Subject: Re: [ppc-mobo] Re: GPL-like hardware design license?
From: "Stephen J. Turnbull" <>
Date: Wed, 6 Oct 1999 15:36:13 +0900 (JST)

    >>>>>>> "John" == John Metzger <>
    >>>>>>> writes:

    John> Stephen Turnbull writes:

    >> If the designs were publically available, they would be
    >> useful in education.  Etc, etc.

    John> Sure, but where's the profit in that? I'm not part of the
    John> subsidized education establishment.

    >> The point is, where's the _cost_ in that?  _No cost to you,

    John> No, there is a cost. The cost to me is that competitors get
    John> to sell hardware for less than I can because they get to use
    John> the R&D I paid for at no cost.

There is an opportunity cost to the public of not releasing the designs to
educational users for free; the loss of the students' chance to study

But there is no opportunity cost to you of simply releasing the design
to use in education (only), if we assume that they wouldn't be willing
to pay enough for it to recover the marketing costs of selling it to

We need to decouple the issues of costs and benefits, which are
empirical facts that are directly measurable in principle and whose
signs are least can usually be derived from theoretical arguments,
from the business and market models which are under debate.

    John> There is no cost to me (other than my time) to improve
    John> Linux, up load the changes and distribute them to everyone
    John> else. The same is not true of hardware. Hardware costs
    John> something to manufacture and distribute.

Not clear under what circumstances that's relevant.  See my response
to Ben Tilly <>.

And, of course, software also has manufacturing and distribution
costs.  Howcum Red Hat can charge 25 to 50 times as much as CheapBytes
for a physically identical CD-ROM?  We're still debating that one;
nothing that I've seen suggests that there would not be analogous
possibilities in hardware.

    >> no subsidy to education necessary._

In order to communicate the design to the educational institutions.
Of course educational institutions are subsidized in general; this
particular action (freely licensing to them) requires no subsidy.

    >> This looks like perpetual motion, but it's not against the laws
    >> of thermodynamics.  Of course the question, how to get you your
    >> profit in the case where certain such uses are permitted,
    >> remains.  Economically difficult, but not a physical
    >> impossibility.

    John> Explain how. I'm all ears.

You find a way of restricting the "no-fee" use to educational purposes.
That's _physically_ possible.

Eg, there is the Aladdin-style license (Ghostscript) which basically
says that if you publish modifications under the same license, you can
do anything with this idea or its derivatives---except make money.  If
you want to make money, you gotta talk to the boss first (and get a
license with different terms, perhaps in exchange for royalties).

That's not "free" as in "free software business," nor is it obvious
how to best do to it in any case (but note even Microsoft has special
educational rates for site licenses---their _goals_ are different, but
the principle of discrimination across customers is the same).  That's
why the FSB list exists: to discuss ways of accomplishing the goal of
making "good" returns within the self-imposed constraint of using open
source licenses (which probably makes discrimination harder, but as
usual, not impossible).

    John> Of course not. And if there is no way to earn a good return
    John> on investment then I won't use anything that touches a
    John> GPL-like license. That's my point.

Who said there's no way to earn a good return?  It depends on your
definition of "good", of course.  If "good" == "maximum", then open
source is probably not the way to go.  It is arguable that if
_everybody_ was forced to use open source licenses, then total profit
will go up in the long run.  But in the current environment, use of an
open source license is by and large going to reduce your possible
profit.  If "good" == "I can retire reasonably early", then maybe it
works.  There are several candidates for early retirement---from their 
current businesses :) ---on FSB.

    John> But by allowing people to make use of the public design
    John> without forcing disclosure of all improvements many more
    John> useful products may come to market sooner than if everyone
    John> goes off and uses proprietary technology. That is a benefit
    John> to everyone, even if the designs are not totally open.

The operative word is "may".  The empirical reality is dubious in any
given case; in theory arguments can be given either way (see below);
and on average we don't have enough experience with open source to
compute a reliable average, nor to extrapolate to the parameters
relevant to hardware.

    John> Why would anybody invest the money to make the changes to
    John> the hardware when they can't recapture the investment?

It is you that has given up and believes they can't.  The burden is on
advocates of publication-required licenses to show that one _can_
recover development costs, of course, before you start applying such a
license to your work; we don't ask you to buy a pig in a poke.  But if
your mind was as closed to new ideas in design as it is to new ideas
in economics, well, you'd have a different career entirely.  Difficult
is not impossible.

And ten years from now, the "easy" answer may be obvious.

    >> Why do you care who pays for it, as long as it gets paid for?

    John> If I have to pay for them out of my pocket I need the money
    John> to be replenished because, unlike Gates or Universities, I
    John> don't have infinitely deep pockets. I don't see what is so
    John> hard to understand about that.

I understood the fundamentals, oh, while I still counted my age on my
fingers.  It doesn't take a PhD in economics to understand your point.
So maybe you should consider the possibility that you're missing mine?

It is you that is refusing to understand that the whole point is that
although we need to find third parties to pay for at least some of the
investment, we don't necessarily need to do so according to the tried
and true business model of using a monopoly on an idea (= design) to
recover capital costs (= plant).

That old model does work well, and if your primary worry is about
money, I recommend that you abandon discussion of free licensing and
use a purely proprietary model.  Free licenses have little to offer
you, except that if others use them, there may be a free ride.[1]

We are here (on FSB) looking for new models that will help us achieve
our goals, only one of which is more money, and that one becoming
fairly satiated at low levels (compared to the billion-dollar figures
we are throwing around).

    >> But you know, there are many self-styled "inventors" who spend
    >> lots of time and money on their careers/hobbies, and never make
    >> a dime---and realistically, never have a hope.

    John> That' fine if you can afford it. I can't. I ain't that rich.

Sure you are.  Many of them are dirt poor, living on peanut butter
sandwiches and dreams.  You could, too; you just don't want to.  I
don't blame you, I don't either.

But from a social perspective, free licensing might be a way to
harness all that energy _before_ it goes down the entropy drain.

    John> The point I'm trying to make is about the license for this
    John> groups efforts and that OpenPPC should be freely available
    John> to for anyone to use, anyway they please, without further
    John> restriction because that is likely to get us more
    John> improvements to the hardware than with the more restrictive
    John> GPL like license.

That's a reasonable position, but it's backed up by nothing except
opinion.  It's also a common one in the free software community, used
by advocates of the "BSD" (more accurately, "XFree86") style of licence.

I agree that you're right, in the first round.  In the first round the 
provision of free technology (the 0th round) with unrestrictive
licensing will result in large coordinated (in the sense that they all 
use version 0 as a base) contributions from the proprietary sector,
and some further contributions from the free sector.  These small
extra contributions from the free sector are tiny as are their use and 
(public) development benefits, and the use benefits from the
proprietary versions are large, but the extra (public) developmental
benefits from proprietary versions are _zero_.

I don't know what to do with the future (proprietary) developmental
benefits, because although the proprietary companies will finance more
development from their revenue streams, some portion of this
investment will be duplicative and therefore non-productive from the
social standpoint.  I see no good way to estimate the amount of
overlap offhand, and it is also clear from the theoretical and (very
few) empirical studies that have been done that very competition that
creates wasteful developmental overlap is _necessary_ to drive an
incumbent monopolist to improve their product.  (In fact, many
theoretical results suggest that an incumbent monopolist has _less_
incentive to for further innovation than rivals do, as the monopolist
is just going to be eating her own lunch.)

BTW, the theory suggests that the amount of waste is typically on the
order of the value of the monopoly profit, which is typically on the
order of the use value to consumers net of production cost.  (That is,
although these numbers are ordered in the obvious way, the smaller
ones are not "negligible fractions".)

On the other hand, with restrictive licensing in the first round the
new free contributions are presumably the same, while extra
proprietary contributions have small use value (because there are
fewer of them, but _positive_ development value.  Thus, the base of
free designs will increase in value over time in this scenario more
quickly than in the unrestrictive license case.  This allows the
restrictive license to achieve (in principle) exponential growth of
the public knowledge base, whereas companies are limited in size.

You apparently believe that the net present value of the stream of
public improvements and their associated improvements in use value
each period is almost zero.  Linux proves that this is not true
universally; a "virtuous circle" is possible.  I know of no analysis
that differentiates between the hardware and software cases: all the
analyses you have proposed work in the same way to show that Linux is
impossible.  So something is as yet missing.

The statement that hardware incurs non-labor costs is not an analysis,
it's merely a relevant fact.  But I don't see how that necessarily
implies the infeasibility of free licensing, nor how it implies that
hardware design cannot achieve the virtuous circle that drives the
Linux phenonmenon.  Unless you also assume the classic business model
of monopolizing an idea to amortize plant investment is the only
possible one.  Then, you've already assumed away open licenses.  Game

So your argument for an unrestrictive _open_ license is presumably
based on the idea of coordination efficiencies (see footnote [1]), and
openness is not really a goal in itself?  Otherwise I should think the
argument unrolls backwards, and that you would argue that OpenPPC is
simply a hobby for people with more time and money than they know what
to do with.

    John> And that a major reason for that is because developing
    John> hardware is significantly different and more expensive than
    John> making incremental improvements to the Linux kernel.
    John> Software is cheap and easy to change.  Hardware is expensive
    John> and difficult to change.

And minds, impossible?  :-)

Look, if you're going to say "I'll believe it when I see it," that's
fine by me.  "I prefer to dream of things that never were, and ask
`why not'?"  Helluva way to run a business, of course ;-)  But why not 
let somebody else try it, since you consider it too risky?

If you want to argue "OpenPPC is very important to me, and I want to
develop products based on it and am not able to see how I can recover
costs under a GNU GPL-like license," fine.  Just declare your self-
interest, rather than insisting that the whole idea is false.[2]

Of course, this risk-averse position is a very important one, as there 
is a reasonable chance the bulk of the members will be risk-averse.
In that case, I can't advocate experimenting with a GPL-like license.
On the other hand, it's possible that some members would like to
experiment with it, and it is a disservice to them to change "higher
than normal risk of failure" to "impossible."

    John> The same applies to software.

    >> Yup, you can argue that.  But some people simply like the idea
    >> that their software, and anything based on it, will always be
    >> available to anybody who wants to create more software out of
    >> it for the sake of creating software.  An economy whose
    >> currency is ideas, not money.  The GNU GPL is a great
    >> instrument for this purpose.

    John> Actually not. I can't use GPLed software to build some
    John> products precisely because I have to publish the changes I

There you go again.

You have _decided_ not to use GPL'd software to build some products
precisely because you _believe_ that none of the business models you
are _aware_ of permit you to recover the financial investment plus a
reasonable profit if you publish the necessary changes to the source,
which is required by the GNU GPL.

There are several hypotheticals in there, all of which are under
attack in discussions of free software.  I admit that I must challenge
those hypotheticals; if you will not admit they need defense, I have
nothing further to say to you.  But I will not permit you to sweep
them under the rug and declare victory without pointing out your ruse.

BTW, my "economy of ideas" does not need to subsume the economy of
money.  Just because you can't cover the ante doesn't mean that others

    John> make to the software. That means the use of that software to
    John> create new products is limited more than it needs to be.

But conversely, if the new products are proprietary, the use of those
products to create new products is limited more than it needs to be.
In fact, it's a monopoly.  Ugh.

Limits bind both ways.  The issues are the rate of tradeoff between
the different types of limits, and whether we can affect that tradeoff
without going to the extremes of creating private monopolies on goods
with zero cost of reproduction (namely, ideas) or complete failure to
develop those goods in the first place due to the free rider problem.

    >> And other people simply disagree with you.  They believe that,
    >> at least for the domain of software, the GNU GPL is a more
    >> effective way of creating large amounts of useful, economically
    >> valuable products and making them available to users and
    >> developers of derivative products than the alternative
    >> licenses.

    John> We'll have to disagree on that one.

If you aren't going to produce or respond to analysis, let's.

    >> And yet others believe that it will work for hardware too, if
    >> done right.

    John> Well I'm waiting for someone to step forward and donate the
    John> design for their north bridge (the improved one mentioned
    John> above...).

Then you may be looking in the wrong place.  Different strokes ... the 
free ideas business models as yet have rather restricted applicability 
and are relatively untested.  I don't think anybody claims universal
applicability of the free ideas models; in fact there are clear
counterexamples that show that (in some toy models) they will fail big 
time.  You have presented some of them yourself!

But they _are_ toy models, and do not include all relevant factors.

[1]  This is hypothetical, not a reference to your interest in
OpenPPC.  Apparently what you have in mind with OpenPPC is not really
free licensing, but a model in which the relevant developers form a
club, and that the club is sufficiently large that balancing the legal
costs of designing and enforcing a cross-licensing arrangement against
the costs of leakage to non-club-members, you've decided that latter
costs are lower so you adopt a free license to avoid legal costs.

[2]  "Not In _My_ Back Yard" is a valid, if not particularly edifying,
motivation for asking experimenters to go somewhere else.  (I live
about 50km from the "critical incident" in Tokaimura, fortunately
upwind.  I definitely sympathize with NIMBY, especially when unskilled 
workers are carrying around fissionables in open buckets IMBY.)

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."