Subject: Re: "incentive void" (was Re: A different patent covenant...)
From: <stephen@xemacs.org>
Date: Sat, 7 Oct 2006 21:29:06 +0900

Ben Tilly writes:

 > I already told you that I thought you were attributing someone else's
 > comment to me.  This reply verifies that you definitely are.  The
 > quote you're attributing to me was not my quote.  If you think that
 > I'm lying, feel free to go back and search.

You're not lying.  You're simply not getting the point.  You have not
said what it is that you do think in context.  You did write "I think
'so'" in a context where "so" referred to something that you have now
denied thinking about five times.

What am I supposed to do except ask you what "so" is referring to?

 > And this is the comment that clarified things.
 > 
 > There is some justice in what you say [about "relevant existing
 > technology"].

*sigh* It's not about justice.  Reread what I wrote about using
existing technology to support or deny patent applications.  The main
point is that the propatent lobby is going to take your idea as an
"unfair" restriction, and you're going to have a damn hard time
getting it through because anybody sitting on the fence is going to
see it as a double standard, once it's pointed out to them.  This
might even cause trouble with the judiciary.

 > debatable.  See, for instance, the space of *X on the web" patents
 > just a few years ago.  Lots of people (me, for instance) found them
 > obvious.  But the patent office was very generous in granting them.
 > Would-be patent holders were very aggressive about pursuing them.

And with a period of public review before any grant, the patent office
would clean up their act (nobody likes looking like an idiot,
especially when there's an indelible record that they were told so in
advance), and the aggressive would-be patent holders would stop
throwing away money on stuff that is a non-starter.

Trying to legislate an indirect criterion for obviousness is asking
for all kinds of trouble---like the potential for a patent system
that's even more useless as an incentive than Norbert claims the
current one is, but still grants patents.

I'm not saying that I think that *will* be the result, but consider
Seth's suggestion for mandatory patent auctions.  I'm pretty sure
that's seriously buggy as Tom and Don pointed out.  If your proposal
has such bugs, they're more subtle, but my mechanism designer's
intuition says the chance is pretty high.

 > What I'm suggesting is that we keep the patent office's default, but
 > we introduce a new criterion, which is that your invention has to have
 > been possible for a long period of time.  Then if nobody else came up
 > with it first, that is evidence that it really wasn't obvious.

Not for demand-driven innovations.

 > like the solution.  If you think the patent system has issues, and
 > those issues are the ones that I am naming, then you might think this
 > is worthwhile.

It should be obvious that I think the patent system has issues.

But I think the criterion you're suggesting (a) has an insanely long
waiting period driven by a criterion (comparison to term of patent)
that seems pretty irrelevant, (b) has been motivated in vague and
heuristic terms, and (c) is going to be complex to understand and
tune, however simple it may be to apply (and I'm still not convinced
about that, given your claim that Xanadu didn't exist as a technology
until somebody actually released a software product).

To expand on that last point, suppose I've got a technology which
makes no practical sense if it costs more than a click to traverse a
link, but can be described in terms of abstract "links" whose sample
implementation is gopher menus?  It's "obviously" a "doing it on the
web" innovation, and by analogy to the blue LED case, I've got the
patent.

 > To recap, I'd like to add a very objective criterion that would result
 > in far fewer patents being issued, but which would make the issued
 > patents far harder to challenge on the basis of, "Oh, that's obvious."

 >  The result is fewer patents, but those patents would be of much
 > better quality.  I think this would improve the quality of the system.

Other things equal, I agree that fewer patents of higher quality would
be a major improvement.  The other things are not equal, and
pragmatically I don't see how you're going to sell this to fence
sitters.

 > Those two subsets are very different, and are used in very different
 > ways.  Lumping them in together serves only to confuse the issue.

But it clarifies the politics that are going to come at you.

 > Particularly when you lump them together and do NOT clarify that
 > you've lumped them together.

Ben, you had a blind spot, and you've just spent the last five
messages berating me for guessing what you were thinking.

 > > Who says we aren't doing very well?  All I've seen are people who look
 > > at an invention *after* it has been patented and say "I could have
 > > done that."  Not a reliable witness, and worse, who's to say that the
 > > person in question is merely a competent practitioner?

 > Who says?  I do.  I've seen too many patents that I look at and say,
 > "That patent should not have been issued."

[...]
 > (That one is not only obvious, but there is prior art for it.  I
 > saw it on chess servers in the mid-90s.)

That's sorta problematic.  You know of prior art that you saw before
you saw the patent, and now it's obvious.  Well, of course it is, and
it was before you ever saw the patent!

We do not want to restrict patents to ideas which are so unobvious
that even after knowing the answer you say "I wouldn't have figured
that out."  We want to restrict patents to ideas that won't make
enough money to be worth taking to market unless they're protected,
but then they will.  I see no reason to suppose those properties are
correlated.  So any criterion based on trying to demand "bar-jumping
creativity" of innovations is going after the wrong thing.

The Amazon "one-click" patent flunks the "goes to market only if
protected" test, and in that sense is provably a bad patent on
economic grounds.  (Why?  Because the one-click patent is most useful
in a very public interface, and it's so useful that once you've had
the idea you will surely implement it and deploy it, patent or no.
"Going to market" doesn't *have* to be in the form of a product for
sale, although the innovator would like that to be the case.)

However, as you once said to me, you bet that I could look at your
pages and never realize which innovations were the ones that had
dramatically improved your business.  Similarly, I bet you could line
up hundreds of "web designers", and a distressingly small number of of
those would recognize the importance of one-click shopping---and thus
it clears the bar of "unobvious".

Just how many of those patents would you judge as "too obvious" *if
you didn't know they were patented*?

 > > I'm much more confident that great swathes of prior art have been
 > > improperly ignored and patents granted on non-innovations than that
 > > many patents have been granted on "ex-ante obvious" claims.
 > 
 > What distinction are you drawing between a non-innovation and an
 > "ex-ante obvious" claim?

A "non-innovation" is a claim such that if it were granted a patent,
prior art would infringe it.

An "ex-ante obvious" claim is one which any trained practitioner could
use as a matter of course in a context where it is applicable.

 > As contrary evidence I'll point to software.  The current enthusiasm
 > among lawyers and some big companies notwithstanding, plenty of real
 > innovations are created and spread through the software world without
 > need for patents to drive them.

Sure.  See my example of the one-click patent.

The problem that patents are supposed to address is threefold (at
least).  It may be a bad idea to try to address all three with one
mechanism, but I'm not sure how to split them out.

(1) Some questions are hard to answer, and require much cost to do so.
    That cost must be covered.  If not, the invention will not be made.

(2) Some answers have benefits that cannot be appropriated if they are
    made public, and it is costly to make them public.  That cost must
    be covered.  If it is not, the answer will have to be reinvented
    repeatedly.

(3) Some answers have costs (eg, giving the competition a leg up) if
    made public, and can be concealed at lower cost.  Those answers
    will be concealed, and must be reinvented repeatedly.

These problems are real.  However, there are many answerable problems
for which the costs of invention and diffusion are covered, and which
can't be cheaply concealed if used.  Those innovations will be
diffused without patents.

Nobody has ever denied this, so your anecdotes are just anecdotes.
Measurement is needed.

 > I can point to a number of examples where a nascent field (eg early
 > flight in the USA) was stifled because of patents.  I can point to no
 > examples where a nascent field was clearly assisted into existence by
 > patent activity.  (Beyond the initial patent that is.)

With *that* caveat, of course you can't, because *following*
development is inherently inhibited in part by the initial patent.
What else is new?

I mean, I can point to a number of examples where students were unable
to graduate because a few professors gave them Fs.  I can point to no
clear cases where students graduated because professors gave them Fs.
I guess you would be comfortable in suggesting that it's a good thing
for universities to refrain from giving grades?[1]

As for the case of airplane flight, it's suggestive, but in more ways
than just the one you have in mind (which is valid, of course, despite
the bias I point out above).  IIRC, the Wright brothers had several
patents, and were quite buttheaded about licensing them because they
didn't believe that a mono-wing plane could fly, while their would-be
competitors were similarly buttheaded about offering enough
compensation.  While Beavis and Butthead are a recurring human
phenomenon and I expect that we'll see unbelievable obstinate refusal
to license by the occasional curmudgeon forever.  But what we see
prevailing in today's is much more calculating, even on the part of
the trolls.

 > > That's what I thought you'd say.  So in theory the shoe maker can use
 > > Nakamura's invention without paying Nakamura, because it couldn't be
 > > invented until well after ten years before it was invented, so it
 > > isn't patented.
 > 
 > Having one permission doesn't give you all permissions that you need.
 > If blue LEDs are patented,

But blue LEDs can't be patented, because by hypothesis they depend on
technology that wasn't invented until too late.

 > BTW how does this work in situations where one country says something
 > is patentable and another does not?  For instance the USA issues a
 > software patent, I presume that the USA has treaties with France, but
 > France doesn't recognize software patents.  What then?

I believe France is stuck recognizing the U.S. software patent under
the treaties.  In most important cases I don't think it much matters:
for the many many little infringers, it's too expensive to enforce
across national boundaries, and the big potential infringers want to
sell in the U.S.  AFAIK, this is one of the main drivers (I don't
claim the most important) for software patents in Europe---they
decided that they weren't gonna get the U.S. to see it their way, and
there are a number of advantages to having everybody in step.

Bernard Lang or Norbert Bollow should be able to confirm or vice versa.

 > > IANAL, but I don't think the law works that way.  First, *they* take
 > > you to court, not the other way around.  This takes time; you get to
 > 
 > Point, I did say that backwards.  They take you to court, not vice
 > versa.  However in this kind of negotiation you're generally given an
 > opportunity to avoid court, and most companies leap on it.

Cite for that heuristic statistic?

[on estimating the cost of a patent troll]

 > I'm not sure of the details, but I notice that you've left the cost of
 > my lawyers out of it.  That's a significant consideration.

Stipulated.  I should have mentioned it.  I left it out because it's
hard to say how much of that cost is due to the mere threat.  Ie,
surely you're going to spend a couple $000 on simply having your
lawyer make the "70% chance of beating the patent" estimate.  But yes,
I suppose it could cost $50,000 or so which would be very painful even
if you end up caving in without a real fight in the end, and it could
probably cost more if there was a real fight.

 > Plus if they take me to court and I'm not using it by the time the
 > case ends, I'm sure they still get something for the fact that I
 > was using it at the start of the case.

That's supposed in my estimate already.

 > (And until recently, they could deny me the right to use anything
 > that depends on their claims while the case was being litigated,
 > I'm glad that no longer happens!)

Really?  The case that Tom mentioned said that *permanent* injunctions
are no longer allowed as a standard remedy.  But I thought that
temporary injunctions were still available, but I thought those were
always at the court's discretion.

 > > Sure, none of this is pleasant, but I suspect that you're painting a
 > > much blacker picture than average.

 > I don't know what is average.  I do know that this does happen.

But averages are what matters, not worst cases.

 > > That's not what Larry says is his experience.
 > 
 > Where did Larry say that?  (I'm not saying he is wrong, I just don't
 > remember the comment.)

A few posts back in the the thread.  Somebody brought up the patent
troll bogeyman, and he said lots of threatening letters are sent, most
just go into the trash, or similar.

 > My problem with a mandatory community patent review: there are so many
 > patents out there and the community has sufficiently little motivation
 > that the community doesn't do a very good review.  But then the fact
 > that the patent passed review can be introduced in court to make the
 > patent harder to defeat.  This is not a win.

Judges and juries are not that dumb.  If the reviews mostly suck, the
patent busters will introduce that in evidence.

I think the standard "cheaper to catch bugs early" argument will apply
here, even if the countervailing effect you suggest is significant.

 > What incentive is there for the adversaries to do their job well?

Well, some people just hate patents.  Some people will do it because
RMS says it's a good thing to do.  Some people will be working in the
field and come out to shoot down anything that a rival does.

 > It still costs money for the adversaries to litigate it.

*What* litigation?  I'm talking about a big Patent Tracker Wiki, where
in principle anybody can log in and post comments, which the examiners
can take into account in evaluating applications.  (Of course it can't
actually be a wiki!)

 > That your taste in music sucks? :-P
 > 
 > /duck

No need to duck.  De gustibus non disputandum.


Footnotes: 
[1]  BTW, I fantasize about that idea all the time, but on other
philosophical grounds, not because I think that the F's are an unjust
hindrance to graduation.