Subject: Re: Option & Software Pricing
From: "Stephen J. Turnbull" <stephen@xemacs.org>
Date: Mon, 08 Aug 2005 10:59:44 +0900

>>>>> "Laurent" == Laurent GUERBY <laurent@guerby.net> writes:

    Laurent> The only section of the text where I spot a value / price
    Laurent> swing is the following:

That's exactly what's wrong with the text, it ignores the difference
between value and price.  Most of the time price is the best measure
of value we have, but it's exact only at the margin.  For all other
purchases it is an underestimate, and there is nothing in theory or
practice to say that it can't be millions of euros for a given buyer,
or even almost all buyers.

The Black-Scholes theory, like all other pure arbitrage models,
assumes that value, though unknown, is the same for all buyers.  This
is obviously incorrect for software (unless there's only one buyer).

    Laurent> Well given that this is the first paper where I see
    Laurent> software support issues just even mentionned, I'd say
    Laurent> it's quite good just for that point.

Then you're looking in the wrong places.  Perhaps you've heard of Eric
Raymond and the Open Source Initiative?  The GNU Manifesto?  Lack of
application of Ito integrals does not make something bad science.

As for the contribution of noticing that something called a
"maintenance option" can be modelled as a derivative asset, it's
negligible.  The only thing this essay adds to previous work on
support issues is a mathematical way of deriving a huge underestimate
of the relative value of software services based on a huge
underestimate of the absolute value of software services.  Giving a
very precise estimate of a parameter known to be estimated
inaccurately is not good science.

    Laurent> CTOs number one priority is support quality, so yes legal
    Laurent> monopoly rules that push support quality towards zero are
    Laurent> an area where CTO would love a bit more capitalism/competition.

I'm sorry, that blade has an edge on each side.  If they would love it
so much, why don't they just buy source licenses, then?  And in fact,
we observe them doing so for mission critical apps all the time---
that's what keeps Red Hat and Sleepycat in business, after all.  What
we see is CTOs conducting a balancing act among quality of app,
quality of support, and expenditures, while vendors conduct a similar
act on the other side of the market.

Furthermore, what is meant by "number one priority" is that this is an
area where a competent CTO can give his employer a big competitive
advantage over less competent ones, by doing a better job of balancing
those options.  It's not obvious to me that leveling the playing field
by removing ex ante options (ie, at software purchase time) from both
buyers and sellers in favor of giving buyers (only) more ex post
options (ie, at software maintenance time) is necessarily good for
anybody.  Least of all the CTOs whose current expertise becomes much
less valuable to their employers.

    Laurent> why there's a derivative market and no untradable options
    Laurent> (where did you get that idea from? :).

It's in the B&S paper, of course.  Didn't you read it?  It's free
economics, you can find it in any library. :) In any pure arbitrage
model, at disequilibrium prices the commodity becomes untradable in
the sense that absolutely all traders are on the same side of the
market---you can't find a trading partner.  That is why arbitrage
models are so widely accepted, the incentives are extremely clearcut
and powerful where they apply.  That's also why they're restricted to
financial economics---they simply don't apply to other kinds of
markets in a straightforward way, because excess demand in other
markets is not discontinuous at the equilibrium price, even with the
fairly extreme assumption of trader symmetry made in the Black-Scholes
model.

    Laurent> I agree that a simple B&S model is very weak to get
    Laurent> insight in software and support prices, but by adding
    Laurent> default risk and some transaction cost I think you could
    Laurent> get more information at the cost of more model work...

Sure, but you're still stuck with the problem that you know you have
an underestimate of the value of the software, and you know you have
no way to estimate the size of the bias (since the price of the
software itself does not reflect user value, it only reflects extreme
competition among sellers).


Footnotes: 
[1]  Of course you can have a proprietary version of glibc; that
simply implies that it is undistributed.

-- 
School of Systems and Information Engineering http://turnbull.sk.tsukuba.ac.jp
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.