Subject: Re: IC's mystery patent claims
From: <>
Date: Thu, 28 Sep 2006 14:47:23 +0900

Thomas Lord writes:

 > Quantization.  The existing market comes in units of 0years and 20
 > years with no other options.

 > I'm saying that between 0 and 20 there are a lot of sweet spots.

 > I'm also saying that allocation of a unit is at the indulgence of
 > the general public as represented by their government.  The bargain
 > driven for a 20 year unit ought to be pretty darn tough in most
 > cases.  The bargain for 6 years ought to be (available and) rather
 > less tough.

Yeah, but this is still not a "market."  It's a regulatory procedure.
That's my point.

It's theoretically possible to create market-like mechanisms for
political issues, but they're pretty damn cumbersome.  You can
probably do better than a nationwide referendum per patent
application, but 4 orders of magnitude (the difference between a
Presidential election which nowadays costs well more than $100M and a
patent application at $10K) is unlikely.

Also, merely allocating the monopoly does a pretty good job of smoking
out the demand-side value (a proportional share of which gets captured
by the customers, and that share is usually large, ie, near 1/2).[1]
That's automatically a pretty good bargain, and this is the extreme
case where the market doesn't exist at all without access to the
technology---the "consumer surplus" of course grows by leaps and
bounds as competing technologies, even "obsolete" ones, are available.

So all that's left for you to talk about when you say the bargain
should be tough is that the compensation should be correlated with the
"effort" or "creativity" involved---and that is what Larry Rosen means
by "resurrect the 'sweat-of-the-brow' theory."

 > Why would we want a one-size-fits-all patent term?

We don't.  "Optimal" patent terms would reflect the incentives for
innovation.  However, by definition we don't know what innovations are
going to happen, so it has to be rule-based.  We *can* do better by
looking at the conditions for innovation, but that is an industry-wide
property.  So pharma, autos, and software might have different optimal
term limits.

But you cannot give appropriate evaluation of individual claims by
saying "RSA should get ten years and IC should get 3" until you've
seen *at least* the patent applications (and preferably the market
response).  At that point, *the incentive stage is over*, and if you
try to bargain them down, *you're cheating*.[2]

 > How about some options, here?  Would I begrudge IC 20 years?  You
 > bet.  Would I begrudge them 1 year?  Prbly not so much.  Somewhere
 > in between is a spot that amounts to both incentive for inventors
 > and a good deal for the public, especially if there's a sane capped
 > price for a public license.

Again, after netting out the rhetoric about a "deal for the public",
your practical suggestion is a cap on the reward for coming into
possession of a good idea, *regardless of how valuable it is to the
public*.  Sweat-of-the-brow theory.

 > Markets, adversarial processes, democratic processes --- these can
 > be deployed for a fine-grained allocation of proportions of
 > monopoly.

Sure, but you have not described a process.  You've described
desirable outcomes (Tom Lord can go back to the architectural work
he's good at, IC doesn't make so much money that Tom pukes, etc).  You
have not described a systematic way to get there, except (implicitly)
making you IP Czar, and I bet you'll refuse the nomination.  (Read all
that boring crap?!  No way! :)

 > IP is good, imo.  Specific regimes, such as we have, can have nasty
 > bugs -- but the general idea is sane.

The bugs are inherent.  IP is a restriction on trade.  You can't get
around it, and you can't make money without exercising the right to
restrict.  All we can do is try to tune the system to maximize
incentive and minimize friction.

[1]  Technically, the point is that as usual marginal cost is
perfectly elastic at media cost.  This means that unless demand is
basically flat at the monopolist's price, most buyers will have a
"choke price" or approximate value well above that.

[2]  Well, technically speaking you could make that "the deal" and it
wouldn't be cheating.  It would still give much less than the optimal