Subject: Re: economic efficiency of free software
From: "Stephen J. Turnbull" <>
Date: Sat, 07 Feb 2004 12:57:30 +0900

>>>>> "Phil" == Phil Hughes <> writes:

    Phil> I see some useful data from the book-publishing racket to
    Phil> toss in here.  Major publishers (that means the companies
    Phil> that just do books rather than companies that are in an
    Phil> industry and also publish books just in that industry) have
    Phil> a success rate of about 1 in 10.

    Phil> What makes this interesting to a free software developer is
    Phil> that while the publisher of the books has some costs
    Phil> (editing, printing and distributing) the author is also
    Phil> taking a big chance with a lot less control. The book
    Phil> publisher is happy with this 1 in 10 ratio and will put a
    Phil> lot of effort into turning a profitable book into a very
    Phil> profitable book rather than trying to help something that is
    Phil> marginal become profitable.

    Phil> Besides the downside for the book author, you have the same
    Phil> thing at the retail level. Bookstores want to maximize their
    Phil> income per inch of shelf space. What that means is that they
    Phil> would rather have copies of a book that somehow got a lot of
    Phil> press on their shelf that copies of an assortment of books
    Phil> covering the same topic in different ways.

Very interesting observation!

So let's turn that on its head.  Suppose the locus of the problem is
not the publishers, it's the bookstores.  The traditional bookstore
faces a physical/physiological constraint: because eyes must see the
books, books must take up substantial shelf space (and this is why
shrink-wrap software comes in boxes that are usually more than 100
times the volume of a CD jewel case).  The publishers turn around and
push the successful books because that helps their customers, the
bookstores, move stock.

An elementary exercise: applying this idea, why is
successful (at least in terms of revenues, I dunno their P&L)?

For (non-game/non-porno) software, we should be able to push the
Amazon strategy back upstream to the publishers, because nobody buys
an application on the basis of seeing it on the shelf.  (They may
choose a big eye-catching Red Hat box rather than a buried-in-the-
stack Debian jewel-case, but the decision to buy Linux has nothing to
do with that picture of a fedora.)

I think there is a real opportunity here.  Figure out a way to get
people to pay for software indexing and review, and channel their bug
reports and feature requests to the developers, and generate a revenue
stream from that.  It would work like electronic publishing in
academia, except that (see below) the problem is aggregating the
signals of customers, not peers.

OTOH, maybe Red Hat has beat us to it.  :-)

Note that several people have proposed similar "virtual software
development firms" here on FSB, but always with the hackers driving
the bus.  The point of this idea, on the other hand, is to put people
like Phil (you're the Phil, right?) and Tim O'Reilly at the
wheel---people who aren't necessarily hackers themselves, but know
about topic and author selection, marketing, and distribution.

    rn> The open source world doesn't have the price signal.

    Phil> Doesn't have or doesn't currently have.  I don't have a
    Phil> solution here but it does seem like there must be some model
    Phil> that would "work".

First, I think you misunderstand what Russ means by "price signal"
here.  It is a happy accident that (where markets are undisputably the
best way to organize trade[1]) the same price signal that (1) tells
the buyers and sellers how much to buy and sell also (2) extracts
compensation from buyers for their use of social resources and
compensates sellers for the resources they provide, and (3) allows
outside observers to measure the social value generated by the market.
All of this "falls out" quite directly from the so-called "marginal
analysis", ie, differential calculus, and the identity of price with
marginal value that the market enforces.

This "price signal" is therefore much more than a guide for producer
activity, and in fact role (3) is the most important from the point of
view of moral philosophy and economic policy.

Now, how is it that the market enforces identity of (market) price
with marginal value?  Because all buyers pay the same market price,
the one with the lowest value, the marginal value, also pays the
market price.  Thus marginal value must be "close" to market price.
The same argument forces marginal cost to market price.[2]

What goes wrong _with the economic analysis_ in the open source model?
Very simply, there is no constraint to pay market price, and most
buyers don't.  This means that the buyer who actually pays market
price is randomly selected, and I can think of no reason why market
price will be pushed to marginal value.[3]

Of course, it's possible to do various kinds of market research, and
discover values outside of the market mechanism.  But that's an open
admission that the price signal has been lost.

Of course, we _can_ put a price on developer/maintainer/QC/support
staff effort, and on S&H.  This is what and sourceXchange
did, as brokers, and Cygnus did, as a vendor.  But we cannot charge
the customers depending on the value society receives.  This means
that there's no way to differentiate between something that was barely
good enough to take to market, and a killer app.

Especially not from the point of view of the venture capitalist.

    Phil> That is, demand would be established,

How?  This is the $64 billion question.  In the proprietary model,
there's no problem.  You ask for the money, and when it comes, you
count it and divide by price.  That's "demand."  With free software,
though, you ask for the money, and it comes---but if your price is
substantially higher than reproduction cost, demand -> 1.  Joe
First-User makes copies for all his friends, and _all the other
"demanders" are not counted._  This is a problem only for economic
statisticians: it's exactly what _we_, as free software advocates,
want!  Right?  Except that it makes it really hard to accurately
assess, even ex post, what the "bigger wins" were.

    Phil> initial development would be funded

How?  I mean, where does the money come from?  The users aren't
compelled to pay, are they?  More important is, "Without the price
signal [in its social value role], how do you know which projects
deserve initial funding?"

    Phil> and "profit" would be shared in such a way that future
    Phil> projects could also get seed money.

Well, sure.  But again, the problem is "how?"  If it's "profit", not
_profit_, who decides?  The proprietary scheme says "vendors who have
satisfied customers in the past are a good bet to satisfy them in the
future---let them keep the profits (they'll reinvest to make more)."

The approach used to allocate research grants, peer review, seems not
suitable, either.  Developers are not the peers of typical users, and
will choose to satisfy goals unsuited to J. L. User, while users are
not the peers of developers, and will regularly mistakenly award
contracts to vendors who (in good faith, even) promise Arpege but
deliver skunk musk.

    >>> then the extra 90% is market friction.  The users would be
    >>> much better off if they could give $50 million directly to the
    >>> Word developers, perhaps through an administrative
    >>> organization like the United Way that added on 30% for
    >>> administrative costs.

    rn> United Way??  How about suggesting the Public Software Fund?
    rn> :-)

    Phil> Yeah, that's closer but does it work?  That is, do any big
    Phil> potential buyers get excited about it?

I would think that "big players" would be handling things themselves,
rather than donating their R&D budgets to the PSF.  Eg, Henry Ford
didn't contribute to the United Way, he established the Ford

One real problem is that there are lots of people who would be willing
to ante up in terms of $10 or $100, but even with PayPal you've still
got the problem of "should I contribute $10 to Conventional Project X,
which has some funding (but not enough) and momentum and is likely to
produce something useful with my $10, or to Wild-Eyed Hacker Y, who is
a certifiable genius, but who might starve to death before getting to
market, after eating his last meal with my $10?"

The PSF can help a lot by providing something of a scoreboard, and
maybe (eventually) a reviewing service.

[1]  And it turns out that where markets are inappropriate, if you
understand the "nature" of the distortion, the price signal, though
"wrong", is wrong in known ways.  That is, it can help you estimate
the size of the distortion and the degree of remedy needed.

[2]  How close depends on how smooth the distribution of buyer values
and seller marginal costs is.  In the limit, when both distributions
are continuous, "close" reduces to "equal".

[3]  The technical problems are actually even bigger than this, but
this is close enough to give the flavor of how the price signal gets
drowned out.

Institute of Policy and Planning Sciences
University of Tsukuba                    Tennodai 1-1-1 Tsukuba 305-8573 JAPAN
               Ask not how you can "do" free software business;
              ask what your business can "do for" free software.