +81 (298) 53-5091
What are those two straight lines for?  "Free software rules."