as a PDF

advertisement
ID: 112
1
A Framework Analysis of Business Models for
Open Source Software Products
with Dual Licensing.
Working Paper, CBS/INF, No. 1, January 2007
Jesper Holck, Roberto V. Zicari 1)
Copenhagen Business School
Department of Informatics
Howitzvej 60
DK-2000 Frederiksberg, Denmark.
Abstract
Aim of this research is to identify the key elements that play a significant role in the success of
business models for companies which produce
software products based on open source using
dual licensing. For that we have defined a new
framework analysis that can be used to evaluate
existing and new business models.
Index Terms—Open Source Software, Databases, Licenses,
Dual Licensing, Business Models for Open Source Software,
Software Products, Innovation, Commodity Products.
I. INTRODUCTION
Recently we have seen a growing interest in business models based on open source software, and
in this paper we will focus on a subcategory of
these: dual licensing. In dual licensing models, the
company owns the copyright to an open source
software product and receives its main income
from selling product licenses. Even though the
product is open source and thus can be freely
copied, customers may want to buy a “closed license”, as this gives them the right to embed or
include the open source software in their own
products without having to make these “derived”
products open source also.
II. ECONOMIC DYNAMICS OF SOFTWARE
McKelvey observed that “software development is
representative of a larger category of knowledgeintensive products and sectors … Unlike information, however, software can and is increasingly
often sold based on the tasks it can perform, not
the underlying code” [1].
Software shares many similarities with R&D intensive sectors where the production cost is low
compared to the cost of discovery, development
and testing. For a software producing company,
the production and distribution costs (i.e. making
CD copies, printing manuals, distribution to resel-
1) Professor, Database and Information Systems (DBIS), Johann Wolfgang Goethe University, 60486 Frankfurt, Germany, Computer Science Institute
(email: zicari@cs.uni-frankfurt.de) and Copenhagen Business School (CBS), Department of Informatics
ID: 112
2
lers) are usually very small, compared with the
cost of creating the software program.
license; we will later deal with these licenses in
more detail.
McKelvey also noted that software is sold on a
particular kind of market, where commercial
products often compete with free software, so users of a given software product are not necessarily
buyers [1].
Two other elements play important roles for open
source software:
This is also true for other “knowledge products”
on the Internet, an obvious example being the free
encyclopedia Wikipedia competing with equivalent proprietary products like the Internet edition
of Encyclopedia Brittanica.
These free and commercial products can be
viewed as competing in terms of users, as a large
number of users can be important to create momentum for further development [2, 3].
This is clearly the case for the communities of
open source developers on the Internet, where
each individual can be seen both as a user and a
producer of the software. Their role in the software development process, especially in testing
and suggestions for improvement, is fundamental.
Internet communities are also important to initiate
development of complementary software and future products, thus potentially attracting even
more future members to the communities.
Philip Armour stated it in [4]: “Software is not
treated as a medium, it is treated as a product, and
this is the problem. The product is not the software, the product is the knowledge that goes into
the software”.
III. OPEN SOURCE SOFTWARE
It is important to note that open source software is
not in the public domain and without copyright. In
contrast, the copyright holders use their copyright
to transfer certain rights to users of the software:
•
Community, the existence of a virtual group
of users and developers, actively engaged in
using and improving the software
•
Development Model, a highly iterative lifecycle where many developers in parallel
contribute with additions and changes to
existing source code
Even though the principles of open and freely
distributable source code were already widely adhered to at universities and government institutions back in the sixties, the term “open source”
was first coined in 1998. At that time some of the
more mature and well-known examples of open
source software were Linux, FreeBSD, Apache,
Perl and the many Gnu tools.
Key business/technological questions at that time
were [5]:
•
Is open source really a new way of building
software?
•
Is open source a repeatable and reliable
methodology?
•
Why would companies produce open source
software and in that way give their products
to competitors for free?
•
Is open source a threat to current methods
for building software and doing business?
From the perspective of a software vendor, the
key questions were [6] (and partially still are):
•
Is open source a reliable model for conducting software development for commercial
purposes?
•
Free Distribution: The right to make copies
of the software and distribute these.
•
•
Are open source software business models
sustainable in the long term?
Source Code: The right to access to the software’s source code.
•
•
Derivative Works: The right to change the
software.
How will open source software businesses
face the developmental and marketing difficulties associated with what Moore [7] calls
“crossing the chasm” from early adopters to
the mainstream?
The copyright holders are usually the programmers, but may also be companies. The mechanism
used for this transfer of rights is an open source
Since then things have changed, open source is
now a business strategy with broad applicability;
ID: 112
for some IT companies open source has become a
strategic vehicle to use against competitors.
A recent special report on open source by Business Week [8] forecasts that “the open-source
movement is making another big thrust forward.
Entrepreneurs, investors, and many analysts say
they’re confident that all of a company’s business
software—representing hundreds of millions in
sales—will soon be available as open source.”
IV. OPEN SOURCE BUSINESS MODELS
McKelvey [1] studied on a theoretical level the
main characteristics of a business model for open
source software such as Linux that she called
“Network-based development of knowledge, with
node control, which is not linked to firm-based
control of economic returns.”
She characterized the business model as “software
development which develops through open networks, much like public knowledge, although
there are nodes of control. Development runs parallel, or else instead of commercial proprietary
development. Here, both individuals, communities
of programmers and firms can be important for
different aspects of organization and control.”
In her model of Linux, she illustrates that software
development and economic returns are decoupled
from control of one firm, and instead rely on the
action of many, distributed developers.
She points out how Linux is an example where
one set of knowledge (i.e. the software) gains in
popularity and in number of users, thus attracting
people interested in appropriate benefits according
to other economic models (e.g. selling “distributions”, bundles of software, service, manuals etc.).
The model relies on some sort of attraction so that
individuals (i.e. developers) and organizations (i.e.
firms or projects) invest their time and energy into
developing the software.
She defines what she calls “nodes of control”; in
the case of Linux referring to the early role of
Linus Torvalds as overseer of a large, cooperative
development process. One can easily generalize
the concept of nodes of control to any firm or organization that coordinates a distributed cooperative software development process.
3
She also points out how both individuals, communities of programmers, and firms are important for
different parts of the process.
Asif Khalak defined an economic model for impact of open source software [9]. He analyzed the
effect of introducing an open-source software
product onto a proprietary market in an idealized
setting. He discovered that demand structure
strongly influences the effect of open-source.
Not surprisingly, he suggests that the open source
model works at best when the demand is for low
prices.
Krishnamurthy in a recent study classified business models for open source into four models
[10]:
•
Distributor
•
Software Producer with a non-GPL Model
•
Software Producer with a GPL Model
•
Third Party Service Provider
In his study, he calculates the potential profit of
open source Products using two dimensions:
•
Customer applicability: the proportion of
the market that can benefit from the software
•
Relative product importance: how important
a program is to the functioning of the user’s
computer.
In his research, products with highest profit potential are classified as having high relative product importance and high customer applicability.
Krishnamurty suggests that so-called open source
high-profile nichers may have potential to become
profitable [10]. These are commercial product
companies based on open source, and regarded
very highly within their specific market niche;
however, beyond that, they are not well known.
V. OPEN SOURCE DUAL LICENSING
Producers of open source software have gone
through a complex process of defining adequate
licensing schemes. The licensing scheme is a key
element directly affecting the profitability and
even the existence of any business based entirely
(or partially) on open source software. A brief review of the main licensing schemes offered for
open source is shown in exhibit I.
ID: 112
One of the most popular licensing schemes for
open source is the GNU GPL, even though it has
proved difficult to build up viable economic models based on this. An important aspect of GPL
(copyleft) is that whenever a GPL-licensed piece
of software is (partially) included in a new software product, this new product must also be under
GPL.
It is interesting to recall what back in 1999, Bill
Joy, at that time chief scientist at Sun, commented
when asked if Sun would consider GPL for licensing their products [11]: “I can’t license all of
Sun’s intellectual property under the GPL, because it just won’t work. I don’t see any reason
why I should give somebody access to stuff that
cost me millions of dollars to do. We’re spending
over a billion dollars a year in research. I can’t
just throw it all on the street… If I make code
available under the GPL, I’ll lose control of
it…The GPL just doesn’t solve my business
problem at Sun. I would like all of our intellectual
property to be available in source form, but I can’t
economically do that under the GPL.”
In the attempt to find viable business models for
open source, partial open-source licensing models
were introduced with the following characteristics:
4
base). Revenues are typically generated by selling
commercial licenses.
Mikko Välimäki has analyzed dual licensing and
observed that by offering dual licensing [12]:
•
“Copyright and control of the core product
development is held in one hand, the original developer.
•
The ability to license the product with other
terms than open source requires full ownership of all rights to the product.
•
The users of the free license have an option
to obtain a proprietary license. If a software
product with an inheriting copyleft clause is
embedded to become a part of another product then the combined product should be
distributed for free. A proprietary license
may free the user from this restriction. In
this way, third party product businesses become also possible
•
The development community does not have
development power to start competing
products (forks).”
Välimäki also identifies three elements that in his
view are necessary for successful dual licensing
[12]:
•
Open source software can be mixed with
non-free software
•
The existence of network effects, only
possible with a large community of users.
•
Modifications can be taken private
•
Price discrimination.
•
The software produced can be re-licensed by
anyone
•
No major requirements for the enforcement
of copyright.
•
Special privileges for the original copyright
holder over other developers’ modifications
According to Välimäki “GNU GPL license seems
to be a viable marketing tool in dual licensing”.
The most notable among the partial open-source
models is dual licensing. Dual licensing is based
on the idea of simultaneous use of open source
and proprietary licenses; companies using dual
licensing can be seen as using both pure-play and
hybrid business models [12].
The acceptance of partial open-source models is
neither straightforward for established proprietary
software product companies, nor for developer
communities identifying open source software
with free software.
The dual licensing scheme has generated new
business models for commercial companies.
Among commercial, established companies using
dual licensing we can mention Sleepycat (relational database), MySQL (database server), TrollTech (application development framework and
tools), and db4objects (embedded object data-
VI. A FRAMEWORK TO ANALYZE
BUSINESS MODELS WITH DUAL
LICENSING
In this section we suggest a new framework for
analyzing the potential success of open source
business models based on dual licensing, offering
both GPL and a commercial license.
ID: 112
We focus our research on commercial software
product companies using open source software
under the dual licensing scheme.
In our framework we look at the following dimensions:
• Revenue Generation
• Dual Licensing
• Products/-Technology
• Entry Market
• Synergy
• Community
• Price
• Software Development
• Spending
• Organization and Management
• Custom Support
• Sales and Distribution Channels
• Financial Support
• Market Strategy
We will in the following examine each of these in
turn.
Revenue Generation
The number one challenge for any business model
is revenue generation. In particular for companies
selling software products, revenue is mainly generated by selling licenses to end users and to companies that embed the software product into their
own products (“embedders”).
Both end users and “embedders” can evaluate the
software product by simply downloading and
testing a free copy from the Internet. “Embedders”
may also find it important to evaluate and examine the freely available source code, as this code
will be included in their own software products.
However, a high number of downloads does not
necessarily lead to a high number of sales; it has
been reported that on average only 2% of customers who have downloaded software for free will
eventually end up paying for it [8].
According to Marten Mickos from MySQL the
numbers are even more extreme, only 10% of
5
downloads cause the installed base to grow; the
other 90% are upgrades of existing installations or
just for testing and playing around. And from the
installed base only around 0.1% end up as paying
customers. MySQL does not perceive this as a
major problem though, as those who do not pay
would not pay anyhow, and in fact help MySQL
by adding to the installed base and recommending
MySQL to potential buyers [13].
That software can be downloaded and tested for
free may both help in qualifying the downloader
into a potential buyer and in indirect qualification,
where the downloader does not him/herself buy
the software, but recommends the product to a
potential buyer.
When looking at sources of revenue, it is important to break down the end user category further:
•
Enterprise end users: employees using the
software as part of their work
•
Individual end users: people using the software for individual purposes.
These two groups are intertwined and interact;
someone who is an individual end user at home
may be an influencer at work, for instance.
Typically enterprise end users contribute heavily
to the revenue generation, whereas individual end
users don’t, as they seldom will pay for the software. The individual users may, however, have
important roles in the decision-making processes
in enterprises. Having adopted products at home,
they will put a pressure on the enterprises in
which they work to adopt the same products. And,
because they add to the products’ installed base,
they also through network effects increase the
companies’ value of these products.
The combination of the viral nature of the World
Wide Web and the role of a large user and developer community can form a powerful promotion
channel at virtually no cost for the open source
vendors, further increasing the size of the community. The larger the community, the more
chances are that customer will pay. A large community is therefore an important factor for the
success of the business model.
We should note however that there are no research
results available that clearly demonstrate this assumption. On the contrary, some of the most
ID: 112
popular open source products (Apache, BIND,
Linux) seem to be the ones with fewest paying
customers (proportionally).
We have defined the following parameters to help
in discovering and quantifying the above relationships:
# downloads over time t
, the number of
# members of community
downloads over time t, with respect to the total
number of members of the community.
p1 =
# sales
, the number of sales with re# downloads
spect to the total number of product downloads
p2 =
revenue
, revenue generated with re# downloads
spect to the total number of product downloads.
p3 =
Dual Licensing
The economic incentive of redistribution makes
dual licensing an important element in a successful business model for software products that are
often embedded in other products. But dual licensing may also be relevant for end users. In this
case the commercial license can be seen as a support model and will mainly be attractive to enterprises. So the dual licensing scheme allows two
flows of revenue: from “embedders” (in case the
product is suitable for that) and from end users.
This is not a general rule, however. For MySQL
AB, dual licensing does not lend itself to sales to
end users, it only applies to embedders. For end
users MySQL has a different model – the subscription model [13].
It is important to note that most members of the
open source software developer community perform an important support function but do not
generate revenue directly.
In case of MySQL (having 50,000 product
downloads per day [14] and 10 million active installations [15]), most community members are
non-paying users who help building a large community, but most of the revenue is derived from
the 0.5% with MySQL embedded licenses [16].
This confirms the assumption that the dual license
is an attractive feature for redistribution.
6
On the other hand, the revenue generated by
MySQL Network subscription1 (for end users) is
steadily increasing [16]; this can be seen as an indicator that a large community may be a good
base for revenue generation.
Thus we can identify different categories of paying customers:
• End users, including
o Buyers: Users who buy the software product
with a non-GPL license, typically “enterprise” end users
o Subscribers: As in the case of MySQL, end
users of the GPL-licensed product who pay
for the services included in the MySQL
Network Subscription [13].
• Redistributors who buy the software with a
non-GPL license and embed it into their own
product(s).
We have defined two more parameters that can
be useful to quantify the relation between revenue and business model:
# buyers
, the percentage of
# community members
community members who buy a product license.
p4 =
revenue
, the “average customer value”, ex# buyers
pressed as the generated revenue with respect to
the number of buyers.
p5 =
An indirect source of revenue might come from
the rest of the Community members, the non-buyers. We can therefore see relationships as follows:
Size of community → # downloads →
# buyers → direct revenue
⎧
⎨
⎩# non-buyers → indirect revenue
That is, we expect that a large community will
lead to a large number of downloads, in turn increasing the numbers of both buyers (leading to
direct revenue) and non-buyers (and indirect revenue). We should however note that research has
not yet shown a direct link between downloads
1
MySQL Network subscriptions include enterprise-grade support,
automatic updates (MySQL Update Advisor) and Technical Alert Advisor,
which notifies customers about issues related to their specific computing
environments. The Network subscription also offers warranties and
indemnification for MySQL’s database [17].
ID: 112
and revenue generation. One factor to consider is
a possible large time lag (of several years) between first download and becoming a paying
customer, which makes difficult to prove the causality [13].
We suggest another parameter to distinguish between direct and indirect sales:
# direct sales
, which expresses the number
total # sales
of direct sales with respect to the total number of
sales, including indirect sales generated by actions
of non-buyer community members.
p6 =
Software Products
We now consider the software product offering.
An important distinction which may affect the
success of a business model is between:
• Commodity products
• Innovative products
In our definition, commodity products are products with functionalities and benefits that are already well understood and accepted by a large
majority of the target entry market; innovative
products on the other hand are products with less
well understood and accepted benefits and functionalities.
The type of product has an influence on the size of
the community: Commodity products have potential to appeal to a large community of users,
whereas innovative products initially tend to be
used by only a small community.
An important question here is whether an open
source company with dual licensing has a good
chance to succeed as a technology innovator or
should instead focus on a commodity product.
On this point, venture capitalist Danny Rimer expressed a strong view in a Business Week interview [18]: “Open-source companies absolutely
can’t have a new, innovative technology. They
have to be smarter approaches to existing technology. They have to be technologies that developers
and buyers already understand. Open source is
about coming up with an alternative that’s
cheaper, not going after a new area.”
Rimer also identified the following requirements
for a successful business model based on open
source [18]: “The existence of a large active
7
community. It has to be a commodity product, the
company has to produce software that is commonly understood by users. It has to have a price
cushion. The open source alternative should be
considerably cheaper than the closed software alternative and still allow for a price cushion so that
over time open source companies can charge more
but still remaining interesting to customers.”
This is the perspective of a venture capitalist. For
Rimer, the above requirements apply for companies aiming at a revenue potential of $100 million
a year.
At present, no or few commercial companies
based on open source software have proven able
to build a big business beyond supporting Linux;
MySQL with $40 million in annual revenues may
be a notable exception.
An important case that seems to contradict
Rimer’s definition of a successful business model
for open source software is the Apache web server
software. According to Netcraft, 62% of all Internet web sites run Apache today, compared with
31% running Microsoft Internet Information
Server [19].
If we look at the value proposition of Apache, we
notice that they offer a product that is:
•
cheaper than their direct competitors’
•
very reliable and matches market requirements
In our framework, the Apache web server can be
characterized as an innovative product with price
differentiation and a large developer community
(more later in this section under Price and Community).
Apache’s software is not intended for redistribution and is only offered with a “pure” open source
license; this seems to confirm the notion that dual
licensing is most useful for products for redistribution.
On licensing, the Apache Software Foundation
(ASF) declares:
“The ASF will not dual-license our software because such licenses make it impossible to determine the conditions under which we have agreed
to collaborate on a collective product, and are thus
contrary to the Apache spirit of open, collabora-
ID: 112
8
tive development among individuals, industry, and
nonprofit organizations.” [20]
•
receives help from full time staff at large enterprises
So ASF sees dual licensing as incompatible with
the way (a part of the) open source community of
developers operate. Because of the direct correlation between the developers’ community, their
active participation and support, and the success
of any business model based on open source, this
is a very critical issue that must be considered before deciding on a business model.
•
uses only Apache licensing (Apache License
Version 2.0), does not offer dual licensing
•
has a horizontal entry market (web server
software)
•
has major, direct competitors (proprietary
software companies) in their market
We asked Christof Wittig (CEO of db4objects, a
commercial open-source software company using
dual licensing), what is, in his experience the reaction of the open source developers’ community
on dual licensing. He replied: “It’s always a question about the upfront contract. We were always
commercial/dual licensed and are open (and consistent) about our economic incentives, and people
respect that. Don’t forget that all community
members make their living with software, so it’s
not that any of these guys are altruists themselves.
It’s always the deal: What are my benefits, what
are the costs, what are the alternatives.”
We also asked him, since Apache is mentioned as
an example of an innovatively successful (though
maybe not commercially successful) open source
product company, if he thinks that their statement
on dual licensing has some negative influence on
the acceptance of dual licensing for the developers’ community. Wittig’s reply was: “Maybe. But
I think you may underestimate the power of open
source/community-driven software production and
distribution. It is certainly easier/faster to ‘commoditize’ a well understood mass-market with a
low-cost product. But the economics also say: it is
cheaper to build and distribute software. If it then
happens to innovate, it can only be better. It just
takes longer, because we have 2 adoption hurdles
to overcome: open source in general (which less
and less is a hurdle, but it has been 5 years ago)
and [the] product category itself…. But the economic fundamentals are on your side, and that
matters to win a business.”
Looking at the main characteristics of Apache we
can see that Apache
•
is open source software
•
is not a commercial company
Possible explanations why Apache does not need
to dual license:
•
Most of the Apache core developers have
steady incomes as they are employed by
companies that want Apache’s software.
Apache gets its funding through its foundation.
•
Apache started off with the license they
have, and to change it now would require
the agreement of all the people who have
contributed with code in the past. In contrast
e.g. MySQL, started off with a private company’s software, and could set up any type
of licensing they needed.
•
Changing licenses is not easy unless you
have the developers all agree to assign their
copyrights to one entity, so that entity can
determine what license changes occur. This
is what the FSF does for its GNU code.
We asked Chris DiBona, Open Source Programs
Manager, Google Inc., why in his opinion Apache
does not support dual licensing, like for example
MySQL. DiBona’s reply: “The only thing I’d
mention is that MySQL doesn’t actually do a dual
license in the spirit in which Apache talks about,
they have an open source license and a commercial license. Yes, two license, but Apache fears
that an Apache/GPL or Apache/MPL mix leaves
users in a situation where they do not know what
software is what and often, dual licensing is just
stupid, as the software developer can just choose
the least restrictive license.”
The main reason Apache does not use dual licensing is probably their limited interest in revenue generation. Their prime goal is to have very
large usage numbers. For such a goal, the most
logical model is single-license with an academic
(permissive) license – just as they have.
ID: 112
We believe there are two important factors that
need to be considered regarding the acceptance of
dual licensing in the specific community of developers for whom the product is targeted, no matter
of the type of product:
1. The interpretation and explanation of dual licensing given by the company
2. The willingness of the user community to accept such a licensing scheme
In addition, for companies offering a commodity
product:
3. The market must be price sensitive and a considerable percentage of users must be dissatisfied with direct competitors’ proprietary products.
For companies offering an innovative product:
4. Open Source software products based on innovative technology have to either solve specific
problems not well handled by main stream
competitors, and/or expand the market by offering a brand new value proposition and addressing a new market segment. Also here the
market needs are key for success. Price differentiation does play a role as well, but not as
strong as in the case of a commodity product.
Entry Market
The next dimension we consider is the entry market as an important discriminating factor.
We distinguish between companies operating in a
niche entry market from companies operating in a
horizontal entry market.
According to Cusumano [21] a horizontal market
is a market across industry or functional specialization, with a broad set of customers. A niche
(also called vertical) market is a market in a specific domain, e.g. specific industry, technical specialty, or specific platform.
For companies operating in a niche entry market,
we have identified the following important factors
for success:
1. They have to produce highly sophisticated
products that solve a specific set of problems
for which main stream competitors do not currently have a satisfactory solution.
9
2. They have to produce products that are optimized for the specific niche market, giving a
clear competitive advantage compared with
full-fledge general versions of similar products
done by direct and/or indirect competitors in
the broad market they operate.
3. There should be no competitive, dominant
open-source or proprietary software products
in their niche market
We believe that if the above conditions are satisfied, and if the software offered is innovative, then
operating in a niche market reduces risks and increases chances of success.
However, since the size of the community might
affect revenue generation, it is important that for
companies who choose to operate in a niche market with open-source software innovative products, that qualification of leads is effective and
supported effectively via redistribution channels,
e.g. companies redistributing the software embedded into their own products.
Offering a commodity product in a niche market
is possible but does not seem to make much business sense. Instead, a horizontal market should be
considered for that.
For companies operating in a horizontal entry
market, we have identified the following important factors for success:
1. They should offer a commodity product (both
for redistribution and to end users).
2. A considerable percentage of users must be
dissatisfied with competitors’ proprietary
products (e.g. price issues, or technical issues
such as reliability or performance), increasing
the market’s willingness to try the open source
alternative.
3. The product must satisfy the needs of the market and compare favorable with direct competitors’ products.
4. They must build a large community. As Asif
Khalak in his work on an economic model for
impact of open source software points out,
“even if open-source is of equivalent quality
and is free of charge, it must still gain a ‘critical mass’ of the market share from those dissatisfied with the product of the market
leader” [9].
ID: 112
10
5. They must have a lower price than the direct
competitors and also a price cushion (see later
in the section on price)
•
Contributors, active in testing, error reporting, writing documentation, and other related services
If the above conditions are satisfied, a company
offering a commodity product in a horizontal
market has chances to get a significant market
share.
•
Users of the product, for private or business
purposes
•
Multipliers, writing product evaluations, doing public relations and marketing, act as
referential, participate in forums etc.
The strategies of exchanging/changing the entry
market will be discussed later when we look at the
market strategy dimension.
An alternative way, suggested by Christof Wittig,
to look at the market for open source software, is
to consider open source markets not as segmented
with the traditional regional or industry view (an
end-user view, depending on physical or domain
knowledge proximity), but on programming languages, technology stacks and application types
(devices vs. websites, for instance). Wittig also
suggests that a demographic segmentation works
very well: Who is in charge of selecting the tools?
That looks very different for a developer at
Merrill Lynch compared to a start-up entrepreneur
[22]. Although this is a very interesting distinction
we will not consider it a subject for future research.
Synergy
For a company offering an open source software
product, synergy with other open source software
products in the same or related markets increases
chances of adoption.
This is e.g. the case for MySQL and the so-called
LAMP (Linux, Apache, MySQL and Python/ Perl/
PHP) stack. The existence of synergy reflects in a
larger user base, which in turns produces greater
revenues.
Community
A successful open source software business model
requires active support from the developer community. We can give a more precise definition of
active community by looking at the roles played
by its members:
•
Company-employed core developers
•
Other developers, not employed at the company
Inactive members of a community may:
•
Download but not use the product
•
Not download the product, just keep themselves informed.
An in-depth discussion of the various roles of
open source community members can be found in
[23], further investigation on this issue is outside
the scope of this paper.
An important factor for success is the ability for
the company to create a single coherent community out of the various communities with interests
in the product: the free open source developer
community, the community of developers at
commercial customers, the product core developers at the company, etc. All of these developers
collaborate using the Internet in a virtual coherent
community associated to the product and its technology. The feeling of belonging to the specific
community is an important factor motivating
members to take an active role.
This is important since the size of the active community is directly related to the entry market size
and the source of revenues. We suggest a parameter to measure this:
p6 =
Size of active community
.
Size of community
Dual licensing has interesting, positive aspects in
this respect: the GPL-license facilitates creation of
an open source developer community, and can be
used as an efficient marketing and referential instruments; at the same time the non GPL-license
allows to add commercial developers and end users to the community.
Price
Price is an important factor for the success of any
business model. For dual licensing we consider
ID: 112
two price-related aspects as important for a successful business model:
•
Price differentiation: the product must be
cheaper than the direct proprietary competitors’.
•
Price cushion: the possibility to over time
increase the product’s price while maintaining price differentiation.
Price differentiation and price cushion are two key
factors for offering a commodity product in a
price sensitive, horizontal market. To a lesser degree the two aspects are also important for offering an innovative product in a niche market; however the niche market might be less price-sensitive
and more looking for new, innovative and reliable
products. If the niche market is new, the issue of
price differentiation might not be critical, at least
not until a direct competitor enters the market.
11
There is a risk that the development community
might start competing products, based on the same
software base (forks). This risk is however minimized exactly from the decision to allow source
code contributions only from company-employed
developers.
Spending
Spending plays a major role for any business
model. One of the advantages of open source
software in general is to allow the spending
mostly in R&D, and not in other (traditionally expensive) activities, such as marketing and sales.
This is due to the nature of the Internet and the
role and support of the active developers’ community and of the various channels.
GPL can be seen as a powerful, low-cost marketing vehicle to support the adoption of a product.
The above applies to dual licensing as well.
Software Development
Organization and Management
Development of open source software products is
typically characterized by the following factors:
Companies operating with open source software
in general tend to have flat hierarchy structures.
This is due to the supportive role of the developers’ community, to the companies’ culture, and to
economic factors, since a flat organization minimizes costs.
•
Globally distributed software development
and Internet-enabled processes.
•
Large software projects with little management overhead (by some authors referred to
as open source Globalization in [24])
•
Heavy support from the developer community.
•
Entirely Internet-enabled production system.
•
Company-employed core developers are often selected from the open source developer
community.
Specifically for companies offering dual licensing
one factor is very important:
•
Only company-employed core developers
contribute source code, thus ensuring the
company’s intellectual property rights for
the product.
This confirms what has been already observed by
Mikko Välimäki [12] that copyright and control of
the core product development is held in one hand,
the original developer, and that the ability to license the product with other terms than open
source requires full ownership of all rights to the
product.
This is not different for companies adopting dual
licensing; such companies are often organized as a
virtual organization, with few local offices. As a
result, core developers mostly work from home,
using the Internet to communicate and share code,
and seldom meeting face to face.
Customer Support
Contrary to service companies, open source software product companies provide limited direct
customer support, but rather use the developer
communities or third parties for that. The developer community provides a key dynamic support
resource and add-ons for all users of the product.
One of the value-added benefits of dual licensing
is that software developed under the GPL-license
enhances the commercial product quality without
consuming resources at the company, for example
when developers perform testing and write bug
reports and documentation.
Sales and Distribution Channels
Usually, companies operating with open source
software do not use classical sales and marketing
ID: 112
12
strategies, instead preferring viral Internet marketing as the prime strategy.
Market Strategy
Customers’ product evaluation is conducted exclusively via free downloads of the open source
software. This enables a “bottom up” sales approach (from the developers’ community to the
customer’s management) and reduces the costs for
new customer acquisition.
Strategy 1: Start Offering a Commodity Product
in a Horizontal Market and Expand Market Share
Even though some sales capabilities such as telesales and dedicated account resources are still
necessary, the low cost of sales and distribution
channels makes customer acquisition considerably
cheaper than for proprietary software companies.
For companies with dual licensing, GPL is used as
an effective marketing vehicle, as it helps creating
a user base, whereby the non-GPL part of the
business allow ISVs (independent software vendors) to play the role of business escalators.
Because an open source product is distributed
with its source code, ISVs have an important advantage in packaging and debugging their own,
final product. Most importantly when ISVs operate as business escalators this directly affect revenues.
We therefore have the following relationships:
Community buyers → Direct revenue ⎤
Community non - buyers → Indirect revenue⎥ →
⎥
ISVs → Direct and indirect revenue⎦
Revenue generation
Financial Support
The type of financial support for the company has
a major implication on the market strategy and
product offering of the company.
Companies with commodity, open source software products with large communities may receive VC money. To achieve the return of investment required by the VC, the company must often
have an efficient expansion strategy against the
proprietary software competitors in the same market
Without VC money, companies can operate initially in niche markets, trying later to expand to a
horizontal market by using for example an “aggressive” horizontal strategy.
We consider two market strategies with respect to
their potential success for a business model for
OSS with dual licensing.
This strategy works under the assumptions that:
•
The product offered is a commodity product
•
There is a significantly large number of customers in the specific market who are dissatisfied with the product(s) of the market
leader (for example due to high prices) and
are willing to try the open source alternative
product.
•
Price differentiation is an important decision
factor for the chosen market.
An example is MySQL’s focus on the web-based
database applications market, as seen as part of
the general database market.
Strategy 2: Start Offering an Innovative Product
in a Niche Market and Expand to a Horizontal
Market
This strategy can be used to establish what Geoff
Moore (using D-Day and the taking of Omaha
Beach as metaphor) calls a beachhead before assaulting the mainstream [7]. Like the beaches of
Normandy, high cost and certain losses can be essential to establishing the toehold that stages victory. Also important for Moore is the idea that to
reach the mainstream market, vendors must provide the complete solution, including support.
Moore claims that lack of service and complete
solutions is one big reason technology firms fall
into the “chasm” with a technically good, innovative product.
The main challenge for companies operating in an
entry niche market is to evaluate if the transition
to a horizontal broader market is possible and
sustainable.
We asked Christof Wittig what makes the innovation technology successful for a horizontal market,
as opposed to a niche market only? His reply was:
“If you can effort to be horizontal, it’s better, because you get economy of scale. The question is
whether the market needs or the differentiation
ID: 112
imposed by competitors forces you to specialize
on verticals.”
A byproduct of the free downloads approach is
that use of the technology/product spreads by itself across regions, platforms, applications, industries in a totally scattered and to some extent
uncontrolled way. This results in a so-called “aggressively horizontal” approach as defined by
Jerry Fiddler [25].
This means that a company that starts up with a
niche as entry market, finds its way outside the
niche market due to the combined effects of:
13
•
Another element that may prove not be any
more true is the limited direct customer support
VII. TWO BUSINESS MODELS BASED ON
OPEN SOURCE SOFTWARE WITH DUAL
LICENSING
Based on our analysis we have identified two
classes of business models that in our opinion
have potential for success. We call them:
• Open Source Software Commodity Products
For Horizontal Markets (CPHM), and
Open Source → Free Downloads → Community • Open Source Software Innovative Products
In practice this is a low cost strategy to extend
For Niche Markets (IPNM)
market reach, and this has a direct relationship
Both business models share the following comwith potential revenue generation.
mon properties:
This approach opens up the company’s entry mar• Software products are open source: The
ket in a somehow uncontrolled way.
company produces software products (not
Assuming that in the niche market there is no
services) based (often exclusively) on open
market leader, and that for the targeted horizontal
source software
market—where a market leader exists—there are
• Dual licensing: The company offers both
dissatisfied users willing to try open source softGPL and non-GPL licensing.
ware.
• Revenue from product licenses: Revenue is
The major downside of this is that influence of
generated primarily by selling product licompetitors (direct and indirect) will start to be
censes: to companies for product redistribuevident. This might in turn negatively effect profit
tion, and to end users for use and support.
generation. The challenge for such companies is
to prove themselves able to transition from a niche
• Proprietary open source: The software is
market, with a small community and relatively
written exclusively by a core of companysmall revenue, to a horizontal market with a large
employed developers. This is an essential
community, direct and indirect competitors with
prerequisite for the dual licensing model,
both open source and proprietary software prodbecause the company owns all intellectual
ucts, and large revenues.
property of the software products. This is in
sharp contrast to early open source systems
based on GPL-like licensing such as Linux,
also referred to as non-proprietary, where
software is not owned by a single vendor.
In particular, when implementing this strategy
some of the elements that positively characterized
a sustainable business model may no longer hold.
Therefore:
•
The existence of dominant market leader(s)
is more likely.
•
Need to increase the size of the community
•
Spending increases
•
The choice of a horizontal market, which
may imply a geographically distributed
market means that the virtual organization
and management will have to be revised:
•
Price Differentiation: The product is
cheaper than direct competitors’ proprietary
software products.
•
Price cushion: The company may over time
increase the product’s price, while still
maintaining price differentiation.
•
Synergy with other open source products
helps adoption of the product and expands
the customer base.
ID: 112
•
•
•
Focus spending on R&D, not on other
activities such as marketing and sales.
Virtual organization and flat hierarchy
structure: Core developers work outside
company premises, using the Internet to
communicate and share code, and with few
face-to-face meetings
Software development as open source globalization: With little overhead, companies
manage large software projects of globally
distributed, individual contributors (also referred to as open source globalization in
[24]). The production system is entirely
Internet-enabled; open source developer
communities provide most core developers
and contribute heavily to project support.
GPL-licensed software may be enhanced by
voluntary developers, without requiring
company resources.
•
Limited direct customer support: Contrary
to service companies, only limited direct
customer support is offered; it is left to developer communities or third parties to take
care of this. The developer community provides the key support resource for all users.
•
Internet as marketing, sales and distribution
channel: The Internet is the primary marketing channel, and provides the opportunity
for customers to freely download software
products for evaluation and/or licensing.
This enables a “bottom up” sales approach
and reduces the cost of attracting new customers.
•
ISVs as business escalators: Because an
open source product is distributed with its
source code, ISVs have important advantages in packaging and debugging their own,
final product. Most importantly ISVs operate as business escalators, directly affecting
revenues.
The two business models however differentiate as
follows:
14
Open Source Software
Commodity-Products
for Horizontal Markets
Open Source Software
Innovative-Products for
Niche Markets
Large community
Smaller community
Commodity product
Innovative product
Horizontal market
Entry niche market
May obtain VC
financing
Privately financed
Strategy 1:
Expand/consolidate the
horizontal market
Strategy 2:
“Aggressively” expand
from niche to horizontal
market when market
conditions make this
possible
Interestingly, both business models can be used by
the same firm for different product categories.
This is the case for example of MySQL AB: the
regular MySQL product can be seen as a highly
popular commodity, but special versions like
MySQL Cluster are technology and innovation
leaders [13].
VIII. CONCLUSIONS
In this paper we have defined a new framework
analysis that can be used to evaluate existing and
new business models based on dual licensing of
open source software products. We also identify
market strategies for two advantageous business
models, based on dual-licensing of open source
software products. At present, both the framework and the strategies lack substantial empirical
evidence, which is obviously something we will
work on establishing in our future research.
EXHIBIT I:
OPEN SOURCE LICENSING
BSD-Style Copyright
This is the copyright used by Apache, and BSDbased operating systems.
The copyright basically allows everybody to use
the software provided it was credited. This kind of
copyright does not incentive companies to build
business based on open source software.
ID: 112
15
GNU General Public License
As defined From Wikipedia, the free encyclopedia, the GNU General Public License (GNU GPL,
or simply GPL) is a widely used free software license, originally written by Richard Stallman for
the GNU project. The latest version of the license,
version 2, was released in 1991, but at the moment
(early 2007) a lot of work is being put into making
version 3 of GPL, GPLv3.
GPL grants the recipients of a computer program
the rights of the free software definition and uses
copyleft to ensure the freedoms are preserved,
even when the work is changed or added to.
The Lesser GNU Public License
This is a modified version of the GPL, intended
for software libraries.
Copyleft is a licensing scheme to facilitate open
and decentralized software development. Its key
feature is that once a program is licensed by the
author, the subsequent programs based on the
original must also be licensed in a similar manner
[26]
A nice overview of various licenses may be found
in [27].
EXHIBIT II:
OPEN SOURCE DATABASES
The most notable examples of open source DBMS
(database management systems) are MySQL,
SleepyCat (Berkeley DB), PostgreSQL, MaxDB,
Ingres, SQLite, and Firebird. Of these, MySQL is
by far the most successful open source database
company. MySQL used viral marketing to penetrate markets, using the waves of interests in
Linux and Apache, and developed an effective
strategy in promoting itself in the developer community. Also the use of related open source products such as Linux, Apache, Perl, and PHP helped
adoption of MySQL.
•
In 2004 MySQL had 4 million active installations
and 10 million product downloads.
MySQL software was exclusively written by employees of MySQL. This was an essential element
to enable the dual licensing model, since MySQL
owned all of the intellectual property of the software produced.
MySQL’s development cycle heavily used the developer community for testing and debugging,
thus lowering development costs considerably if
compared with a closed software company.
Much discussion occurred on the pros and cons of
an open source testing process vs. the one of a
closed software company.
MySQL is a light version of a DBMS, lacking
some of the full-fledge functionalities typical of
established giants such as Oracle, IBM. MySQL
executives, however, do not share this view [13].
PostgreSQL developed from a project started at
the University of California, Berkeley in 1986,
and has had a strong root in research.
PostgreSQL is distributed under the BSD license
by a non-profit OSS community of individuals
and companies. This, in our opinion, has limited
the commercial potential of PostgreSQL.
Sleepycat’s Berkeley DB offered a pre-relational
DBMS mainly used in the embedded market.
The company started originally with a BSD free
license and then later moved to dual licensing
with GPL-derivative for the open source version.
The change in licensing combined with the choice
of a niche market (embedded systems), boosted
the company’s commercial possibilities. Sleepycat’s business model is similar to MySQL’s
The company was acquired by Oracle in February
2006.
EXHIBIT III:
OBJECT ORIENTED DATABASES
The business model of MySQL was based on a
dual licensing scheme: GPL for non-commercial
use, and non-GPL for commercial use.
In particular, commercial MySQL software:
•
could be distributed with non-open source
software
•
included a warranty from MySQL AB
came with development support.
DEFINITION AND SHORT HISTORY
2
An object database management system
(ODBMS, also referred to as object-oriented database management system, or OODBMS), is a da2
Reproduced with permission from ODBMS.org [28]
ID: 112
tabase management system (DBMS) that supports
the modeling and creation of data as objects. This
includes some kind of support for classes of objects, and for inheritance of class properties and
methods by subclasses and their objects.
In their paper, The Object-Oriented Database
System Manifesto [29], Malcolm Atkinson and
others define an OODBMS as follows:
An object-oriented database system must satisfy
two criteria: it should be a DBMS, and it should
be an object-oriented system, i.e., to the extent
possible, it should be consistent with the current
crop of object-oriented programming languages.
The first criterion translates into five features: persistence, secondary storage management, concurrency, recovery, and an ad hoc query facility. The
second one translates into eight features: complex
objects, object identity, encapsulation, types or
classes, inheritance, overriding combined with late
binding, extensibility, and computational completeness.
ODBMS were originally thought of to replace
RDBMS because of their better fit with objectoriented programming languages. However, high
switching cost, the inclusion of object-oriented
features in RDBMS, and the emergence of objectrelational mappers (ORMs) have made RDBMS
successfully defend their dominance in the data
center for server-side persistence.
16
Francois Bancilhon from MCC starts development
of O2 at INRIA (France).
Tom Atwood at Ontologic produces Vbase, which
supports the proprietary language COP (C Object
Processor), eventually eclipsed by C++. Ontologic
becomes ONTOS, and the database is rewritten to
support C++. Tom leaves Ontologic in the late
1980s and founds Object Design (now part of
Progress Software) with ObjectStore (based on
C++).
Another product from that time is Objectivity/DB.
Drew Wade has been one of the founders of its
vendor, Objectivity.
1991 – ODMG
Rick Cattell (SunSoft) initiates ODMG (the Object Data Management Group) with 5 major
OODBMS vendors. The first standard, ODMG
1.0, was released in 1993. Throughout the 1990s,
the ODMG works with the X3H2 (SQL) committee on a common query language. Though no specific goal is achieved, the efforts heavily influence
the ODMG OQL (object query language) and, to a
lesser extent, SQL: 1999.
1995 - The OODBMS Manifesto
Malcolm Atkinson et. al release “The Object-Oriented Database System Manifesto” [29].
1990s - First Growth Period
Object databases are now established as a complement, not a replacement for relational databases.
Market for commercial ODBMS products grows
to some $100M, peaks in 2000 and shrinks since.
SHORT HISTORY OF ODBMS
A final ODMG 3.0 standard is released. Shortly
thereafter, the ODMG submits the ODMG Java
Binding to the Java Community Process as a basis
for the Java Data Objects (JDO) specification.
Afterwards, the ODMG disbands.
Early 1980s - Orion Research Project at MCC
Won Kim at MCC (Microelectronics and Computer Technology Corporation) in Austin, Texas,
begins a research project on ORION, which will
later lead to two ODBMS products: ITASCA (no
longer around) and Versant.
Late 1980s - First wave of commercial products
Graphael, a Lisp-based system appears from the
French nuclear regulatory efforts. Graphael eventually goes through a re-write and becomes Matisse.
Servo-Logic (later GemStone Systems) begins
work on GemStone.
2001 - Final ODMG 3.0 standards released.
2004 - Advent of open source
db4objects is released as free, open source
ODBMS. In November 2005, db4objects is first to
implement Native Queries [30] as an object oriented data access API that relies entirely on the
programming language (Java/C#) itself.
ID: 112
17
EXHIBIT IV:
EMBEDDED DATABASES
An “embedded” database system is a database that
is inside another software product. Typically the
embedded database is not visible, and it is customized to fit the requirements of the application,
which often does not need all functionalities of a
“classical” database management system, for example reporting.
ACKNOWLEDGMENTS
Marten Mickos, CEO MySQL and Christof Wittig, CEO db4objects, provided useful feedback to
earlier versions of the paper.
REFERENCES
[1]
Embedded database systems must be very robust,
since in the environment where they operate there
is no database administrator.
According to IDC [31] the embedded database
market grew 15% to reach 1.8 billion in 2004.
IDC [32] has identified two key elements to characterize the potential of the embedded-DBMS
market:
•
avoid direct competition with large DBMS
vendors
•
enable rapid growth (by royalty or run time
license fees) through the ISV channel without requiring a corresponding growth in
support or sales staff.
[2]
[3]
[4]
[5]
IDC suggests the following key elements for a
strategy to enter this market [32]:
•
Make the DBMS self-managing, requiring
no DBA, with an API to allow the tool or
application in which it is embedded to manage the database.
•
Decide what sorts of products to target for
embedding the DBMS
•
Identify the targeted software product types
based on a determination of which markets
make the most sense for the DBMS product
based on its characteristics and the growth
potential for those markets
•
•
[6]
[7]
[8]
Ensure that the size, performance, and reliability characteristics make sense for embedding in such products
[9]
Price the product so that its royalty or runtime fees do not distort the overall price of
the target product.
[10]
McKelvey, M. (2001): The Economic Dynamics of Software: Three Competing
Business Models Exemplified through
Microsoft, Netscape and Linux. Economics
of Innovation and New Technology, 11: p.
127-164.
Katz, M.L. and C. Shaprio (1985): Network Externalities, Competition, and
Compatibility. American Economic Review, 75(3): p. 424-440.
Katz, M.L. and C. Shaprio (1986):
Technology Adoption in the Presence of
Network Externalities. Journal of Political
Economy, 94(4): p. 822-841.
Armour, P.G. (2000): The Case for a New
Business Model: Is Software a Product or
a Medium? Communications of the ACM,
43(8): p. 19-22.
Behlendorf, B. (1999): Open Source as a
Business Strategy, in C. DiBona, S. Ockman, and M. Stone (eds.): Open Source,
Voices from the Open Source Revolution.
O'Reilly.
Feller, J. and B. Fitzgerald (2000): A
framework analysis of the open source
software development paradigm, Proc.
Twentyfirst International Conference on
Information Systems (ICIS), Brisbane,
Queensland, Australia: Association for Information Systems.
Moore, G.A. (1999): Crossing the Chasm:
Marketing and Selling High-Tech Products to Mainstream Customers. revised ed,
New York: Harper-Perennial.
Lacy, S. (2005): Open Source: Now It's an
Ecosystem, in BusinessWeek Special Report.
Khalak, A. (2000): Economic Model for
Impact of Open Source Software, MIT,
Department of Mechanical Engineering:
Boston.
Krishnamurthy, S. (2003): An Analysis of
Open Source Business Models, in J. Feller,
et al. (eds.): Perspectives on Free and
ID: 112
[11]
[12]
[13]
[14]
[15]
[16]
[17]
[18]
[19]
[20]
[21]
[22]
[23]
[24]
Open Source Software. MIT Press: Cambridge, Massachusetts. p. 279-296.
Kim, E.E. (1999): The Joy of Unix. Linux
Magazine, (November).
Välimäki, M. (2003): Dual Licensing in
Open Source Software Industry. Systemes
d'Information et Management, 8(1): p. 6375.
Mickos, M. (2006). Personal communication, November 28, 2006.
Kawasaki, G. (2006), Ten Questions with
Marten Mickos, CEO of MySQL,
http://blog.guykawasaki.com/2006/08/ten_
questions_w_2.html
Anonymous (2006), HP Extends Open
Source Service to MySQL Software,
http://www.hp.com/hpinfo/newsroom/pres
s/2006/060425a.html
Burgelman, R.A., S. Inkinen, and C. Wittig (2005): MySQL Open Source Database
in 2004, in R.A. Burgelman, A.S. Grove,
and P.E. Meza (eds.): Strategic Dynamics:
Concepts and Cases.
Kerner, S.M. (2005), MySQL Grows Its
Network,
http://www.internetnews.com/devnews/article.php/3483336
Anonymous (2005): Rimer's Rules for
Open Source. BusinessWeek online.
Anonymous (2007), Web Server Survey,
http://news.netcraft.com/archives/web_ser
ver_survey.html
Anonymous, Apache License v2.0 and
GPL Compatibility,
http://www.apache.org/licenses/GPLcompatibility.html
Cusumano, M.A. (2004): The Business of
Software: What every manager, programmer, and entrepreneur must know to thrive
and survive in good times and bad, New
York: Free Press.
Wittig, C. (2006). Personal communication, Dec. 1.
Ye, Y. and K. Kishida (2003): Toward an
Understanding of the Motivation of Open
Source Software Developers, Proc. International Conference on Software Engineering (ICSE 2003), Portland, Oregon,
USA.
Wittig, C. (2006), The 4th Element: Open
Source and Globalization,
http://www.db4o.com/about/news/publicat
18
[25]
[26]
[27]
[28]
[29]
[30]
[31]
[32]
ions/visionen_2006_5%20Wittig%20(abri
dged).pdf
Barnett, B. (2002): Entrepreneurship: Formation of New Ventures, Lecture Notes
S353, Stanford Business School.
Mustonen, M. (2003): Copyleft - the economics of Linux and other open source
software. Information Economics and
Policy, 15(1): p. 99-121.
Widmer, U. and K. Bähler (2006): OpenSource-Lizensen - Wesentliche Punkte für
Nutzer, Entwickler und Vertreiber, in B.
Lutterbeck, M. Bärwolf, and R.A. Gehring
(eds.): Open Source Jahrbuch 2006:
Zwischen Softwareentwicklung und
Gesellschaftsmodell. Lehmanns Media:
Berling.
Grehan, R., Introduction to ODBMS,
http://www.odbms.org/introduction.html
Atkinson, M., et al. (1989): The ObjectOriented Database System Manifesto,
Proc. First International Conference on
Deductive and Object-Oriented
Databases, Kyoto, Japan.
Cook, W. and C. Rosenberger (2005),
Native Queries for Persistent Objects,
http://www.odbms.org/experts.html#article
2
Olofson, C.W. (2005), Embedded Databases: The Invisible Engine That Could,
http://www.db4o.com/about/company/bac
kgrounder/IDC_419_web.pdf
Anonymous (2004): Worldwide Embedded
Database Management Systems 2003 Vendor Shares, IDC.
Download