Subject: scarcity and markets in commercial free software
From: Tom Lord <>
Date: Thu, 12 Dec 2002 14:23:49 -0800 (PST)


	  Natural sources of scarcity exist in the free software and
	  world do create markets.  The free software markets that
	  currently exist existed as proprietary markets before free
	  software came on the scene, and adopted themselves to free
	  software easily.  There was a proprietary market for for
	  strategic _proprietary_ R&D, but it can not easily adopt to
	  free software because it was built around licensing fees and
	  software ownership.  Nevertheless, by identifying how the
	  proprietary R&D markets operated, we can envision a form of
	  those markets that works in a free software world.

  Ok, so, for markets we need a scarce resource, multiple buyers, and
  competing sellers.  Let me see if I can sketch where I see these
  elements in the free software world, and where they exist for
  strategic R&D, and then re-raise the question: "How do we make this
  theoretical market into reality."

  Let's look just at the medium and big vendors, picture the
  free software world in four parts, and identify three markets
  along the boundaries that join them

       	market       \	 |    sell<------------.    such as employees 
	    	      -> |    ^                |	              
		      |       v      |         v        |
		      |      buy     |        buy       |
		      |              |                  |
		      |		     | MAINTAINERS      | INNOVATORS
                      |              |			| 
   such as Enterprise | such as IBM  | such as Suse and | e.g. GNU ca. 1987
   IT		      | and Suse     | mandrake         | Linus ca. 1992
		      |		     |			|
                  buy<->sell     buy<->sell             |
		      ^		     ^
		      |		     |
	   a market for		     a market for
	   "solutions"		     distribution

  The bits are not scarce, but "solutions" built from the bits are
  scarce. For example, individual attention to customers is a part of a
  solution;  reasonable assurance of future support is part of a solution;
  future responsiveness to changes in customer needs is part of a solution;
  update and admin assistance is part of a solution.

  The bits aren't scarce, but a sustained and reliable engineering
  process to turn those bits into a distribution is scarce.  For example,
  porting to specific hardware is part of distribution maintenance;
  advocacy and implementation of standards is part of distribution 
  maintenance; code vetting is part of distribution maintenance.

  The bits aren't scarce, but solutions and distribution engineering
  do create a (currently rather modest and over-supplied) employment
  market for tactical and custodial engineers (some of whom do dabble,
  as their overriding duties permit, in innovation).

  Now, the bits aren't scarce, but strategic innovation is.  Moreover,
  it produces the raw materials which enable the three markets that do
  exist -- so it is rather vital.  So, why is there no market there?
  I think I know why.

  The markets that do exist in that diagram are not novel.   They are
  slight reformulations of markets that already existed in the proprietary 
  world.  The buyers already existed, the sales channels already existed --
  free software businesses were able to enter these markets looking very
  close in form and function to proprietary sellers who were already playing.

  There _was_ a market for proprietary strategic innovation: the VC market.

  	(a) There isn't much VC capital these days

	(b) Traditional VC-generated innovation mostly produced single products, 
	    not distributions or solutions.  In the proprietary world,
	    vendors and distribution maintainers (usually under the same 
	    roof) and end customers were all buyers for those
            single products but that market was supported by scarce license 
	    rights.  But this is the _one_ mechanism that doesn't translate
	    at all from proprietary to free software -- so a VC has little or 
	    no incentive to fund strategic free software R&D.

  So we didn't have a ready-made market for strategic free software innovation
  and, in fact, we're down to just a few non-market mechanisms:

	(i) custom development for deep-pocketed end-customers

	(ii) self-funded development by innovators spending their own 

	(iii) innovation paid for by public spending

  Can there be and should there be a market for free software innovation?
  I think to answer that, we can look a little more closely at what 
  VCs used to do.

  VC-funded innovation had a couple of very interesting properties.

	(A) It made early, informed guesses about where future 
	    product markets would come into being.  It "thought ahead"
	    about the general direction in which innovation was
	    needed.  In effect, the VCs were paying their suppliers (the
	    innovators) to direct their attention in particular 

	(B) It managed risk by imposing competition.  VCs individually
	    and as a group, when they identified future markets,
	    created seeming _over-supplies_ of innovation.  They'd
	    fund 5 or 10 companies, all working towards the same goal,
	    then wait to see which 1 or 2 actually won.  In effect,
	    the VCs paid their suppliers to take risk -- to each take
	    on a different branch of a semi-systematic search aimed at
	    discovering the best technical strategy in the face of

	(C) IPO bubbles aside, license fee payers and corporate 
	    acquisitions paid for the R&D burn rate, plus a 
	    healthy premium for the project selection services
	    implicitly offered by VCs.

  So, the bits are not scarce, but the _attention_ and _risk taking_ of
  innovators is.  

  This is what led me, way back when, to propose (here, to VCs, and even
  to a few vendor VPs whose email addresses I miraculously found) the 
  idea of commercial free software research labs.

  Such labs would sell attention and risk taking.  The form I had in
  mind for these products was pretty straightforward: sell information
  exchange between vendors and labs (directing lab attention and
  implementing technology transfer), and sell research labor (the labs
  would be focused on producing running code, taking it to a stage
  where tactical/custodial engineering could take it over -- but the
  labs would also live in an environment where they were competing
  with other labs, and where most projects wind up losing).  I had in
  mind that, having identified a probable future market, customers
  would ideally want that same seeming over-supply of labs to explore
  the alternatives for how to meet those needs (e.g., the lab doing
  `arch' and the lab doing `Subversion' might both be R&D suppliers to
  Sun and IBM).

  I still think that makes good sense.  It can be implemented with
  much lower burn rates than we used to see from VC start-ups (for all
  the usual reasons and more).  Moreover, R&D in this form doesn't
  have to cover a VCs ROI -- it _only_ needs to cover the sum of the
  burn rates -- so it's a lower cost approach to R&D for that reason

  But it is far from automatic -- since nothing resembling this kind
  of market already exists.  There are no sales channels for this.  It
  isn't written into corporate budgets.

  How do we make this theoretical market into reality?