Subject: Re: Thoughts on GPL
From: "Jonathan S. Shapiro" <>
Date: Mon, 2 Mar 1998 16:19:42 -0500

   Nobody is making money from selling software that's GPL'd.
   There are add-on companies, there are bundlers, there are
   support organizations, hardware vendors and so forth.  I need
   an example of a company that wrote an application, GPL'd it,
   and is now selling it, expecting to make a profit on those
   software sales alone.


Even *I'm* missing something here.  The reality is that software
companies don't sell software.  They sell some combination of
software, support, documentation, packaging, integration, etc. etc.
The fact that a GPL software company doesn't get to expense the
software component of the product does not preclude it from making
money, so you're setting up a bit of a red herring here.

   >> Even if your work isn't highjacked by someone, you aren't going
   >> to even begin to recover your engineering costs by selling
   >> support and/or enhancements.  I won't justify that statement
   >> here, but believe me, it's absolutely correct.

The issue, IMHO, is not whether the engineering cost can be recouped,
but whether a sufficient compelling advantage can be built to justify
the investment you need to build marketing and channel.  I have today
realized something I had not before considered that suggests it can.
I'll outline that below.

   Cygnus started by taking free software and selling support, so
   there's little capital investment there.  They also did
   enhancements, on demand, for that software for customers --
   again, little or no capital investment.  

   ... If they'd had to develop gcc from scratch to sell it,
   do you believe they could they have started the company on $20K?

Half right.  A better way to look at this is perhaps that Mike got
Stanford (grad stipend) and Penn (parents tuition, if you like) to
support his work on GCC, which he later converted into a hidden
investment in what would later become Cygnus.  Simultaneously, John
Gilmore invested his own time in GDB.  The $20k number is therefore
more than a bit fuzzy.

Aside: people argue that "labors of love" are not investment.  That is
both inaccurate and absurd.

In practice, Cygnus's capital investment was very nearly what it would
have taken them to build the software.  The difference is that they
got customers to pay for it, and they got to market 18 months
earlier. Key issue here is value of time to market.

Cygnus actually did some pretty serious development under the free
model, partly because of their leadership.

I shall respond to the following in a bundled response:

   This whole discussion started with the statement by someone that
   you couldn't recover initial software development costs if you
   GPL'd the result....

   Microsoft might be able to GPL applications and still make a

   I don't dispute that it's possible to GPL an application and
   then build the company that offers support and enhancements.
   But my statement was that a company with a larger marketing
   and/or support organization can take over your work...

I think the original statement was mine.  I also think, as I said
above, that you are mis-stating the problem: no one sells pure

If you'll indulge me, let me turn around and argue the opposing

The percentage of total product investment that goes into the actual
software is generally less than 10%.  System software may be an

What you get for that 10% is not really a compelling monopoly.  In the
absence of patent protection (which, I'll note, GPL'd software would
not preclude; an interesting twist), the only thing stopping your
competitor from entering your market is the *time* to redevelop, not
the *cost*.  If they give away the software, the developing company's
remaining advantage is expertise.

Aside: this suggests that a GPL model with a 12-18 month
redistribution delay would have much the same result as proprietary
development.  I'm curious if people feel that this would meet the
spirit of free software?

What I realized today is that there is one kind of market condition
under which a free software vendor can enter and dominate: any
established market lacking a single dominant player (or a pair of

In an established market, customers are in fact buying support,
contracting, and maintainance.  Provided that the market does not have
a dominant provider, purchasing decisions are made on the basis of three

	price, but only if the discrepancy is large

If, however, the incoming competitor is free, and the support they
provide is comparable, the newcomer has an opportunity to end-run the
politics issue; the customer can simply grab the software and see how
they like it.  If it provides a good technical solution the problem
will be solved before the politics can fuck things up.

Once a free product is gaining market share, there is no incentive for
commercial players to continue to innovate -- they cannot do so
competatively, and gradually drop out.  Look at the decline in
compiler vendors.  I believe (without any supporting data) that a
great deal of this decline was due to the promiscuous availability of
GCC, which was "good enough".

The free software provider now waits a while as the for-pay people
bail out of the market.  With a good enough technical product, they
will in due course own the market due to attrition.  If they do a
decent job of branding (which Cygnus has *not* -- think about their
revenue in light of that), they can then charge whatever they want for

The cost of playing this game is basically the cost of continuing to
innovate and thereby continuing to maintain a monopoly on expertise,
which prevents the competative entry you describe.  This may actually
be *cheaper* than dominating a market by conventional means -- I need
to think about it some.