Subject: Re: small worlds and better than ransom
From: "Stephen J. Turnbull" <stephen@xemacs.org>
Date: Mon, 08 Oct 2007 10:27:07 +0900

Thomas Lord writes:

 > Stephen J. Turnbull wrote:

 > > I have no quarrel with your concept of ideal customer.  My problem is
 > > that I don't see why any of them will consider prepurchasing an ideal
 > > way to spend money, whether from you or from anyone else.
 > 
 > Risk management.

"Risk management" means different things for individual investors and
joint-stock business firms.  For individual investors, there is a
tradeoff between risk and return.  They can and should accept a lower
return in order to reduce risk.

Firms exist to shoulder risk.  They don't want to take unnecessary
risk, but (in theory at least) it is desirable that a firm minimize
risk only within constraint of maximizing profit.  The reason is that
if the firm trades off return for lower risk, the individual
investor's ability to diversify is blunted.

So, if in fact the firm's management judges that there is some "best
project in its field", then in that field they should choose to invest
in that project only.

 > Those models place all of the up-front risks of a project on
 > just one firm.

Correctly so, in my expert opinion.  Merely opinion, I don't know of
any "theorems" here, but definitely informed by expert judgment.

 > So, this is a high-stakes, high-risk, low (but positive) ROI kind
 > of research funding.

High risk, low ROI should be avoided at all cost.  This just doesn't
make sense, either from a pure economic standpoint (firms should
engage in high risk, high ROI projects) or from an institutional
("management for the managers") point of view, where risk should be
minimized to protect the managers's salaries and privileges.

So you think that the managers of Red Hat and IBM are just stoopid?

 > It works well enough that companies do it, but they can't afford to
 > do *much* of it and that translates, in this model, to only
 > exploring a very small number of projects.

Did I see you write "IBM" and "can't afford to do *much*" in
succeeding sentences?

 > Moreover, each investment of this sort tends to have a
 > lot of momentum so, once a company makes a plan to spend a
 > boatload on one speculative project, even when it becomes clear things
 > aren't going to work out well, it's hard to revise the spending
 > quickly.

Doesn't it make sense to fix the management problem that allows
"momentum" to take over, rather than locking in a guaranteed low
return?  Don't you realize that a department whose mission is to
distribute $X/N to N prepurchases, and must devote managerial
resources to liaison with each of the prepurchase vendors, is likely
to gain a certain amount of momentum in each of those prepurchases,
too?

 > Prepurchases are an alternative.    A firm could spend $X
 > on one, internal, open source R&D project -- or they could spend $X/N
 > on each of N pre-purchase projects.   And because the pre-purchase
 > transactions are vastly simpler than creating new internal development
 > teams, and because prepurchases are bought in small increments,
 > a firm can continuously adjust exactly *which* N projects it is playing
 > with.

You're ignoring the monitoring expenses, which will be substantially
higher than for an internal project (or equivalently result in
substantially poorer information), and will increase more than
proportionally in N.  You're also ignoring the expenses of enforcing
the prepurchase contract (eg, where the buyer thinks they're buying an
impending major release but a minor release is delivered with a major
release number and the developer claims the contract is fulfilled).

Of course you can get rid of both monitoring and enforcement expense
by simply acknowledging that you're not talking about "pre-purchase,"
you're talking about patronage.

 > In essence, IBM could have basically sponsored a professional
 > competition (i.e., paid competitors) in the form of pre-purchases.

I really don't think so.  If the competitors were going to be on an
even footing, they'd have to be big.  That's expensive, and not
"continuously adjustable".  Worse, a lot of what Eclipse is about is
the open architecture.  IBM would simply have sponsored a Blu-Ray
vs. HDD battle, to everyone's detriment.  Nobody in their right mind
does that.