Subject: Re: patent trolls and X-licensors
From: "Ben Tilly" <btilly@gmail.com>
Date: Thu, 8 Jun 2006 11:34:27 -0700

On 6/8/06, Stephen J. Turnbull <turnbull@sk.tsukuba.ac.jp> wrote:
> >>>>> "Ben" == Ben Tilly <btilly@gmail.com> writes:
>
>     Ben> It is not beneath me.  I post on public lists expecting
>     Ben> people to be able to understand that I have different degrees
>     Ben> of evidence for different opinions, and some opinions I have
>     Ben> less objective evidence for than others.
>
> Expect away.  Also expect to be misunderstood.

I do, I do, and in that case I expect to have to explain myself in
more detail. :-)

>     Ben> Here is the problem.  There are a lot of good ideas in the
>     Ben> wavelet world that get into one or two products (if that) and
>     Ben> that is it.
>
> What makes you think they'd necessarily get propagated if they weren't
> patented?  I see a dozen good ideas a week on the XEmacs lists; you
> can look 'em up in the archives, but you won't find them in the code.

Personal judgement call.

>     Ben> People don't put them into standards.
>
>     Ben> They can't get implemented in free software.
>
> That's a definitional problem.

What do you mean by a definitional problem?

Besides no matter how you define things, what I've just said is still
true.  If I've patented my idea, and you know about it, you're going
to be less inclined to put it into a standard unless I'm willing to
agree to terms that mean that people can actually use my patent.  Even
if I say that I I'm willing to license under reasonable and
non-discriminatory terms (RAND), a lot of people are going to
complain.  One of the big reasons why they will complain is because
licensing fees mean that people can't implement the patent in free
software.

>     Ben> And the result is that the ideas are *only* taken advantage
>     Ben> of in specialized products that really need what wavelets
>     Ben> provide.
>
> Really?  I believe you that they only get used in such specialized
> products; it's entirely unclear to me that the blame can be entirely
> laid at the door of patented wavelet technology.

I'm basing my statements now on memories of what I observed a decade
ago.  I accept that it is not clear that the blame can be entirely
laid at the door of patented wavelet technology.  I can't cite enough
examples to convince anyone either way.  But my impression then
certainly was that patents were the big issue for would-be
implementers, and I've run across comments since that indicate that
they remain so.

>     Ben> However those same ideas are not finding their way as quickly
>     Ben> into, say, digital photography.  Whenever I watch a slow jpg
>     Ben> fill in on a web page I can't help but think that I saw
>     Ben> software in 1996 that looked better at low bit rates and
>     Ben> whose quality filled in a lot faster as you downloaded data.
>     Ben> That is sad to me.
>
> But MP3s are everywhere, despite the encoders being patented, and the
> patents being agressively enforced AFAIK.

The standard became widely adopted and THEN they started enforcing the
patents.  It is exactly examples like that which make people leery of
patents in standards.

Also you'll find that support for a lot of audio and video
applications is sadly lacking on Linux because of patent issues.

But even though some patent holders are successful at getting their
patents into standards, that doesn't change the fact that people
generally don't like that, and it doesn't change the fact that there
is a considerable amount of resistance to getting caught in that trap.

> Nor did the Unisys patent really put a dent in the market for GIF.

See the above comment about submarine patents.  Also note that the
Unisys patent only affected encoders - decoders weren't touched.

> And surely there is unpatented undergrad textbook wavelet stuff that
> would beat the pants off interlaced JPEG, and plenty of naive (or
> rebellious) undergrads to implement it?

Unpatented wavelet stuff has found its way into lots of places.
Including video standards.

> Or maybe you're just getting elderly and misremembering the good old
> days?  ;-)

I don't think so.

>     Ben> The famous example [of an innovation that the organization
>     Ben> was unable to exploit] that people like to offer is Xerox and
>     Ben> the GUI.  But it is not isolated.
>
>     >> Emacs and GNU comes to mind. :-)
>
>     Ben> You wouldn't be biased, would you? :-)
>
> I am, but (at least until quite recently) in the direction opposite to
> the one you imply.  Sad, isn't it?

Odd, I'd thought that you'd be biased because you were a fan of XEmacs
and that whole Emacs vs XEmacs split.  Which happened a while ago.

>     Ben> However I'll note that I have yet to hear of a real case of
>     Ben> someone who needed to solve a technical problem choosing to
>     Ben> solve it by doing a patent search looking for a previous
>     Ben> solution.
>
> Don't you think that's the saddest thing you've written today?

Not particularly.  It is the wrong tool for the job.

>     Ben> I know that I could tell you all of the potentially
>     Ben> patentable ideas that my employers business has, and if you
>     Ben> implemented all of them and tried to compete with us head on
>     Ben> you'd lose money hand over fist for a very long time.
>
> Yeah, but what if I hired a real programmer to do the implementation,
> instead of doing it my own-wannabe-self?  :-) Specifically, suppose I
> hired *you*?  Don't say "you can't"; there are plenty of hackers for
> sale, even if you aren't.  And don't say "I signed an NDA and a
> no-competition agreement"; doesn't that have just as strong chilling
> effect as a patent (in terms of diffusion of innovation from the kinds
> of organization we're discussing)?

Getting real programmer wouldn't help.  Not even one like me who knows
the successful business well.

What distinguishes a good business from a great one is execution.
Which depends on a ton of small details.  For instance one recent
subtle interface tweak increased our profit margin by over 20%.  You
could spend weeks analyzing our site and not even notice the change
that is responsible.  We could face a direct competitor without that
tweak and bid up their cost of advertising to the point where we make
money and they don't.

>     Ben> Basic science historically has a lot more value in the long
>     Ben> run than the short run.  When we try to make the path to
>     Ben> value more straightforward we run a big risk of improving
>     Ben> short term gains at the cost of long term gains.  That's not
>     Ben> a win.
>
> Have you forgotten the GNU Manifesto?  We don't hack for money.  The
> immediate effect of improving appropriability of short-term gains is
> not going to be on basic researchers.

There is a lot of pressure on universities to judge themselves based
on how many patents they get and business they generate.  That
pressure does affect basic researchers, even if it reasonably
shouldn't.

> Granted, if we told all basic researchers that they need to get
> patents to fund themselves, there'd be a problem.  I certainly hope
> you understand that that is nothing I want any part of!

I accept that.  But changes to our funding system can easily create
pressures that way without coming out and saying that.

>     Ben> [Wavelet applications to OCR] ties back to my impression that
>     Ben> patents have slowed research and adoption of wavelets.
>
> Well, wavelets might have enough hack value to get somebody to *do*
> something in open source if that were legal.  But the problem I had
> with gocr, for example, was that it was just plain a low-quality,
> badly-designed application.  It wasn't hackable without an inordinate
> amount of effort.  I misdoubt that its problems were due to lack of
> availability of free licenses to wavelet patents.

Fair enough.  I haven't looked at OCR, you have.

>     Ben> Did you try http://jocr.sourceforge.net/?  Never tried it,
>     Ben> but I just looked for open source OCR and that is what I
>     Ben> found.
>
> I haven't tried the OCR yet, but that's not for want of trying to
> build the software.  ;-)  Only once, but hey, why break my head over
> it?  I've got what I need, and I pay my dues elsewhere.  Eventually it
> will get into DarwinPorts or I'll move my scanner where the USB cable
> can get to the Linux box and I'll try again.  In the meantime, I use
> the free as in free beer products.  I don't even have to build those. :-)

No criticism meant.

Cheers,
Ben