Subject: Re: Successful FSBs
From: "Stephen J. Turnbull" <>
Date: Wed, 25 Sep 2002 16:59:29 +0900

>>>>> "Tim" == Tim O'Reilly <> writes:

    Tim> Well, I wouldn't call us an FSB.  But I'm also not sure that
    Tim> making up a rigid definition is all that useful.

I agree, not to FSB operators, not directly.  It is definitely useful
to economists.  I think economic thinking is useful to FSBs as long as
you remember that economic models are benchmarks---not reality, they
primarily emphasize the rapidly decreasing benefits to following the
herd, and they (perhaps) expose assumptions which permit substantial
profit when you find a market where they're _not_ true.

    Tim> I've always liked to point out that humans like to think in
    Tim> terms of boundaries, but that natural events tend more to
    Tim> gradients surrounding an increasingly dense core.

How to be a perfect businessman: become perfectly One with The Market,
and then do what comes naturally.  No argument from me; that's
precisely analogous to what I say when trying to teach graduate
students to stop being _students_ and start being _economists_.

    Tim> I suppose, in practice, you can define a pragmatic edge, but
    Tim> it's always changing, just like the shoreline between land
    Tim> and sea.

Given that almost all humans most of the time are wedded to thinking
in terms of boundaries, if we want to give pragmatic advice, I think
we have to have pragmatic boundaries.

    Tim> I challenge you to distinguish [ISPs] from even the early Red
    Tim> Hat except on grounds of ideology, and the fact that they
    Tim> didn't actually distribute software in the 1980's paradigm,
    Tim> but instead delivered it in the emerging paradigm of the 21st
    Tim> century.

Very simple: unlike Red Hat, their business operations don't care
whether any given piece of software is, was, or will be open or
proprietary, except as directly reflected in the bottom line.  If you
can substitute proprietary software for the free software in your
operation with no change except that you give up your free beer, in
what sense are you promoting _free_ software, or even "access to the
laboratory"?  (Well said, Rich!  Nice phrase!)

This "works" because they only have to deal with a small number of
vendors.  If BIND, the OS, and Apache all go proprietary one day
(including past versions for the sake of argument), the ISPs shut down
for a day, go to Fry's Electronics, buy one copy of each product, and
they're back up.  Raise fees to cover the one-time cost, otherwise no
change in day-to-day operations.  Sure, it's more complex than I put
it here, but (to the best of my limited knowledge) it really doesn't
much affect their day-to-day operations.  (It doesn't count if their
sysadmins moonlight on Linux or Apache dev teams, only if that's part
of the job description.)

If all free software that Red Hat deals in suddenly went proprietary,
the first thing that would happen to Red Hat is that they'd have to
hire a bunch of lawyers and purchasing agents, probably at the expense
of riffing a few now-useless developers who no longer have access to
source, and losing some of their best to now well-paying proprietary
upstream projects---a radical change in the way they do business.

You can draw a complementary distinction for Aladdin, too, though not
quite as sharp.  If they withdraw the AFPL and GPL and make Ghostscript
entirely proprietary, then they're going to want to add marketing (and
maybe support) functionality aimed at the retail market.  They
probably need to add programmers to work on driver development etc
that currently are mostly contrib, and might have to replace existing
contrib work with "clean room" implementations if the authors balked
at the proprietary scheme.  They don't _have_ to change anything, but
clearly their incentives change radically if they give up the current
open source (in spirit, if not to the letter) approach.

Note that although Usenet itself and the ISPs have tended to enable
distribution of free(-ish) software, precisely the same technologies
can and have been used to distribute proprietary software (not to
mention "warez").  The only reason that Usenet/ISPs tend to be
associated with free software is that proprietary vendors prefer to,
and because of restricted distribution can afford to, control the
distribution channels themselves.

So the distinction can be made.  I don't know if it is a useful one.

    Tim> If you want a definition, I'd say something very broad, like
    Tim> "An FSB is any business that uses free software as a
    Tim> significant part of its business strategy, and that profits,
    Tim> directly or indirectly, from the wider use of free software."

    Tim> And yes, that includes O'Reilly, and IBM, and Sun, and
    Tim> ActiveState, and CollabNet, as well as Red Hat.

Not to mention Microsoft and RSA, both of which depend on the Internet
and its largely open source infrastructure for important components of
their strategies.  Oops!  That's too close for comfort.

I think Rich's felicitous phrase "access to the laboratory" is useful
here.  This should be enough to exclude Microsoft's current strategies:

An FSB is a business that _promotes_ "access to the laboratory" via
free software as a significant part of its business strategy, and that
profits, directly or indirectly, from the _substitution_ of free
software for proprietary software.

This doesn't have to be organization-wide.  Also, I would like to
advocate _promotes_ and _substitutes_ as key parts of the definition.
What do you think?

Institute of Policy and Planning Sciences
University of Tsukuba                    Tennodai 1-1-1 Tsukuba 305-8573 JAPAN
 My nostalgia for Icon makes me forget about any of the bad things.  I don't
have much nostalgia for Perl, so its faults I remember.  Scott Gilbert