Subject: Re: Cyclic, Cygnus, and Sendmail: Open Source business models
From: Michael Tiemann <>
Date: Thu, 24 Jun 1999 10:08:59 -0700

I don't think this analysis holds water.  Sleepycat, Cygnus, and Cyclic
all have (or contributed to) technology that is very stable and does not
need a support contract (Berkeley db 1.85, GCC et al, CVS), yet all
three offer "support" as one of the components of the commercial
product.  Why this talk of support?  I think it's because you need to
offer support if you're going to offer people cutting-edge, new stuff. 
If you don't rev the technology fast enough, the value of support is

Sleepycat rev'd the database to 2.0, supporting multithreaded
programming and fixing god knows how many problems with 1.85.

Cygnus revs the GNU software on a daily basis in EGCS, and rolls up so
many changes every 6 months that our release notes are published in a
bound book.

Before we re-start the (bogus) conversation about whether or not vendors
of open source software depend on releasing broken software to sustain
their business, let me just say that as long as the innovations outpace
the bugs, people are going to stay interested, and both innovations and
bugs require some level of support.  (I've dealt with many engineers who
think they are smart enough to use these tools without support, but I'm
a compiler wizard, and I ask questions of our engineers ALL THE TIME.)

So, my hypothesis is that Cyclic was unable to get CVS innovation up the
critical level to make it commercially interesting, and Jim has
acknowledged that he's not going to try any longer.  I'll note that
Bitkeeper (rhetorical question: who really needs _another_ source code
control system?) continues its relentless evolution, and they are
certainly doing a good job of stoking commercial interest.  I don't
think it's their license, but their rate of innovations that are making
the difference.

Which brings me back to Cygnus: as long as we keep innovating, with CPU
ports, with optimizations, with GUIs, with 3rd party innovation, people
will remain interested.  And with support, people will be able to use
this new technology sooner than they would if (1) they waited for it to
become stable, or (2) they try to muddle through it by themselves.


"Michael A. Olson" wrote:
> I saw John Gilmore's post to the FSB mailing list of Jim Kingdon's
> announcement, in which Cyclic discontinued its new support offerings
> for CVS.  This prompted some musings on the business models in use
> by different Open Source companies.
> Cyclic has not been, primarily, a product development company.  Jim
> has financed his activities by selling support contracts, with
> occasional consulting.  I don't know what his balance sheet says,
> but I expect that his recent announcement means that Cyclic hasn't
> been making the kind of money he would need to continue operating
> the business.
> I suspect that there are two reasons for this.
> First, CVS does pretty much everything that a software developer
> needs a source code repository to do.  There's just not much of a
> market for consulting to add features to the system.
> Second, CVS is sufficiently reliable that there isn't much demand
> for support.  Companies may feel that they can support it internally,
> since they have the source, or they may be willing to walk the wire
> without a net because the code has demonstrated its stability.
> The result is that Cyclic's business model can't sustain the company.
> Certainly there wasn't enough money to fund growth -- Jim was the
> only full-time employee -- but there wasn't even enough income to
> keep the principal interested.
> Compare this to Cygnus, which has the identical strategy (support
> and consulting on Open Source tools).  Cygnus makes a good living from
> its new port consulting practice, because silicon vendors need the
> best tools to run on their platforms, and the best tools include gcc
> and friends.  There's a continuing market for these services; the
> silicon vendors are constantly releasing new chips to compete, and they
> don't have the expertise or funding to port Cygnus' offerings themselves.
> I don't know what fraction of Cygnus' revenue is from support, as
> opposed to consulting, but I suspect that consulting revenues dominate.
> That is, Cygnus is, in a sense, selling product, not services.  Companies
> pay for the software that runs on their chip sets.
> I've worked for lots of companies, and have used the Gnu tools at
> all of them.  As far as I know, none of my employers ever had a Cygnus
> support contract.
> At the other end of the spectrum is a company like Sendmail.  Sendmail
> does sell support and services, but it's fundamentally a product
> company.  They want you to buy Sendmail Pro from them, because it's
> ready to install and easier to administer than the source you download
> from  Sendmail, Inc. is committed to the continued
> availability of the MTA from, and pays its employees to
> work on the Open Source code, but their business model says that people
> will come to them for proprietary software around the Open Source
> offering.
> The moral of the story, from my point of view, is that even Open
> Source companies need to sell software to be successful.  If you're
> considering starting a business to support an Open Source project,
> you need to think about your goals.  You may make enough money to
> feed the cat and pay the mortgage, but will you have a business that
> is viable for the long term?  Why will people pay you enough money
> to grow?
> All the Open Source companies I can think of that are making appreciable
> money right now are selling software.  Red Hat does.  Cygnus does, if
> you buy my argument that their ports are really software sales.  Sendmail
> does.  Certainly our business model at Sleepycat counts on revenue from
> Berkeley DB, and not just services.
> Working with freely redistributable source code isn't, by itself,
> a business plan.  I don't think that Open Source changes the rules
> for companies much.  You still need to plan revenues and expenses.
> Open Source is a strategy you can adopt for competitive or other
> reasons, but it's not the answer to any business questions, by itself.
>                                         mike