Re: [tlug] Zurus distributions experience

["Stephen J. Turnbull" <stephen@example.com>]

Lars Kotthoff writes:

 > Hmmm. After reading your reply, I'm under the strange impression that we
 > actually mostly agree :)

Perhaps.

 > >  > So you're saying that basically innovation only happens when
 > >  > somebody has a commercial interest in it?
 > > 
 > > Yes.  Economically, that's more or less the characterization of
 > > innovation, because turning something into a product usable by a large
 > > number of people is deadly dull, painstaking, hard work.
 > 
 > Certainly not done by academics,

Well, no, actually academics probably do as much of it as anybody.  R,
for one example.

 > but there's loads of examples where open source software is turned
 > into something usable (for some definition of the word) by a large
 > number of people -- without commercial interest.

Sure, there are loads of examples of free software innovation, but the
user value is dwarfed by the worst software in the world (I mean
Windows and Office, of course).  The difference is the degree of
"market penetration".

 > That's entirely possible, English is only my second language :) I
 > mean something that's not big enough to be called an invention, but
 > more of an improvement to something that already exists.

I don't know of any definition of invention that puts a lower bound on
the "size".  For example, proposing that aspirin has some value in
preventing heart disease is an invention of sorts, and it's pretty
trivial: some doctor just noticed a correlation in statistical terms,
and published a paper in a medium-prestige journal.  The process of
innovation involves getting people at risk to take the right dose,
which involves the following costs: (1) tests to confirm effectiveness
and the appropriate dose, (2) costs of registration and regulatory
approval, (3) costs of educating doctors in the field and their
patients about the treatment, and (4) other costs of marketing.  Not
cheap.

 > I'm not sure what distinction you're making here.

Whether the invention is accessible to users who would value it.  The
process of making the invention accessible is innovation.  Innovation
can lag invention by a very long time, for example the use of
gunpowder for anything but pretty fireworks lagged the invention by a
thousand years or so.  (A good thing, in this case, since the first
major innovation was making it available to the army.)

My point about academics and the free software community is that
they're OK at innovating within their respective communities, but
these are a relatively small fraction of *all* users who would
benefit, while they're pretty terrible at reaching anybody outside
their communities.  Consider Open Office and Mozilla, both of which
started out as commercial offerings sort of for sale, and were
eventually open-sourced for a very commercial reason (namely,
kneecapping Microsoft).  Surely those generate a very large fraction
of the direct benefits to end-users.  You could talk about indirect
benefits from Linux and the GNU System, but today it's very expensive
to be a regular contributor to the kernel or other core components;
practically all of the leading "lieutenants" are employed to work on
the kernel, as are the main contributors to GNU libc and GCC.  It's
extremely difficult to keep up if it's not your day job.  Most of
those day jobs are provided by commercial firms.

The fact that many of the projects mentioned are GPL shows that the
GPL doesn't completely discourage contribution.  On the other hand,
just as the argument against Microsoft goes, "surely the most widely
used OS today would be an order of magnitude better if it weren't for
Microsoft's market power", you can argue that "surely the most widely
used free OS today would be an order of magnitude better if the BSDs,
Linux, and GNU developers were able to share freely among themselves."

And that is my argument for retiring the GPL.

 > Not at all. Again, I'm not saying that the GPL should be used for
 > everything.

Now you tell me.  What you said before was that the GPL should be used
for academic research results.  You didn't make any exceptions....

 > In these cases it makes IMHO perfect sense to release the software
 > under a more premissible license. The situation is different when
 > you're working on software which is highly specialised and the
 > algorithms used in it are the actual research.

ISTM that in that case, unless you patent some aspect of the
implementation of the algorithm, the GPL is pretty worthless.  The GPL
doesn't cover algorithms, only the expression of them in code.  If the
algorithm is sufficiently good, then "clean-room" reimplementation is
cheap compared to the value of using the algorithm.  Cf. RSA, whose
reference implementation was licensed permissively (ie, not copyleft;
obviously it wasn't a free license), but use of the algorithm was not.

 > > But another, very important purpose of publishing is to
 > > communicate the idea to those who will actually produce an
 > > innovation available to society.  You are entirely neglecting
 > > that fact, which is enshrined in the U.S. Constitution, no less.
 > 
 > That's not really different from publishing to enable other
 > academics to base their research on it. Somebody else works with
 > your ideas. And keeping the things you've found out to yourself
 > doesn't really help anyone.

I don't understand what you're talking about here.  There's a very big
difference between using a license that discourages commercial
research into extensions and thus reserves a small portion of possible
benefits to the academic community, and using a license that enables
commercial exploitation, which provides benefits to a much wider
audience (otherwise it wouldn't be worth trying to compete with the
free version), without taking *anything* away from the academic
community except vaporware that won't ever` be written in the presence
of the GPL.

That's an extreme case, of course, but my feeling is that most of the
work that commercial firms do, like spiffing up the menus and default
styles for generated graphs, are not terribly valued by academics.
OTOH, those breakthough contributions such as the new algorithms in
Mathematica are generally sufficiently difficult to invent and
implement that lack of commercial funding would at least dramatically
delay both invention and availability to those outside of the
mathematical research community (and large investment banks).

 > > Where do you think that "public" money comes from?  Not taxes
 > > paid by professors and graduate students!
 > 
 > They're paid by the general population (which includes professors
 > and graduate students) and companies. Assuming that a particular
 > piece of research will only benefit one or a small number of
 > companies,

That assumption is *spectacularly* wrong.  Remember, those companies
are *businesses*; their revenues (and therefore profits) depend on
offering a value proposition to their customers.  Even under the most
extreme assumptions about demand and supply conditions that are
plausible, the consumer surplus (total value to customer less
expenditure on purchase) is half of producer surplus (revenue from
sales, assuming *zero* variable cost of production).  Given that this
is the extreme assumption, a 2:1 ratio isn't bad.  Now compute actual
profit by subtracting costs of production and distribution (small but
not zero), costs of marketing (possibly large), costs of R&D (probably
large), and general management overhead (accounting, executive
compensation, risk premium to stockholders, etc, probably large), and
you'll get a much smaller number.

I wouldn't be surprised if the consumer surplus produced by Microsoft
over the decades runs to several trillions of dollars, while the total
benefit to MSFT stockholders is probably a couple hundred billion.
For example, much of the consumer value generated by Open Office is
due to the fact that it copies MS Office slavishly.  None of that is
benefitting MSFT, but it's based on MSFT R&D.

This is not to say that consumer surplus would not have been double or
triple that had MSFT used an open source license, I don't know about
that.  But the reverse is also definitely arguable: that without MSFT
proprietary licensing generating revenue to support development of its
new products, the surplus generated by use of software would have been
*smaller*.

N.B. "Consumer surplus" is the technical term used by economists, but
it would be more accurate to say "buyer surplus" or "customer
surplus", because it includes expense reduction or productivity
improvement by firms that use the product to make their own business
more profitable as well as consumers in the home.

 > it seems unfair that everybody else (including competitors of those
 > companies) should essentially subsidise their R&D.

Huh?  What subsidy?  The whole point of open source software is that
it is socially stupid to do the same R&D twice when you can do it once
and distribute it more widely.  How is this different from the
"subsidy" you get when I publish my research results in a journal you
can check out from the University library for free?  Maybe I should
make you pay me?[1]

My point (and you agreed) is that researchers do basic research
because that's what they do.  Pay them a decent but not terribly high
salary (as salaries for people with similar talents go), and they'll
keep doing it regardless of the license; just try to stop them from
doing it (well, holding more faculty meetings and basing promotion on
the ability to attract research grants is a pretty effective way to
shut down basic research, but that's a different rant).  They're not
into dotting i's and crossing t's, and if you leave those aspects up
to them (at their present salaries and job descriptions), they just
won't get done.

Companies on the other hand are not terribly good at doing basic
research (AT&T Bell Labs notwithstanding, RIP).  Since the academics
are going to do it anyway (BTW, this is a fundamental assumption of
Stallman's GNU Manifesto, except s/academic/programmer/), nothing is
lost by using a permissive license (except maybe royalties to the
academics, but we'll ignore that because it's just personal greed).

 > > But why not?  That's an awfully hypocritical position, considering
 > > that is what the vast majority of academics do.[3]
 > 
 > I think academics are taking stuff to base their research on and,
 > eventually, publish.

I'm not talking about "stuff", I'm talking about the taxpayers'
*money*.  Both for "research" costs and salaries.

Note that the commercial software firm doesn't care much about
"stuff", it cares about money.  I'm reducing profs and grad students
to that level, that's all.  And the amount of "stuff" that gets
published in useful form is actually rather small.  Instead, companies
like O'Reilly and Wrox spring up to pay competent writers to translate
the incomprehensible gibberish that academics fill their journals with
into technology that can actually be used by practitioners, and
companies like Elsevier hire some academics to write textbooks so that
a few of the grad students can understand a small fraction of it.

 > And yes, there're of course ways around the GPL. And it might be
 > easier for corporations to contribute to projects with a more
 > permissive license. But they can get involved in the research as
 > well if they don't want to directly contribute code. Given the
 > selfish nature of people in general, I think saying "You can have
 > it, but you must share it."

But that's not what the GPL says.  What the GPL says is "you can have
it, but if you give it to somebody else you have to give them the
whole thing."

The problem with this is that the costs (including loss of revenue to
competition) of giving the whole thing are potentially large and borne
by the distributor, while the benefits are diffused into society, and
the fraction that can be captured by the distributor is very small.
So it doesn't matter in the end that the cost to the distributor is
more than offset by benefit to society; the product doesn't get
innovated, and nobody ever knows what they've lost.

 > is better than "You can have it and do whatever you want with it." 
 > given that a license can be changed.

Feel free to try to get the license on FSF-owned copyrights, or the
Linux kernel, or OpenSSL, changed.  I don't think you can do it.

Footnotes: 
[1]  In practice, what academics do instead is withhold results from
publication.  I was once told by my professor that it was probably
worth $50,000 to the professor whose data I wanted to withold it from
me.  This is very common in social sciences, where data sets are
collected at government expense, and the summary statistics are
published quickly, but the data sets themselves are not published for
a decade, if ever.  On the other hand, both public and private surveys
exist where you only have to pay $500 to $5000 for a full copy of the
data set.  These sums are beyond the means of almost all grad
students, but well within reach of all but the poorest professors in
universities in OECD countries.