Subject: Re: EY invests in online patent exchange
From: "Stephen J. Turnbull" <turnbull@sk.tsukuba.ac.jp>
Date: Thu, 16 Sep 1999 12:29:26 +0900 (JST)

>>>>> "Craig" == Craig Brozefsky <craig@red-bean.com> writes:

    Craig> Brian Behlendorf <brian@collab.net> writes:
    >> There is another possibility, one which may be in this
    >> company's future, I don't know (and I'm offline so I can't
    >> check).  Premise: if we can't get rid of software patents, the
    >> least we can do is make licensing them simple, predictive,
    >> perhaps even a known quantity.

Ah, I'm not so far off the track if BB comes up with a closely related 
idea.

    Craig> I think that only serves to legitimize software patents,

Agree.  Nothing wrong with that, if they're a net social benefit.
(Devil's advocate here, you understand.)

    Craig> and in doing so would have very negative effects on
    Craig> software in general,

This is very debatable.  (Although I tend to agree.)

    Craig> Free Software in particular.

This is more plausible.  But still debatable.

    Craig> Regardless of how low the transaction cost is for
    Craig> purchasing rights to use a patent, a program could require
    Craig> purchasing on the order of dozens of patents, and as more
    Craig> software techniques are patented, the number will just get
    Craig> higher and higher.

Zero times any arbitrarily large finite number is still zero.  Don't
speak in absolutes.  Furthermore, there's a continuity here in
economic practice, so "low enough" is good enough:

    Craig> As that number increases, and you have different license
    Craig> criteria for different patents (CPU, user, transaction
    Craig> based) the purchasing process becomes nearly impossible to
    Craig> manage.

For you with your current level of expertise.  Somebody, perhaps even
you, will make a lot of money (or contribute the bulk of the potential
riches to the social benefit in your case) by figuring out how to
manage that process.

Most patents are barely this side of obvious.  Those patents will be
licensed and crosslicensed in bulk.  Important patents will be bundled 
with ancillary patents that facilitate implementation of the important 
patent in common situations.  Etc, etc.  Even an academic like me can
figure that out; somebody who really knows what's going on should be
able to do a lot better.

In any case, what you say is all true for hardware patents, too.
Patents seem to work there.  Where's your beef?  Be explicit.

    Craig> It does not help users, but just creates another teir of
    Craig> middlemen producing nothing, and adding their markup.

This is the usual "What's wrong with steal^H^H^H^H^Hsocially
appropriating an invention once it's been invented?  The inventor
can't uninvent it!" fallacy.

"Does not help users" is trivially true only if all past and future
inventions and their date and place of invention have already been
immutably set.  If in fact the original logic of patents is true, that
financial incentives will increase the rate of innovation (not mere
invention), then the burden is on you (well, actually, I've done part
of the work in a previous post; more and stronger arguments would be
helpful) to identify the ex ante expected social costs (not merely the
ex post private ones, which are obvious enough, but are exactly
balanced by the revenues received by various producers) of the patent
system applied to software, and show that they outweigh the benefits
of the increased rate of innovation.

    Craig> That is an understatement.  It is not suboptimal, it is
    Craig> seriously damaging.

We know that you and many people believe this.  That's not evidence.

I'm sorry, but all the arguments I've seen so far have focused on the
ex post costs to users.  These arguments basically boil down to "Why
don't I have the right to take what I want?"  It is not true that the
argument is "Why don't I have the right to take what I want if it
doesn't hurt anybody else?"  There is a plausible mechanism (slowing
the rate of innovation) that does impose costs on others (besides the
developer in question, who I am leaving out).

The burden is on you (us; I'm with you in spirit) to show that (1)
software patents actually have the perverse effect of slowing
invention and innovation; or (2) anyway the costs of the software
patent system outweigh the benefits of the increased rate of innovation.

I have a plausible argument for (1), but it's not a flat-out knockdown 
gold-plated proof yet.

    Craig> Free Software and enforced software patents are in no way
    Craig> compatible,

Sure they are.  Invent around the patents.  (Costly, but feasible.  So
much for "in no way.")

The fact is, that if it is that hard to invent around the patents,
then simultaneous invention is rare.  If simultaneous invention is
rare, then the unique inventions are making a substantial social
contribution.  If unique inventions are making a substantial social
contribution, it may be good social policy to encourage more
innovation, and patents are one way to do that, a way which so far has 
shown no evidence of destroying the innovativeness of American
society.

    Craig> legitimizing software patents by claiming to provide a
    Craig> low-transaction cost mechanism for licensing them only
    Craig> encourages their use, and makes it more convenient for the
    Craig> user to give up their freedom.

"Legitimizing [drivers licenses] by claiming to provide a
low-transaction cost mechanism for [driver training] only encourages
their use, and makes it more convenient for [would-be drivers] to give
up their freedom."

Every contract involves giving up some freedom in return for some
benefit.  The freedom to drive a car is not inalienable; I don't think 
the freedom to use RSA public-key encryption should be inalienable.
The question is "are we getting enough back from patent-holders to
make their monopolies (== our loss of freedom) a net social benefit?"

    Craig> Any Free Software company that pursues software patents
    Craig> IMO, does not deserve the name.  It does not fit into the
    Craig> O.S. definition as I read it either.  This practice would
    Craig> also not fit with the GPL as I understand it.

Arrant nonsense.  The patent can be Freely Publically Licensed, with
or without a "poison pill" clause.

    Craig> As a strategy it's entirely unworkable.

Tell that to the NYSE, who said exactly the same thing about
electronic trading.  Now their backs are up against the wall just as
the wrecker ball comes through....

Creating markets is a very powerful, though not universally
applicable, method for solving economic problems where the goal is
end-user benefits.  Once again, the burden is on absolutist nay-sayers
to justify their positions.  I think the case against software patents
can be made, but it's non-trivial.

-- 
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 two straight lines for?  "Free software rules."