Subject: Re: BSD, GPL and macroeconomics
From: "Jonathan S. Shapiro" <shap@eros-os.org>
Date: Fri, 15 Feb 2002 06:48:16 -0500

> >What follows is merely a personal opinion, but since some of my work is
> >funded this way I have given the matter some thought. [...]
>
> Jonathan, this is a very good essay; very well put.  Even the
> anti-libertarian points were put in a way I could sympathize with, though,
> as I'm sure you expect, I continue to disagree.

In case anybody else took my posting as anti-libertarian, it wasn't intended
as such. I would characterize myself as distinctly pro-libertarian, but
skeptical of anything until it is demonstrated to work. I tried to make this
clear, but perhaps I wasn't clear enough.

>  From the reasoning in this essay, how do you get to GPL rather than LGPL?
> This matters to me mostly for non-ideological reasons -- GPL is
incompatible
> with Mozilla, and E is covered by Mozilla.  (I'm not arguing for Mozilla,
> but this historical fact is something E has to live with.)

First, I should have written (L)GPL, as there are clearly cases of things
that need to link but retain a separate identity. Libraries and scripting
engines are examples.

Where possible, however, I favor GPL when we are talking about national
technology policy. The reason is as follows:

LGPL provides link flexibility. An unscrupulous user of LGPL'd code could,
for example, extend glibc with proprietary routines. I say "unscrupulous"
because this behavior would not match the understanding of the community
about what LGPL is supposed to be for, but such behavior is within the scope
of the license text as I understand it. That is, a very fine-grain mixing
and matching of licenses is possible.

The tradeoffs I see are as follows: the finer the granularity at which a
company can mix and match licenses, the less incentive they have to reinvest
in the open portion, and the less "return" the public sees on their
investment. It has been my experience that the GPL barrier is very hard to
weasel around, and tends to force companies to put *some* reinvestment back
in. Usually the cost of this to the company is not excessive, and the
resulting system is often better from an engineering perspective because it
is forced to make effective use of its software elements as seperable and
reusable components. LGPL is much much easier to subvert in this respect.

On the other hand, I want companies to invest in free software, and I
therefore don't want to go as far as requiring that all software be free. I
believe that in the end what matters in the success of software is the
price/function/support tradeoff. Of these, the only cost that matters is
support. All other things being equal, free software will ultimately win
because its cost of entry is zero. We don't *need* to require that all
software be free. It's much more effective for the proprietary world to pay
for the process of being "nibbled to death by ducks."

I hasten to add that my use of (L)GPL in this discussion is not meant to be
exclusionary. The point, in my view, is to strike an appropriate balance
between usability and public return. Many licenses do that in some form.
Which is appropriate greatly depends on the code and the objectives.


Jonathan