Subject: Re: EROS license
From: shapj@us.ibm.com
Date: Mon, 28 Jun 1999 10:59:19 -0400

>I predict that since Linus needed a source code control system, and
>didn't like CVS, somebody would have eventually written a free source
>code control system which he found acceptable...

Tsk, tsk, Ian!  That's not a prediction.  It's a belief.  Predictions are
falsifiable.

>I also don't think Larry sees the power in free software.  I believe
>that if he did, he would not require the program to pass the
>testsuite.

One problem is that success as a free software type requires better personal
marketing skills than success in the proprietary world.

However, I don't agree with the second point.  I haven't read the requirements
on BitKeeper, but in principle requiring conformance to a test suite isn't a bad
thing.  I believe Larry is doing this because he views branding as important.
His strategy is to brand BitKeeper strongly with his name, and to then ensure
that all distributions of BitKeeper are stable and robust so that customers will
know that it works well and pay him to improve it.

This could be done with a pure GPL license by building brand identity on a
trademark, but that is a much more expensive approach.  User's don't benefit
from paying for advertising.

Provided that there is a means to debunk erroneous tests, placing provisions
that set a lower hurdle for product quality is a definite benefit to the user.
The difference between the trademark strategy and what Larry has done is that
his strategy actively limits the ability of damaged goods to propagate and
reproduce.

Can you articulate more clearly why this is inconsistent with free software, and
also can you suggest an argument under which it's not a beneficial thing?  I'm
probably missing something obvious again.


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