Subject: Re: Studies
From: Russell Nelson <>
Date: 27 Nov 1997 20:00:36 -0000

[ by the way, this list is archived on the list server.  Send mail to  -russ ]

L. Peter Deutsch writes:
 > I'm sure they don't.  So IMNSHO, Adam's lumping AFPL'ed software in with
 > commercial software simply because it doesn't allow Yggdrasil and others to
 > make money on a free ride off the authors' efforts is quite unfair.

Adam wants to distribute solely libre software.  He goes out of his
way to remove non-libre software from other vendors distributions when
he redistributes them on a CD-ROM collection.  As a consequence, he is
free to distribute FPL'ed software, but as a further consequence, he
chooses not to.  Seems like there was an old Greek guy in a similar
circumstance, but I couldn't tell you his name.

It's my impression that Debian as the same policy.

 > 	1. Linux vendors cannot expect to make money on the
 > distribution function per se.
 > 	2. Linux vendors can make money by selling non-libre products that
 > enhance libre products.  But they could do that whether or not they
 > distributed the underlying libre products.  And to the extent that their
 > products are non-libre, they forfeit the claim to be participating in a
 > particularly community-benefiting activity.

"To the extent".  No Linux vendors does this to a great extent, not
even Caldera, the most commercial distributor.  The "Redhat"
distribution per se does not include any non-libre software.  It is
only the "Official Redhat" distribution which includes some.

 > 	3. Linux vendors can make money by providing services.  But they can
 > do this regardless of whether the topic of the services is gratis, libre, or
 > whatever (although their own costs will be lowest if the topic is at least
 > gratis).

You forgot #4 and #5, Peter:

4. Linux vendors can make money on their brand name.  "The Official
Redhat distribution?  Oh, Redhat makes a good product, I'll buy it."

5. No market is perfect -- and changing markets are *never* perfect.
There's almost always a little slop, a little room to tack on a little
extra on each copy sold.

 > In my '96 paper, I argued that polished end-user applications cannot be
 > developed on the libre model, because the cost is so high (much higher than
 > the cost of developing system-like software) and the rewards to the
 > developers are negligible.

It depends on what you mean by rewards.  The direct rewards of writing
libre software ARE often slim.  But there are also indirect rewards,
like reputation.  Look at Linus Torvalds.  He could walk into any Unix
house and say "Hello.  My name is Linus Torvalds.  You ki...[ sorry,
wrong script ] Where is my desk?"  Now imagine the value of the
reputation of someone who wrote a really great free stock analysis
software, or a really great free word processor.  Look at David
Harris.  He's a big mucky-muck in the email world with his Pegasus
Mail software, and he is *antithical* to the concept of marketing
one's software, to the point where he refuses to cooperate with
national magazines that wish to review his software.

 > Indeed, it is my understanding that the one great counter-example
 > -- a Photoshop work-alike whose name momentarily escapes me -- has
 > now been abandoned precisely because the developers could no longer
 > afford to continue subsidizing its development with their time. is still up, and has news dated November 13.

 > To my mind, the most interesting model of libre software development -- and
 > the only one that will be able to take the next step, to competing in the
 > producing of end-user applications -- is one in which no single developer
 > has to pay this subsidy: i.e., a model in which there are many developers
 > each doing a small amount of work.  I think the key issue in this kind of
 > dvelopment is how to maintain architectural and stylistic coherence of a
 > large program being worked on by many people, without which software rapidly
 > decays into unmaintainability.

Yup.  It's a problem that business struggles with all the time.  In
all my reading, I have not found anyone who has an actual solution to
the problem, although there are many poseurs.

 > It is my understanding from Linus that Linux is closer to the
 > Ghostscript model: he is personally responsible for maintaining
 > coherence.

That is my understanding also.

 > The truly cooperative model of software development is a subversive one for
 > the "free software business",

Hehe.  If it happened that free software business opportunities dried
up, then this list would mutate into a free software management list,
because I'll bet that's second on list members' interest lists.

 > If I had known about the Apache model when I started writing Ghostscript in
 > 1986, I probably still would have done what I did, but I would have given a
 > lot more thought to architecting the program specifically to support
 > distributed development.

Hmmm...  An ounce of design saves a dollar of maintaining coherence.

Makes me wonder how much you could break a complex program down into
independently cooperating parts.  If you had good API maintenance,
that might be sufficient.  That might be the way, for example, to make
a libre desktop publishing program.  Have one person just write and
maintain APIs, never code.

We know how a large complex economy works best -- through free
(there's that word again -- I mean libre) markets.  Perhaps a similar
mechanism could be used to write libre software?  It needs more

-russ <>  | Freedom is the
Crynwr Software supports freed software | PGPok |   primary cause of peace.
521 Pleasant Valley Rd. | +1 315 268 1925 voice | Taxes feed the naked
Potsdam, NY 13676-3213  | +1 315 268 9201 FAX   |   and clothe the hungry.