Subject: Re: lifestyle businesses
From: Brian Behlendorf <>
Date: Thu, 19 Sep 2002 21:54:23 -0700 (PDT)

On Thu, 19 Sep 2002, Kragen Sitaker wrote:
> I'm sure running a T&M business takes a lot more work than it would
> appear; Brian, did it take a lot more work than other businesses
> you've run?

Hard to tell, every job I've had I put in serious hours on.  :)  T&M, at
last as a "small" company, was pretty instable - you hit the nail on the
head when you said you don't get rewarded for efficiency, so you have to
be at least as efficient as your competitors, but no more so, and the
customer has to believe that you're by far the most efficient.  There was
no reliable, dependable revenue stream; every sale required starting over
from scratch, it seemed like, and there was nothing in particular (other
than the quality of your work, which anyone in this fields knows is a
highly subjective thing to a customer) that hooked repeat business.  T&M
businesses also rarely can justify R&D expenses or taking chances; I think
that's ultimately what dissatisfied me the most.

A T&M business, or even a lifestyle business, would not have provided Tom
Lord the R&D retirement he wanted.

> I think software freedom drastically reduces the transaction costs of
> cross-company cooperation, but doesn't reduce the coordination costs
> within companies much.  Coase's "Nature of the Firm" therefore
> predicts free-software firms will employ fewer people each than
> proprietary-software firms.
> For example, Cyclic had two employees, but CVS had perhaps half a
> dozen other developers; this kind of small-scale cross-company
> collaboration simply doesn't happen with proprietary software
> projects.

Sure it does; not often, but it does, and that's part of the business
we're in, making that happen more efficiently.  It's not as accidental or
as efficient as it might be in the open source world, and a public open
source project definitely has the lowest transaction costs of any kind of
collaborative project, but it's not the only kind.  But it happens to
companies large and small.

> Consequently, free-software developers can leverage the advantages of
> several collaborators without paying the efficiency costs imposed by
> working in a larger company under a single manager.

I think the efficiency costs are roughly the same if you have a means for
engineers who report to different managers/divisions/etc to discover and
communicate with each other, examine each others' code and outstanding
defects, etc.  What you may miss is the self-selective nature of open
source, in that usually it's motivated volunteers working on code rather
than time-clocking engineers whose true passions are fishing or raising
their kids.  But I think that impacts companies large and small equally.

> Perhaps, then, very small FSBs can ultimately employ many more
> free-software hackers than medium or large FSBs.

Maybe.  A counter argument might be that larger companies can more easily
acommodate the overhead that funding someone to work "on interesting
things that might lead to more revenue someday" can.  How much work did
Cygnus do that was not under a T&M contract may be interesting to ask.

> Brian Behlendorf wrote:
> > FSB, Class A: companies (or divisions of companies) whose services are
> > entirely or mostly focused on (perhaps privately branded) open source
> > software; this includes support, custom engineering, consulting, etc.
> >
> > . . .
> >
> > The hierarchy implied is intentional - I'd have a lot of respect for
> > someone who could build a large Class A company.  Perhaps there's some way
> > for a Class A company to not be a "lifestyle business" as Larry describes;
> > that would be interesting.  I don't know, though.
> Do you consider Cygnus, pre-V.C., a "lifestyle business"?  Or not a
> member of Class A?  Why?  They seem to me like they should belong.

I would consider them class A, and while larger than a mom-and-pop
business, I think they pretty much maxxed out their immediate market, and
would have had to consider a new area of business in order to have, say,
doubled; and it's not clear to me that any other technology space had the
mix of qualities that the compiler/toolchain space had that made contract
development there so interesting.  I really struggle with figuring out how
generalizable the Cygnus example actually is.

> On dual-licensing: even RMS has no objection to the kind of
> dual-licensing TrollTech and Aladdin/Artifex do; even though they sell
> proprietary licenses to their free software, they write no proprietary
> software.  I vote we honor them as FSBs.

So if I say I'll release the source code to today's SourceCast under an
open source license in 10 years, can we be called one too?  :)  Seems a
bit odd.