Subject: Re: Paying for development on a services model?
From: Ben_Tilly@trepp.com
Date: Tue, 9 Oct 2001 20:53:48 -0400


Zimran Ahmed wrote:
> Ben Tilly wrote:
> > Cygnus Software managed to develop GPLed software,
> > for profit, and did
>
> while a handful of companies developed GPLed software
> and made some profit, they're small and vastly
> outnumbered by proprietary firms. Burden of proof
> still lies with OSS developers who try to make a
> profit. I am not familiar enough with the intimate
> details of Cygnus to comment further.

Excuse me?  It really seems that you are not bothering to bother
reading what I wrote.  You seem to be presenting random opinions
with no context for what has been said and no consistent understanding
of what position is outlined.

You claim that it is IMPOSSIBLE to make a profit developing open
source software.  This in reply to a post which gives specific
examples of companies that did so, says it is hard, and explains both
the factors that make it hard, and suggests what you need to do to
handle those factors.

Demonstrating that it HAS BEEN DONE is sufficient to prove that it is
possible.  I am not saying it is always feasible.  I am not saying
that it is easy.  In fact I said quite a bit about why it is hard.
But the assertion of impossibility is a very strong one, and it is a
wild claim that you are making with no support in the face of existing
counter-examples.

The burden of proving that what has happened is impossibile is on YOUR
shoulders here, not mine.  The fact that you are not familiar with the
details of how Cygnus made a profit does not change the fact that it
accomplished what you considered impossible.  It sold a service which
required it to do ongoing development of a GPLed project.  It made a
profit doing this for many years.

> > Suppose that EDS and Microsoft were interested in
>
> forget EDS and Microsoft for a moment and look at
> business as a whole.
>
> All sorts of businesses start every year. They have
> every type of margin. Grocery stores start. Software
> firms start. Service firms start. Software firms
> start. They all get financing somehow on a variety of
> different terms. To claim that a company MUST have
> enormous margins to get funding is simply at odds with
> the world.

What do you think the subject of conversation is?

It is not whether service businesses can thrive.  It is not whether
the open source world will support healthy service businesses.  It is
about what conditions are needed to run an open source business that
makes a profit and contributes back development to the open source
world.

[...]
> > investment.  Microsoft has much less of a problem
> > getting together the
> > resources to enter a new market.  Indeed that level
>
> think about this from an investment standpoint. if a
> market has 10% margins in it, then why would any
> company want to invest capital that cost more than 10%
> a year? If Microsoft could invest their capital in
> something that returns 15% a year, they'd be foolish
> to invest it in a 10% business *even if they could*.
> EDS similarly does not need to tap into its cash pile
> to fund entries into new businesses. Any potential
> investment source happy with the risk/return profile
> is OK. Remember, there are debt markets, equity
> markets, bank markets, subsidies, government sources
> etc. etc. etc. if there's a 30% margin in a market,
> ALL investment sources should be happy to lend at up
> till a 29% rate to chase that market.

Several points.

First of all you seem to be confused about what profit margin means.
The profit margin we are talking about is the amount of profit per
dollar brought in.  If you have a very small profit margin, but you
can generate a lot of business for a small initial investment, you
may have excellent return on investment.  For instance grocery stores
do this.  They have a low profit margin on selling fresh produce.  But
if they buy a dollar of apples, they quickly sell that stock and are
back buying another dollar of apples with the proceeds.  Before they
are done they have bought many dollars of apples with that initial
dollar, and generated a good profit for the year on that dollar.

Secondly in different lines of businesses, the structure of your
business supports different business models.  For instance if you
have to make large investments that pay off over a very long time if
they don't bomb, you need to have a very large profit margin to get a
good rate of return on money invested.  Lower the risk of failure,
the portion of your worth that you need to put at risk, and the length
of time to a good return, and you can get the same returns with lower
profit margins.  But the lower the profit margins you are working
with, the more it matters to your bottom line if you can find small
incremental savings.

Thirdly there are barriers to entry.  In 1998, any direct competitor
to Microsoft was committing suicide, and everyone knew it.  It makes
no difference what Microsoft's rate of return was.  Nobody with a
brain was going to invest in direct competitors because they weren't
going to be able to survive.  Instead people invested in Microsoft.

And my last point ties back to the subject of the discussion.  I am
talking about the ability of service companies to fund development,
not the ability of open source companies to be profitable.  Low profit
margins says nothing about your ability to make a good return on
initial investment.  It says a lot about your ability to bear
potentially questionable expenses.  Therefore if you are funding
those expenses, you either are finding ways to make what they are
buying core to what you are delivering (which is what Cygnus did), or
you are finding a way to raise your profit margin (which is what
proprietary software companies do).

> > margin.  The OEM does not.  And Microsoft has the
> > power to decide
> > which OEM gets the best margin, and uses that
> > ruthlessly.
>
> this is because Microsoft successfully commoditized
> away the complementary good (hardware), PC
> manufacturers were unable to commoditize away the OS.
> This is not because proprietary software is
> *inherently* higher margin than hardware, it's just
> the way this particular market played out at this
> moment in time.

Microsoft is not an isolated example.  It is merely a good one.

> Remember, the marginal cost of production (and
> therefore margin) of software is zero.

I know that very well.  I am also aware of the relationships between
profit margins, risk, barriers to entry, and length of time to return
on investment.  Proprietary software tends to involve a lot of risk,
significant barriers to entry, and a long lag between intial
development and potential reward.

> > For instance IBM is unable to compete in the
> > commodity PC business,
> > not because they are unable to run a good business,
> > but because
> > Microsoft is unwilling to give them terms where they
> > could make a profit there.
>
> And why shouldn't Microsoft want IBM to make PCs?
> After all, every PC IBM sells will have Windows loaded
> on it and therefore make MSFT.

Because IBM competes with Microsoft.  For instance Lotus competes with
Outlook/Exchange, and OS/2 competed with Windows.

> Microsoft wants to avoid being commodified away. The
> terms they impose on OEMs (eg. no desktop control, no
> dual-boot) aren't there to stop PC makes from
> generating profit (competition within the PC market
> will do that just fine), it's to avoid having their OS
> commoditized away.

Microsoft plays favorites quite deliberately to encourage OEMs to
support the Microsoft vision of Windows everywhere.

> Microsoft doesn't hate Dell because it has the best
> margins. Infact, it likes Dell because Dell is a great
> competitor in the PC market and causes PC prices to
> drop.

A large contributing factor to Dell being a great competitor is that
it has lower costs on Windows than practically anyone else.  Why?
Because they don't compete with Microsoft anywhere.  By contrast HP
and Compaq support operating systems which Microsoft would Prefer To
See Replaced with Microsoft-only solutions.  Hence they get worse
terms.

[...]
> > margin.  Conversely an open source company that
> > wants to get a better
> > profit margin needs to find or create a barrier to
> > entry.
>
> which you can't really do through keeping completely
> open code. If you want to somehow sell code, you have
> to version it. And if you want to version, it has to
> be somewhat closed. Would you count that as still
> being an FSB?

Your logic is faulty.  In fact in my first email I gave a list of
potential ways to create barriers to entering the business that an FSB
was in.  Only one of those ways was a dual licensing strategy.

> the OSS alternative is to not have any development
> cost to ammortize and so not need to version code,
> just sell services or other complementary offerings.
>
That is not the only OSS alternative.

(Incidentally you are consistently spelling "amortize" with 2 m's.)

> > Wrong.
> >
> > Any company that creates a complementary good to OSS
> > would like to see
> > OSS software be viable.  They do not necessarily
> > want to see it enough
> > to support OSS.  Indeed given a choice, they don't
>
> you're looking at this through a developers
> perspective, not a business perspective.

I am talking about how free software businesses can reasonably fund
ongoing development of free software.

> From a business perspective, if I use OSS at all I
> support it. Even if I don't add anything to the code
> base. If OSS is in my product, I would count that as
> supporting OSS. If I write applications, and I write a
> version for Linux, I would count that as supporting
> OSS, even if I don't contribute to the code. You may
> use a different definition, but if we're hoping to get
> wider use of Linux, i would count any sort of user,
> even those that don't contribute back to the code
> base.

A million companies out there may be willing to come in and set up a
small business.  Give them email, basic web, backups, etc.  Those
companies are good for the usage statistics.  But stamping out more
intranet servers is not contributing to development of more OSS.

By contrast Pixar (which is not an FSB) has contributed a lot of code
to OSS.

[...]
> > rather unhappy that their products were reverse
> > engineered, and many
> > more go so far as to develop and release binary only
> > drivers in the
> > hope of keeping that from happening.)
>
> indeed. I beleive that this is an incorrect business
> strategy driven by not understanding what OSS is. In
> the future, I expect this to be less of a problem. It
> makes no economic sense for hardware manufacturers to
> not have Linux drivers available if they don't need to
> pay to develop them.

Their concern is not whether a few Linux users buy their hardware.  Of
course they love selling hardware to Linux users (as long as they
incur no support costs for it).  Rather they are worried about what
might turn up when people reverse the hardware.  That includes
refutation of company claims about the hardware (frequent problem if
any kind of encryption is supported), public documentation that makes
it easier for people to compete with them, or revealing of other
dirty laundry which Does Not Bode Well.

Indeed these fears have proven to be founded in fact in many cases...

Cheers,
Ben