Subject: Re: small worlds and better than ransom
From: Thomas Lord <>
Date: Fri, 05 Oct 2007 13:41:51 -0700

Stephen J. Turnbull wrote:
>  > Either way, you aren't selling the activity of doing R&D: you're
>  > selling past work.
> This is the way it works in almost all pure-science grant programs
> (NSF, for example), though.  As RMS pointed out in the GNU Manifesto,
> people do speculative programming (and research) because they like to
> do it.  They get themselves funded by showing evidence of competence,
> usually past output (except for Young Investigator-type awards), much
> of which is to be integrated into the allegedly new content of the
> proposal, rarely on the basis of the wonderfulness of the proposal.

NSF and similar programs work well for some things and
quite poorly for others.   I'd rather not digress into an analysis
of NSF if it isn't important.

I don't know what RMS did or didn't say but I know why
I do open source R&D, and how I do it, and why I think
people should care, etc.

I do it because I'm very good at it and I'm not very good
at many things.   I do enjoy it and that probably helped me
to become good at it.   I do it semi-systematically -- a
literary comparison would be to using Burroughs-esque
cut-up techniques to generate new ideas, and then just
technical, improvisational coding skills to systematically
try attractive ideas out.   I think people should care because
I've had some successes this way in the past (projects that
create value, even if I didn't personally reap much) and
my rate of success seems to be going up (suggesting I'm
getting better at it).

People don't generally get themselves funded by showing evidence
of competence -- not competence at software R&D, anyway.
By one route, some people get funded through the relatively
dysfunctional credential-based allocation of research funds (but,
again, I don't mean to digress into NSF).   By another route, people
get funded when there is a "growing concern" -- a firm that
can suck up between .1 and 5 $M cash with a plausible ROI in
3-6 years.   The skills needed for achieving success by either
route are not the same as and are not a superset of the skills needed
to be inventive with software.   Those funding systems are
absolutely wonderful when the work but there has to be a third
way for when those systems just get in the way.

Open source software R&D is a "basis craft".  The skill is to, by
a mix of "feel" and technique, design a basis set of new ideas and
technologies such that that basis set is impressively generative. 
That is, if combinations of elements of the basis, and combinations
of those combinations have a large and relevant "span" then the
R&D has succeeded: it  has created new culture -- a new space of
activities -- degrees of freedom along new axes.

You can see this "basis and span" nature of R&D in famous examples,
like traditional lisp or early unix.   Both technologies feature
"composition" centrally -- functional composition in lisp,
composition of processes and pipes in unix.   A basis set with
just a few types of primitive components and universal ways
to compose them is likely to have the desired large span of easilly
reached applications.   The contribution of R&D, in these cases,
was to define the new basis sets.    Markets then captured the
value of the span much later, and in many scattered transactions.

A later example is the LAMP stack.   It's rules of composition
are not nearly as regular as traditional lisp's or traditional
unix's, but they are still among the simplest rules of composition
that span a large space of interesting web applications.

If there is a systematic way to invent new "basis sets" of
technology, and to reward the investors in that system with an
advantage in claiming value from the resulting "spans" -- then
that is a business model for open source R&D.  It is such
a model I seek and I'm trying for a very simple-minded approach,
rather than something overly fussy and theoretical.

Some would argue, I think, that the systematic way to
invent new basis sets is to just let people scratch itches -- the
basis sets will evolve organically -- no business model is needed.
However, nobody can name a single historical case (not even
LAMP) where that has actually occurred:  all of the big advances
came from individuals and  small teams of deliberate explorers
who purposefully set out to invent a new basis, usually in
pursuit of a definite economic reward.  Open source
itch scratching has proved effective at building out into the
spans of basis technologies.   It has not been good at creating
new bases.

>  > A researcher with a promising project can perhaps bootstrap
>  > by publishing an early version of the project, but then selling,
>  > primarily, "pre-purchases" of future versions at a customer-determined
>  > price.
> This looks to me to have all the disadvantages of VC funding, except
> from the researcher's point of view (he doesn't have to give up
> equity).

Oh, it's quite different.   You're looking at it the wrong way.

People with one set of skills are in the start-up entrepreneur
business.   Their job is to sell chunks of equity in growing
concerns to VCs for big chunks of money.

Let's dub that class of people the "start-up folks".

The start-up folks do some things for themselves and other
things they outsource.    So, by the time they have an ordinary
business office, probably they are outsourcing the emptying
of the trashbins, but they will do the network design for their
new social networking site themselves.   Right?

The start-up folks are pressured to always become more and
more efficient.   One of their biggest obstacles is coming up
with new visions and new founding technologies. 

I'm suggesting that the start-up folks partially outsource
their brainstorming work.    This is heretical in the valley
where the paradigm is that first you invent the transistor
in your garage and then you become chairman of a
multinational corporation but, mythology aside, it seems
a realistic approach to driving the VC engine, to me.

>  > Customers who pre-purchase, rather than wait for, release 0.2
>  > establish a relationship with the R&D vendor.   When they
>  > speak to the vendor they are heard not as a member of the
>  > general public, but as a valued customer whose repeat business
>  > is hoped for.
> A problem is that the valued customer*s* may have a wide variety of
> priorities.  To the extent that those conflict, some customers will
> lose.  This uncertainty is (in many cases) larger than either the
> uncertainty about what "X" is, or the uncertainty about if and when
> "X" will be delivered.

If there are enough customers with conflicting goals and some
are paying well below what the R&D firm wants to spend
addressing the low-paying customer's needs then, yes, the low-paying
customers lose.  (But, not always:  if a lot of low-paying customers
all have similar needs, they can add up to a regular customer.)

Pricing is the big question there.   A losing customer buys
pre-purchases and is disappointed in the end.   So, unless there
is some mitigating circumstance, presumably they don't want
to buy more pre-purchases.    The price for the first ones have to
be high enough to work for the R&D seller but low enough that
customers can take the risk of losing everything.

> After all the dancing is done, what's left is "if you need/want to
> maintain control of the development process but to be paid for the
> product, use a proprietary model."  If you're feeling generous, simply
> *promise* to release as open source under certain (unacceptable to a
> profiteer) circumstances, and keep your promises.  As you gain a
> reputation for keeping your promises, RMS will heap hot coals on your
> head, but both the people who want the best now and are willing to
> pay, and those who are willing to accept last year's version as long
> as it's free (in whichever sense) will come to like you and give your
> their business.  It worked for Aladdin/Ghostscript for many years;
> even ESP Ghostscript was more about the CUPS business plan than any
> real clamor in the community for a more advanced GPL Ghostscript.

That is silly, though.  There is simply no need.

There are natural advantages to being the customer of an open
source R&D firm.   When a customer expresses their point of
view to the firm, that influences the direction of future research.
When the firm reports developments to a customer, that customer
gets a leg up on competitors at understanding the new, evolving

Think of it in terms of mailing lists.   As a researcher, I could
create a public mailing list and declare that that is the main
way to talk to me.   Even if I make no such declaration, many
people will assume that is the main way to talk to me and
will even lash out if it turns out not to be true.   My other choice
is to declare simply that I prefer to speak with customers and
that without customers, I can't do much at all.

I guess I am saying that open source R&D is simply
"the business of having customers".

>  > R&D  vendors are unlikely to use cheap drinks, think steaks,
> I could be attracted by a think steak. :-)
>  > flashing lights, and noisy environments to encourage
>  > reckless purchases by their gambling customers.)
> DNA Lounge, anyone? *chortle*