Subject: Re: street performer protocol
Date: Thu, 18 May 2000 13:09:10 -0400

Stephen J. Turnbull wrote:
> >>>>> "Shimpei" == Shimpei Yamashita <> writes:
>     Shimpei> Stephen--can you give me (us) a quick overview of the
>     Shimpei> financial history of the semiconductor and biotech
>     Shimpei> industries and tell us why it's relevant to the free
>     Shimpei> software marketing model?
> The industries experienced enormous growth with high current profits,
> and invested large amounts in assorted kinds of R&D, with capital
> gained from IPOs and other issues of stock.
> As the industries began to mature (in the case of semiconductors) and
> failed to make good on the promise (compared to other promising high
> risk high return investments == dot coms, in the case of biotech), the
> amount of capital available compared to needs became scarce, cost
> control became important, and a few firms (Intel is the classic case
> in semiconductors) became dominant in R&D and innovation, and others
> were forced to specialize in manufacturing.  Everybody outsourced
> anything that wasn't "core competence".  Since most biotech is still
> very much basic research, you see feverish activity in "building
> alliances", because noone can afford to go it alone anymore.

This seems quite relevant to free software as well...but I think that
the separation between economic investment and intellectual property
changes the dynamic significantly.

> The analogy is that Crispin is claiming that the Red Hat and other
> Linux distribution successes are innovation (in software) driven, and
> thus the stock market rewards them for investing in development.  But
> investors could sour on them just as fast as they did on biotech---how
> is Red Hat going to finance its development programs then?  Maybe a
> couple of firms can do it on their current revenue, but I bet (based
> on the analogous history) that most of them are paying developers out
> of venture capital, and when the venture capital dries up, they'll
> have to cut costs.  The first thing that will have to go, when the
> bean counters take over, is basic development, because that doesn't
> directly move product and generate revenues.  This will lead to
> consolidation in the industry; I don't think that there will be more
> than two or three profitable (on the scale necessary to finance the
> kinds of development programs the major distributions run) Linux
> distributions after say three years.

Here is what I see as differing sharply.  I see at least 3 roles in a
big and successful free software effort.  (Often they are combined.)

1) Raw development.  This is what you are concerned that the
   distributions will not maintain.  I agree with you that in the long
   run it is unlikely to be the Red Hat's of the world driving most of
   this.  Instead it will be various specialist companies.

2) Integration.  Does Linus Torvalds do basic OS research?  I strongly
   doubt it.  However he and a fairly small group have the job of
   maintaining a definition of what Linux officially is.  (Ditto in
   Perl with Larry and the various pumpkings.)  Many believe that
   there are pressures pushing people to want to get things in the
   central version with free software, while in proprietary stuff
   there are similar pressures to diverge.  There is power and control
   in having a hand near (if not on) the tiller.  I strongly doubt
   that Red Hat will soon want to lose having Alan Cox on board.

3) Q&A.  Reporting bugs.  Fixing them.  Submitting fixes to the
   maintainers.  Conducting audits.  There is strong pressure on
   distributions to do this kind of work, and frankly this is the
   largest (and least pleasant) body of ongoing work.

I see successful distributions on big OS projects eventually
offloading the costs of development research.  Perhaps like Cygnus
did, they will actually do the development but they will charge
for it from people with an interest in seeing it be done.  Perhaps
they will leave the development also to other people.

I see dominant distributions attempting to maintain some control of
the overall direction the project takes.  This is a non-concern in
your other examples because IP is cleanly owned by one company.  It
is not something Red Hat or any other distribution can lightly

> And even they will be hard-pressed to keep up in many of the new
> fields: embedded OSes, clustering, etc.  These will be the province of
> specialists (often ex-members of big distribution development groups
> who are sick of the budgetary limitations and intergroup competition
> for support characteristic of maturing organizations).  At least,
> that's what's happened in biotech: extreme specialization in
> application fields as well as in basic technology.

Is Red Hat today the leader in embedded Linux, ports to mainframes,
or driving the clustering initiatives?  Not that I see.  I am not
sure who is on top in embedded Linux (there are many initiatives),
IBM clearly has the lead in mainframe ports, and based on current
publicity TurboLinux, Suse and VALinux come to mind for me on
clustering before Red Hat does.  (Not that I personally care about
any of those areas, I am just making a point.)

OTOH Red Hat put a lot of work into Gnome, a security audit, ease of