Re: Major interview

[Ray Olszewski <ray@comarre.com>]

At 01:29 AM 9/6/99 -0400, Roger Dingledine wrote:
>
>Take a look at
>http://www.seul.org/pub/fsib.html
>This is a document I wrote a while ago, still in need of a bit of work,
>that tries to explain how the free software model is actually better for
>'normal people'. It mostly approaches it from the angle of businesses,
>but the arguments could be adapted to the educational medium.

I read through this, and I agree that it needs some work. It does a good job
of listing the benefits of Open Source (or free or whatever ... the fine
distinctions may matter in some contexts, but not here) development, but it
responds only superficially to the main criticism of Open Source software:
that the model offers no good way for freelance programmers and companies to
make money.

Personally, I'd like to see Open Source catch on and grow, but I'm not so
committed to the philosophy that I am blind to its limitations. So I read
advocacy documents like yours with a skeptical eye, but not with hostility
to their underlying goal.

Let me anticipate the conclusion of my later, detailed comments on your
draft, so as not to lose too soon the readers who find the later details of
my comments too tedious. To interest Edsoft developers in Open Source, we
need convincing examples of how they do Open Source development and be
profitable. If we can't make this case, we won't get them to develop. If we
can, then we will find some interest. 

How *might* Open Source be profitable in education? These are the models
that are worth exploring:

1. Lagged Open Source -- develop first as a proprietary app for a client,
with a commitment that the code will later be opened. (This is one way to
read Cygnus' approach to compiler dvelopment.) In practice, this would mean
contracting with a large education organization -- in the US, a big district
or a state -- to develop apps specific to their needs.

2. Open Bid -- post the project on one of the Open Source Web sites, where
interested individuals have a mechanism for agreeing to pay for development
of a particular product. (Do these Web sites really work yet? I don't know,
but the technical problems I see with this method are hardly insurmountable,
if  end-user are actually willing to support Open Source development that is
not "free" in the pricing sense.)

3. Grants -- doesn't really need explaining.

This list by itself isn't enough to convince anyone, and in any case there
may be other approaches that we should consider. But I hope it is a starting
point for those interested in encouraging Edsoft developers to pursue Open
Source, a pointer to the kinds os specific arguments needed to convince them.

The rest of this message is comments on specific areas of your draft, areas
where I think you need to tighten up the argument, usually by providing
examples (or better examples). I though about sending it privately, but
decided to go this way, since others may be able to suggest cogent examples.
And with the interview coming up, Doug, Jose, and Roman, at least, need to
remain part of this discussion.

Early in the draft, you discuss in general terms how Open Source development
can benefit a business. You write:

"Paradoxically, free software simply makes good business sense. By opening
the source to its programs, a company can achieve much faster development
cycles: often there are other companies out there interested in developing a
similar product, and often many of the customers themselves are able to
contribute suggestions and patches (small changes to the source code that
correct the behavior of the program). Since some of the customers are
technically competent, and others are often large corporations with IT
departments of their own, this assistance translates directly into more
development. Indeed, this system allows the company to keep the overhead
lower, since each software project requires fewer programmers on the
payroll, and more software projects can be undertaken by a given number of
programmers."

But you follow this general discussion with this list of examples: TCP/IP,
sendmail, perl, and apache. None of these was developed by a company that
tried to make its living by writing and selling software -- TCP/IP itself
comes from a variety of sources, from a time when the Internet was very
non-commercial; sendmail only went commercial last year (and I don't think
its commercial product is Open Source); and perl and apache are developed on
a non-profit basis by volunteer groups. The same is true of the other "big
name" success stories you don't mention:, Samba, XFree86, the Linux kernel
itself, and the large collection of GNU software ... as well as packages
like pine and majordomo. The best (only, really) example for your point that
I can think of is the compiler tools developed by Cygnus (egcs, the
native-code Java compiler, etc.).

Later, you write:

"With the customer as a co-developer, and often with the collaboration of
other companies interested in the same product, relatively small companies
can create enormous software projects that rival large monopoly
organizations (for instance, the Maxwell word processor started out this way)."

While Maxwell may be an excellent piece of coding (I don't use it so don't
know), it is hardly a poster child for the success of Open Source in a
commercial setting. The introductory discussion on the Maxwell link you
point to reads: 

"Maxwell was written by Andrew Haisley, Dave Miller, and Tom Newton. It was
originally destined to be commercial software, but for various reasons this is
unlikely to happen. We have decided to release the source code to Maxwell under
the GNU GPL, to ensure continued development on it." In other words, it was
a commercial **failure**, not a commercial success.

You later mention Mozilla, but you yourself note, correctly, that it doesn't
look like a good example any longer "given recent resignations". It too is a
*commercial* failure.

The only place I found in the draft where you claim *commercial* success
(that is, profitability) for Open Source is here:

"First of all, there are several dozen large companies making a living off
free software right now, such as RedHat, Caldera, Cygnus, and Sendmail. They
make money by packaging and selling a convenient distribution of their
software, and by providing after-sale service to their customers. While the
software is available for free from their websites, it is simply more
convenient to receive a CD in the mail along with printed manuals, and
people are willing to pay for this convenience. Many hardware companies
distribute their software drivers as free software -- they make money off
selling their hardware, and they need robust software drivers available on a
wide variety of platforms. In addition, companies such as O'Reilly
Associates and VA Research make money by selling accessories to the
software, such as books, documentation, and integrated computer hardware
systems that run free operating systems. Indeed, Netscape's decision to open
the source code for its popular web browsing client shows the influence the
idea of open-source has had on business."

Let's look at all of this more carefully.

"there are several dozen large companies making a living off free software
right now, such as RedHat, Caldera, Cygnus, and Sendmail"

Are there? First, is any of these companies profitable? (Only RH is public,
I believe; what did its IPO documents say about RH profits?) Second, do they
really "make a living" from *free* software? Except for RH, all these
companies sell both free and closed-source products. Out of the bunch,
Cygnus' mixed model looks like the most promising for Open Source, and I
don't know enough about the company to know how well it actually works in a
business setting.

In this context, it might be interesting to look at what effect the move
from closed to Open Source has had on companies that were pressured by the
community to make the move. An example I can think of off the top of my head
is the QT library -- is its developer "making a living off free software"? I
don't know. What other recent examples are there of companies that made the
move (besides  Netscape/Mozilla, hardly a success story)?

"While the software is available for free from their websites, it is simply
more convenient to receive a CD in the mail along with printed manuals, and
people are willing to pay for this convenience."

While convenience may make companies like Cheapbytes profitable, I doubt
that convenience drives sales of US$70-plus "Official Red Hat". (BTW, I
wonder if the one distribution creator that is making money might be Walnut
Crek CD-ROM, publisher of Slackware. Their business model is under attack
from lower-priced competitors like Cheapbytes, but it does at least focus on
making momey from the convenience of packaging and distribution.) I suspect
the support option with Red Hat, and the proprietary extensions Caldera and
Sendmail include in their products (does Cygnus sell CDs?) motivate most sales.

"Many hardware companies distribute their software drivers as free software"

Examples, please? I see a lot of driver software *binaries* distributed at
no charge over the Web, but I can't think of examples of hardware companies
that distribute their driver *source* code. (Of course, there may be many
examples that I don't know - but you should provide some.) Indeed, some big
video card makers even require the needed X servers to be closed source,
either by releasing their own closed-source servers or allowing someone
(SuSE, for example - XFree86 itself won't distribute closed-source servers)
to develop a server only under an NDA.

"companies such as O'Reilly Associates and VA Research make money by selling
accessories to the software, such as books, documentation, and integrated
computer hardware systems that run free operating systems."

This is the best example you have (though I find it a bit odd to call a
US$2000 computer an "accessory" to the Linux OS). Does O'Reilly support any
Open Source development, or does it only contract for books on existing
products? 

VA Linux Systems (the name changed) does pay people to do Open Source
development; do you know if any other Linux hardware vendors do? Does Dell,
for instance (you mention later that Dell "recently announced that it has
begun shipping computers with RedHat Linux pre-installed and
pre-configured") -- or is Dell taking a free ride on the work of, among
others, its competitor VA Linux? As Linux grows, this "free rider" issue may
prove a problem even for committed Linux hardware developers like VA Linux.

The problems with your last example, Netscape, don't need additional mention
here.

If we want to convince companies to do Open Source development for schools,
I think we need to identify real, concrete ways they can make money by doing
the development. I suggested the best ideas I can think of earlier in this
message; I hope others can suggest some better ones, because right now I
don't think my suggestions are good enough to be convincing.

One closing thought ... I think Open Source advocacy would benefit from a
dose of humility. Open Source has some clear advantages, but it also has
some problems. People are trying to develop solutions to the problems, and
they may succeed -- certainly a lot of clever people are trying a lot of
different things. Just as Linux advocates these days usually acknowledge
that Linux has both strengths and weaknesses, Open Source advocates should
do the same for the various flavors of the Open Source model -- it would
enhance their credibility outside the circle of true believers.

------------------------------------"Never tell me the odds!"---
Ray Olszewski                                        -- Han Solo
Palo Alto, CA           	 	         ray@comarre.com        
----------------------------------------------------------------