Subject: Re: the walls have bears
From: shapj@us.ibm.com
Date: Wed, 2 Jun 1999 09:01:35 -0400

I agree with Russell's conclusion, but not his reasoning.  The real issue is
incentives.  A brief anecdote, if you will indulge me:

IBM released the Jikes java compiler under a quasi open source license -- please
no flames about the license, as I had nothing to do with it and it has already
been beaten to death elsewhere.

There is a conspicuous hole in that license.  It allows you to sublicense, but
does *not* require that you sublicense under the original license.  Check with a
lawyer on this, but my read is that you could, within the expressed intent of
that license, simply reissue the code under either GPL or "Tom's Proprietary
License" (TPL, not to be confused with PTL of cursed memory).

I had occasion to talk with the lawyer who authored this license, and pointed
this out to him.  He agreed that the hole existed, and said that it was
intentional.  I pointed out that users of Jikes were under no obligation to
release their changes.  He agreed.  Roughly paraphrased, I told him this was
pretty stupid.  He asked why.

I said: "Look, IBM is giving this stuff away.  There are a bunch of problems
with the license, mostly in unenforceable provisions, but let's leave that
aside.  The question is what does IBM get out of doing this?  This license
basically gives everybody in sight the ability to build a proprietary position
against us *using our own work.*  IBM has chosen not to productize this, but
that decision could change.  Choosing to give Jikes away was a reasonable
choice, but why on earth should we give up the advances that other people make
or let them take a proprietary position against us?  At least write the license
in such a way that people using our code will compete with us on even terms!"
After a little thought the lawyer in question agreed.

As it happens, the Jikes license was an experiment.  One of its purposes was to
use as a test case to get certain segments of IBM comfortable with open source
ideas.  The license itself, in my opinion, is a failure as an open source
license, but it served its purpose internally, which in the long run is more
important.

The point is this: companies will use MPL or GPL in preference to BSD simply
because they don't want to give competitors the ability to generate a
proprietary position against them using their own work.  It's simply a matter of
keeping the terms of the game fair.  Universities, for the most part, are not
concerned with this, and therefore find the BSD license palatable.

Jonathan S. Shapiro, Ph. D.
IBM T.J. Watson Research Center
Email: shapj@us.ibm.com
Phone: +1 914 784 7085  (Tieline: 863)
Fax: +1 914 784 7595

> I'm focussing on software distribution because that's the point I'm
> trying to make -- that a company that creates software for
> distribution is more likely to GPL it than MIT-license it....