Subject: Re: street performer protocol
From: "Stephen J. Turnbull" <turnbull@sk.tsukuba.ac.jp>
Date: Mon, 22 May 2000 21:38:31 +0900 (JST)

>>>>> "Crispin" == Crispin Cowan <crispin@wirex.com> writes:

    Crispin> The references are fairly obvious:

Even to me.  I can use Google, and I can recognize an OS distribution/
kernel vendor (as opposed to say a subsidiary of General Foods).

My point was that this discussion should be phrased in a way that
doesn't depend on different people's judgements about relative
attractiveness of given _businesses_.  Especially if some participants
are not qualified to make that judgement at all---I'm not going to
bother to look up Lineo or QNX because I don't understand the
technology at the level necessary to make decisions about buying stock.

    Crispin> My perception is that Microsoft has taught a GREAT many
    Crispin> companies to avoid lock-in if at all possible.  As a
    Crispin> result, being OSS-based is a significant feature, and
    Crispin> pushing proprietary/closed/single-source technology is an
    Crispin> up-hill battle.

My point is that it is not possible to avoid lock-in; you just have a
choice of what kind of lock-in you wish to accept.  Companies that
choose lock-in to a moribund technology because there are several
rivals vying for customers, as opposed to lock-in to the monopoly
vendor of the next "industry standard" may very well regret that
choice.

    Crispin> The point being that even if Lineo, per se, fizzles out,
    Crispin> then several of Lineo's competitors in the Embedded Linux
    Crispin> market will be there to pick up the pieces.

But you just ignored my definition of "fizzle", applying to the whole
technology.  Lineo's competitors pick up the pieces.  Now what?  They
have no King's horses and they have no King's men, and they cannot put
Humpty together again.  They've exhausted themselves in the same price
war that caused Lineo itself to fizzle.  That's the scenario I'm
talking about.

Sure, they can "win" in the sense of driving QNX out of business, too,
but it's not at all clear that the customers are better off---choice
is _reduced_, and development of both alternatives was retarded.  (The
blank for telling me about all the development OSS generates for free
is below.  Here is not the appropriate place.)  And the "superior" QNX
technology is not gone, in fact it's on the auction block---and since
it's proprietary, it continues to offer the same revenue generation
advantage to the buyer at a lower cost than QNX itself paid.

An OSS market is very close to what economists call a contestable
market.  In a perfectly contestable market, the incumbent will be
forced to price such that Revenue = Development cost of Current
Version Software + # of Licenses * Price, leaving no extra resources
for development of Software TNG.

Contrast this with the monopolist, which has at least some pricing
power (by definition), thus allowing it to earn above-normal rents.  A
stupid monopolist can of course pay out those rents to shareholders or
executives, but why do you assume they're stupid?  A smart monopolist
will allocate some (or all---look up a biography of Andrew Carnegie)
of those rents to achieve cost or quality leadership.

This provides a substantial advantage over the OSS firm, one which is
not vulnerable to any improved strategies for the OSS firm (except
abandoning the OS principle).  The reverse is not true for OSS:
everything that an OSS firm can do, a proprietary firm can mimic to
some degree.  And they will.

But that doesn't mean that the emulation will provide all the social
benefits that free software does.  Just enough to make a proprietary
model competitive with OSS models.

    >> OSS-based business models have yet to demonstrate cost
    >> advantages in development (in the abstract sense of lower
    >> cost/performance, however customers define the latter) that are
    >> not (more or less) cancelled out by the competition adopters of
    >> that model must face from other firms who by definition have
    >> fairly equal access to their most valuable resource.

    Crispin> This analysis explains why it is not a significant
    Crispin> advantage in terms of profitability for RH vs. Microsoft
    Crispin> for RH to be in the OSS business.  It completely ignores
    Crispin> the fact that OSS makes the product so much more
    Crispin> attractive that it may drive the entire proprietary OS
    Crispin> sector into the ground.  The low barrier to entry cancels
    Crispin> RH's cost advantage due to OSS, but it does not cancel
    Crispin> the *customer's* advantage due to OSS.

The point was to explain why OSS does _not_ make the product that
attractive.  I'm aware that OSS advocates think that OSS is much more
attractive (I certainly find it so); it's amazing Microsoft has
survived this long, isn't it?  Well, uhm ... no, it's not maazing at
all, and it's not all due to anti-competitive behavior, either.  Sure,
it's easy to imagine some kinds of OSS driving the price down and the
proprietary firms out of business, but it's not clear that's a good
thing, not if it leads to OSS-dominated markets dependent on hobbyist
activity for innovation.

Those aspects of OSS attraction that depend on source availability to
customers are met by proprietary firms using devices like source
escrow.  Those aspects depending on handing out your trade secrets to
your competitors are generated by competition from happy competitors,
which means that unless you are getting enough freebies from the
hobbyists, you are not producing enough new development to match the
proprietary firms' hard-cash-funded efforts.

Remember, development costs are basically fixed and sunk.  "Fixed"
means that in a competitive market price competition can drive the
optimal price to marginal cost, which is well below the price
necessary to recover development costs for the current version, let
alone finance development of the next version.  "Sunk" means that to
minimize losses, loss-making firms will stay in business, aided and
abetted by their backers---but not to the extent of throwing good
money after bad by funding new development.  (This is the point of the
semiconductor and biotech analogies, especially the biotech case.)

There are frictions in these markets, such as lags before competitors
can enter; a labor shortage that guarantees decent developers to get
fairly high compensation---which in the options-driven high-tech
market means they're funding their employers; and source escrow is not
as good from the customer's point of view as open source because of
the lags involved in invoking the escrow mechanism.  I think these
frictions will, in many niches, give substantial advantage to FSBs; in
others, they won't be large enough and the advantage will go to the
proprietary firms.

But the amount of effort spent by proprietary firms in reinventing the
wheel is overestimated by OSS advocates, I think, and the overestimate
will become worse as the benefits that OSS has achieved from reusable
software become apparent.  I see no reason that proprietary firms
cannot find ways of achieving cross-firm reuse that do not involve
freeing their code[1] (and in about the same amount of time it will
take FSBs to figure out how to generate enough revenue to support
second and third generation product development).  In fact, those that
do will crush the ones that don't.  But I don't see FSBs "crushing"
the proprietary model, except in a few niches.

One strategy is the one adopted by the biotech industry and many
high-tech manufacturers: outsourcing large parts of basic R&D to
"alliance partners" (often in explicitly temporary relationships, aka
"virtual firms").  This is precisely Michael Olson's description of
Sleepycat, is it not:

>>>>> "MO" == Michael A Olson <mao@sleepycat.com> writes:

    MO> Our licensing revenue comes from customers who are proprietary

Note that word, "proprietary."

    MO> software vendors.  They're the only ones who have to pay us.
    MO> We don't generally consider our customers to be small
    MO> specialists selling components.  They are more often
    MO> developers who'd rather buy components than roll their own.

[It would be interesting to see how many other partners the downstream
vendors have, and whether firms like Sleepycat can find ways to
"eliminate the middleman" (the downstream vendor) by directly
partnering with other "component vendors".  Probably not; this is not
going to be an exact analogy to biotech (where there are quite well-
defined roles in a 5-or-so-step pipeline, rather than amorphous
networks of relationships that depend on the specific application, as
in software).  The integration function in software is going to
continue to be extremely important.]

That leaves the hobbyists to provide the source of energy required by
the laws of thermodynamics.  I don't think there's enough there to
support a whole profitable industry; although plenty to provide an
useful check on would-be (and actual) monopolists.  :-)

    Crispin> Where as I think that the crushing is about two years
    Crispin> away in some selected markets; considerably longer in
    Crispin> others.

But I don't see why.  I'm not interested in your opinion---obviously
you're willing to bet on your opinion, it's sincere and well-thought
out enough to satisfy you, I have no doubts on that score.

I'm interested in your arguments.  The one you've proposed so far says
"OSS is better," for the same functionality.  I think 'most everybody
agrees.  But I don't see how you support the assumption that the
functionality will be as good as what we've seen so far, given that so
far OSS has limited itself to a small fraction of the software
industry.  It is at least arguable[2] that OSS has "chosen" to work on
the "easy" problems (granted, often doing a better job in many ways
than the proprietary industry), and so the next few steps are going to
be harder and (thus) more expensive.

Those expenses will need to be funded, and OSS intentionally ties one
hand behind its back when it comes to funding.  That's a powerful
concrete disadvantage to overcome with vague allusions to customer
preference, many of which can be answered by proprietary models with
only a slight relaxation of "proprietariness."


Footnotes: 
[1]  This is precisely what software patents aim at, of course.
Whether they do it in the interest of society is a completely
different question.  There are other ways for proprietary firms to
achieve the same end, and I propose one; this is just an easily
observable, very classical example.

[2]  And in fact this argument is "received theory" in economics.

-- 
University of Tsukuba                Tennodai 1-1-1 Tsukuba 305-8573 JAPAN
Institute of Policy and Planning Sciences       Tel/fax: +81 (298) 53-5091
_________________  _________________  _________________  _________________
What are those straight lines for?  "XEmacs rules."