|
IN THE UNITED STATES DISTRICT COURT
FOR THE DISTRICT OF COLUMBIA
UNITED STATES OF AMERICA,
Plaintiff,
v.
MICROSOFT CORPORATION,
Defendant.
STATE OF NEW YORK ex rel.
Attorney General DENNIS C. VACCO, et al.,
Plaintiffs,
v.
MICROSOFT CORPORATION,
Defendant.
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
Civil Action No.
98-1232 (TPJ)
FILED UNDER SEAL
(Seal removed pursuant to court's
October 14, 1998 Order
Civil Action No.
98-1233 (TPJ) |
DIRECT TESTIMONY OF GLENN E.
WEADOCK
Page 1
.I. OVERVIEW
1. I am
providing this testimony based on substantial experience working with,
teaching about, and writing about Microsoft operating system and application products,
and on my conversations with computer managers in large organizations, deposition
testimony, and documents in connection with this case. The major points discussed in
this testimony are as follows:
 It is
difficult to define software products according to any specific grouping of
files. Software products are typically defined according to their
features.
Organizations
generally have different needs for browser products, which they
view
as applications, and operating system products.
Whereas all
organizations using computers need operating systems, some wish
to
have no browsers deployed to some or all users.
Many
organizations wish to use browser products and have compelling reasons
to
standardize on a single such product, but want to select that product
independently of any particular operating system.
Many
organizations, having standardized on Windows 95 for a large portion of
their
operating system needs, have sought, at some cost, to remove Internet
Explorer from their PCs by either deleting the means of access to Internet
Explorer or standardizing on the original version of Windows 95, which
did
not
come with Internet Explorer at all.
Some
organizations, having moved to Windows 98 or considering doing so, are
confronted with the inability to remove Internet Explorer from the
operating
system, and accordingly view themselves as having little or no choice but
to
standardize on Internet Explorer.
Page 2
II. BACKGROUND AND QUALIFICATIONS
2. I am a
seminar developer, seminar instructor, author of computer books and
videos, and computer consultant. My computer knowledge is a combination of education,
years of hands-on experience, and thousands of discussions in formal and informal
settings with other computer users. I am President of Independent Software, Inc., in
Golden, Colorado, a company I co-founded in 1982.
3. Over the
years, my consulting activities have covered a wide range of
computing topics, including hardware platforms and software environments. My early
consulting practice focused on custom programming and system integration. Later, I
focused on technical support and microcomputer networking in organizations of all sizes.
Recently, writing and teaching have occupied more of my time than consulting. I have
submitted my curriculum vitae to the Court as Government Exhibit 1177.
4. In the
seminar field, I have developed seminars about Windows 3.1,
Windows 95, and Windows 98, networking PCs, and Help Desks (i.e., technical
computer support groups inside medium to large organizations), and have led technical
seminars throughout the United States, Canada, and the United Kingdom, in association
with Data-Tech Institute, since 1988.
5. I am the
author or coauthor of the following books: Bulletproofing Windows
98 (McGraw-Hill, October 1998); Windows 98 Registry For Dummies (IDG,
1998);
Windows 95 Registry For Dummies (IDG, 1995); Small Business Networking
For
Dummies (IDG, 1998); Intranet Publishing For Dummies (IDG, 1997);
Bulletproofing
Windows 95 (McGraw-Hill, 1997); Creating Cool PowerPoint 97 Presentations
(IDG,
Page 3
1997); Bulletproofing Client/Server Systems (McGraw-Hill, 1997);
Bulletproofing
NetWare (McGraw-Hill, 1996); Bulletproof Your PC Network (McGraw-Hill,
1996); and
Exploding the Computer Myth (Wiley, 1995). I have also written several articles in
computer and business magazines, and I have been interviewed by Computerworld,
MacWeek, and Inc. magazines on business computing issues.
6. As a video
author, I have developed several long-format technical computer
videos on subjects including Windows 95 and Help Desk management. I also contributed
to the development of several Internet-related computer videos.
7. I am a
Microsoft Certified Professional, certified specifically as a Windows
95 Product Specialist, Windows 95 Migration Specialist, and Windows 3.1 Product
Specialist. I also have participated and am currently participating in a number of beta test
programs (i.e., programs in which the creator of software products distributes
pre-release
versions of products to a limited set of users for testing, experimentation, and feedback),
including Windows 98 and Windows NT 5.0. My current professional memberships
include the American Society for Training and Development and the Association for
Computing Machinery.
8. I hold a
Bachelor of Science in General Engineering from Stanford
University, where I graduated with distinction in 1980. My course of study included
computer science courses and engineering courses requiring computer programming. I
have continued my technical education since then by attending seminars and courses on
Windows, networking, programming, communications, and the Internet.
Page 4
9. In October
1997, the Department of Justice asked me to provide consulting
services regarding Windows 95 and Internet Explorer. I have since provided services
ranging from explaining features of Windows and Internet Explorer to performing
experiments and tests with those products, and other products, regarding features, design,
removability, compatibility, and the user interface. I directed the creation of a videodisc
used in the January 13, 1998 contempt hearing, and testified in that hearing as to the
results of my tests. I have provided information and commentary to DOJ regarding
corporate and end-user computing issues, as much of my writing, teaching, and
consulting concerns those subjects. I have commented on filed documents (briefs and
depositions) in order to clarify technical issues and provide insight relating to end user
and corporate concerns with regard to operating system, browser, and other software
products.
10. This testimony relies upon the following sources of information:
My sixteen
years' experience in the computer industry, including my own work
with
the Internet and multiple versions of Microsoft's Windows operating system
products;
Research that I
have done in the course of writing eleven books, six seminars, and
four
videos;
Thousands of
articles and books that I have read about computing and computers;
Conversations,
both in person and by e-mail, that I have had with Windows
software developers;
Conversations
that I have had in seminar and consulting settings with hundreds of
computer professionals who work in the Windows and Internet fields,
particularly
Page 5
with
regard to issues of support and management of networks of computer users in
large-
and medium-sized organizations;
Interviews
with corporate PC managers in which I have participated at the request
of the
Justice Department;
Summaries of
interviews that DOJ representatives have conducted of corporate PC
managers;
Research and
experimentation done at the request of DOJ with different versions
of
Windows 95, Windows 98, Internet Explorer, and other software products;
My review of
documents and deposition testimony (of Microsoft employees and
other
witnesses) in the months prior to this trial; and
My own
experiences in custom, non-commercial application software
development.
11. The interviews in which I have participated at the request of the Justice
Department included top-level technology managers (Chief Information Officers, Vice
Presidents of Information Services, and the like) from the following organizations:1
American Stores; Citibank; ConAgra; Federal Express; Florida Department of Revenue;
GE Supply; Informix; J.C. Penney; Liberty Corporation; Playboy; Morgan Stanley/Dean
Witter; Sabre Group; and US Steel Group. The total number of PCs under management
by the individuals I spoke with is approximately 246,500.
Page 6
12. The conversations that were summarized for me by the Justice Department
included top-level technology managers from the following organizations: Boeing;
Chrysler; Ford; John Deere; and Motorola. The total number of PCs under management
by the individuals in this group is approximately 366,000.
III. DEFINING SOFTWARE PRODUCTS
A. Given the different ways that software products
can be
compartmentalized into files, the most reliable and useful definition
for
a software product is that product's feature set.
13. Software is among the most abstract and difficult products to understand
and to define. Given the fact, as I discuss in greater detail below, that software
developers can commingle code units (subroutines) with other, unrelated code units into a
single DLL file on disk, it is more appropriate, natural, and intuitive, to think of the
meaning of "software product" as a feature set -- that is, what the user sees -- than as a
fixed set of files.
14. Attempting to define software as a particular collection of files is
ultimately
impossible if code units within the same file are shared, either by multiple applications or
by a single application and an operating system. For example, Microsoft's word
processing software product, Word for Windows, ships with the file COMCTL32.DLL,
but that file is also used by Windows 95. Is the file COMCTL32.DLL part of Word for
Windows, part of Windows 95, or both? The answer is simple. If a file is relied upon by
both products, then it is can be considered a "part" of both, but it cannot be considered an
Page 7
exclusive "part" of either. Attempting to define software strictly as a collection of files
is
a fruitless exercise when some of those files perform double duty in different contexts.
As the director of information technologies for US Steel Group stated during an
interview: "You cannot isolate application code easily anymore." Similarly, Scott Vesey,
Boeing's Windows Web Browser Manager: "You know, it becomes very difficult to
draw a specific line at which you've drawn a boundary between operating system and
application." (S. Vesey Deposition, 155:23-25.)
15. Indeed, both industry professionals and computer customers think of a
software product more as that which enables a set of related features than as a collection
of specific files. For example, when a reviewer evaluates a software product in a
computer magazine, the reviewer typically focuses on that product's feature set: "This
product can do X, but not Y," and so forth. The list of files that come in the box, or the
list of code units that those files contain, is rarely if ever provided. Similarly, when I
teach computer concepts to users, I commonly hear users describe word processing
software as a program that provides text creation, editing, and printing features. In my
experience, users typically think of a browser as a program that provides Web page
display and navigational features on the public Internet or a private intranet.
16. Consistent with my experience, Microsoft employees also tend to describe
software products in terms of feature sets. For example, Chris Jones, Microsoft's Product
Unit Manager for Internet Explorer, has described two of the products Microsoft
associates with Internet Explorer as follows:
Page 8
Q: What is
Outlook Express generally?
A: It's a POP
3 mail client, POP3 and IMAP.
. . . . . .
.
Q: What is
NetMeeting generally?
A: It's a set of
services that let you do video phone and
video conferencing. (C. Jones Deposition, 80:2-4;
140:18-19)
Mr. Jones described Internet Explorer, as developed for the Apple Macintosh operating
system, in similar terms:
Q: How then
would you define what competitive browser
for other platforms, non-Win32 platforms, what that
means?
A: It's a very
interesting conversation. In some senses it's
just what our customers expect and want from us. It's
the thing that will let them go and deploy and take
advantage of the services on the Internet and
computing on the Internet. So what does competitive
browsing mean, if you are asking me what package we
ship on the Macintosh, we'll ship a package IE5 that
contains a set of features that people can use to browse
the Web, that ISVs can target and ICPs can target. (C.
Jones Deposition, 116:2-14.)
17. At the same time, Microsoft executives have acknowledged the difficulty
of
defining Internet Explorer according to files or code. Most notably, David Cole, who has
managed Microsoft's Internet Explorer development, stated in an internal e-mail on
January 2, 1998 (less than two weeks prior to a hearing held by this Court in which Mr.
Cole testified) that:
Taking out all of "retail IE" per the court order renders Win95
unbootable per my long affidavit. We can add files back one
Page 9
at a time and see what starts working, but I don't think there
is PR or legal benefit of explaining that. There is controversy
as to whether some files like comctl32.dll and others that
were part of the original Win95 should be fairly included on
the list, but the court said all of retail IE. We've never sat
down and decide [sic] what files are part of what, we just ship
them where needed. (Exhibit 989, D. Cole message to B.
Chase et al., January 2, 1998 [emphasis added])
B. Typical PC users can only use software products
that are loaded and
readily accessible.
18. The existence of a software product on any particular PC – that is,
whether
it is effectively present or absent from the customer standpoint – depends on both the
presence of the software enabling the product's feature set, and the means to use that
feature set. Without the means of access, a typical user has no convenient way to use (or
even determine the existence of) the software that exists on a CD-ROM disc, PC hard
drive, or floppy diskette. As John Kies, a Senior Product Manager at Packard Bell/NEC,
has testified:
Q: We talked
about the option as providing Packard
Bell/NEC with the option of removing the icon and the
Internet Explorer from the start menu. From a
marketing perspective, how does removal of the
Internet Explorer icon and Internet Explorer from the
start menu affect whether or not customers perceive
that the Internet Explorer is removed or not?
A: Well, if we
provide it without Internet Explorer in the
menu item, the customers feel like there's no browser
installed whether or not the actual code may exist
below the surface of the user interface. So we wanted
to give our customers the idea that they were receiving
as clean an OS as we could provide them. (J. Kies
Deposition, 27:1-16.)
Page 10
19. In fact, it is possible, and sometimes a matter of commercial practice, to
have software that exists on a disk or PC in the sense that its code modules are physically
present, but does not exist in any practical way from the user's standpoint because the
software is hidden, protected, or otherwise disabled. Consider the following examples:
Philip Barrett,
senior vice president of RealNetworks, has described his company's
RealPlayer software in these
terms:
Q: Focusing
on the RealPlayer and the PlayerPlus, does RealNetworks
distribute a version of the Player that includes the PlayerPlus
features not activated?
A: Yes. The
way the Player and PlayerPlus are related is basically
there's one Player. PlayerPlus features are activated by a license
key
that one gets by coming to our Web site and going through a
secure
form and purchasing that license key, although they believe they
are
purchasing PlayerPlus.
Q: So if a
customer doesn't purchase this activation key, what does the
user have?
A: The user
has the standard, what we call, ubiquity Player, the free
Player.
Q: But as a
technical matter, does the user have the code that
implements the PlayerPlus features?
A: Yes.
Q: Is it
correct to say that even if a customer has the bits of code that
make up the PlayerPlus, if the user does not have access to
the
PlayerPlus features, then, as a practical matter, the user doesn't
have
the PlayerPlus product?
A: From the
user's perspective, what they have is the standard Player.
Q: But not
PlayerPlus?
Page 11
A: Correct,
not PlayerPlus.
Q: And that's
because they're unable to access the PlayerPlus features?
A: That is
correct. (P. Barrett Deposition, 10/7/98, 45:)
Adobe Type
On Call is a CD that contains many typefaces, but the user must
receive an alphanumeric password in order to unlock them. This is the
vendor's
method for ensuring that payment is received before the user can work with
any
given
typeface.
Apple's
QuickTime Movie Player is a program that plays back digitized movies on
a
computer screen. As Timothy Schaaff, Senior Director of the Interactive Media
Group
within Apple, stated in his September 16, 1998 deposition:
Well, the business model that we have articulated for
QuickTime for the last six months or so has been a
model where people download the package of
QuickTime software for free. And then if they – and
by spending a small amount of money, they can obtain
an access key effectively that would unlock additional
capabilities. (T. Schaaff Deposition, 9/16/98, 455:5-
11.)
A common
way of distributing software is by making "trialware" or "shareware"
freely
available on the Internet. A customer can download a program and use it for
a fixed
period of time, say, 30 days. At the end of that time, in one common
scenario, the software may remain on the user's PC hard drive, but will no
longer
function until the user pays a license fee and receives a password or a
special
small
file that unlocks the program. Expired trialware or shareware physically
exists
on the PC in terms of bits and bytes, but once expired, the program is
effectively absent until the user pays for it.
IV. CUSTOMER DEMAND FOR
BROWSER AND OPERATING SYSTEM
PRODUCTS
A. For a variety of reasons, organizations generally
value the ability to
make application software decisions independently from operating
system decisions.
Page 12
20. Organizations2 typically evaluate, purchase, manage, train, and support
operating system software separately from application software. For example, operating
system software evaluations and purchases must consider computer hardware issues to a
much greater degree than application software evaluations and purchases, as the operating
system typically "insulates" applications from hardware details. Consequently, based on
my consulting and seminar experience, organizations generally value the ability to make
application software decisions independently from operating system decisions. Indeed,
the decision making process within most organizations is such that operating system
software and application software decisions occur separately, either in time – such as
when an organization evaluates which applications it needs first, and then selects the
operating system later, or when an organization wishes to make application decisions
across a variety of operating systems (as I discuss below) – or in terms of who makes the
decision (as is the case, for example, at American Stores, Boeing, and US Steel Group),
or both. Conversely, no information technology manager has ever told me that he or she
Page 13
does not value the ability to make application software decisions independently from
operating system decisions.
B. Organizations' needs and expectations for
browser products are
fundamentally different than their needs and expectations for
operating
system software products.
21. Managers at many organizations (for example, ConAgra, Morgan Stanley
Dean Witter, J. C. Penney, Playboy, Florida Department of Revenue, Boeing, GE
Supply) have told me that they value the flexibility to make browser decisions
independently from operating system decisions. For example, the CIO of ConAgra told
me on September 15 that his organization's choice of Windows NT 4.0 was not affected
by the browser issue "at all." Similarly, Boeing's Scott Vesey stated in his deposition
that "Those [browser standardization or browser acquisition] decisions have been made
separately from operating system decisions." (S. Vesey Deposition, 104:9-10.)
22. Organization managers have expressed that their needs and expectations
for
browser products are fundamentally different than their needs and expectations for
operating system software products. Managers typically (indeed, as far as I'm aware,
universally) consider Microsoft Windows 95, Windows 98, and Windows NT in the
category of operating system software.3 Microsoft publicly characterizes these products
as operating systems, and to my knowledge has never characterized them otherwise.
Page 14
Conversely, organizations typically consider browser software as application software,
like e-mail or word processing, not as an operating system or as part of a particular
operating system. I have recently heard or read this view from conversations or
documents from Boeing (see Exhibit 636, Boeing "Desktop Integration Roadmap"
presentation (TBC 000348-000390)), ConAgra, Morgan Stanley Dean Witter, J. C.
Penney, Playboy, Florida Department of Revenue, Boeing, and GE Supply. Technology
managers tend to be definite on this point, using phrases such as:
"No question – the browser is the same type software as e-mail or
calendar
applications." (Steve Ruegnitz, Morgan Stanley); or
"The browser decision is the same kind of decision as an e-mail or
word
processing decision." (Paul Kallas, Playboy); or
". . . browsers are considered by most of our customers as a third
party
application." (J. Kies [Senior Product Manager for Packard
Bell/NEC]
Deposition, 9/11/98, 25:10-12); or
"In the same way that we would want to be able to choose what
graphics
editor or what HTML editing product or what word processor we're
using,
we would want to be able to choose what Web browser we're using
for
those same business reasons." (Scott Vesey [Boeing] Deposition, 104:19-
23).
Similarly, Dr. Michael Dertouzos, Director of the Laboratory for Computer Science at
MIT, and formerly a Microsoft witness in this case, made this point quite clearly during
deposition testimony in this case on October 2, 1998:
Q: Is a
browser an application?
A: Historically
and today, it is the case that browsers are treated as
applications. (M. Dertouzos Deposition, 10/2/98, 36:1-2.)
. . . . . . . .
Page 15
I am not prepared to opine about the innards of any of these
systems
today as to how well they do one thing or another. However,
there is
a definite difference between the commands used by browsers
and
the commands used by operating systems. These are two
different
worlds that are always in two different historical contexts. (Id.
at
47:14-21.)
No corporate PC manager, in fact no one outside of the Microsoft organization, has
ever
described a Web browser to me as operating system software or as part of Windows 95 or
any other operating system.
C. Managers in organizations that do desire or
require browser software
on PCs deployed throughout their organizations commonly express
many reasons to prefer one specific browser over another, and these
reasons generally do not correspond or relate to their choice of
particular operating systems.
23. Organizations' needs for operating system and browser software also differ
with respect to the pervasiveness of need for each type of software within the
organization. Whereas all organizations require an operating system in order to address
even their most basic computing needs, not every organization needs browser software.
Internet and intranet browsing is a less universal computing need than, say, being able to
type information at a keyboard. (An intranet is a private computer network that
uses
Internet technologies such as Web servers and browsers.) For some organizations, a
demand may exist to have no browser on the PCs of all or some employees. Based on
my
experience and my interviews, I am aware of several reasons why this is sometimes the
case:
Page 16
a. An organization may wish to make it difficult for certain employees
to
access the public Internet, in order to reduce the amount of unproductive time
employees spend "surfing the ‘Net" on subjects unrelated to their jobs.
Without a
browser, accessing the Internet's World Wide Web is impractical.
b. An organization may wish to minimize computer resource use (disk
space,
memory, CPU power, etc.). Unused applications that reside on a computer
typically tie up such resources. The degree to which this is a concern
typically
depends on the average age of an organization's computers; the older (and
therefore poorer in resources) the computers are, the greater the
organization's
desire
to avoid wasting those resources. As Microsoft's Joe Belfiore has testified:
Because a PC system is a shared set of resources, memory
and processor are shared among all the different pieces of
code that may be running at any time. If any one piece of
code is taking up more of the shared resources than it needs
to, then it will impact the performance of the other pieces of
code…The more we can free up memory and processor so
that other applications get more of those resources, the faster
those other applications will run. And that's not specific to
our code; that's true of any code." (J. Belfiore Deposition,
138:4-18.)
Resource use is often a concern with programs as large as Internet Explorer.
For
example, a November 1997 memo from Microsoft's Chris Jones to the IE
Project
Team
states:
The IE4 browser, while fast, is simply too big for customers
to install and adopt, both in terms of memory usage [working
set] and also in terms of disk footprint [install size]. (Exhibit
364 (MS7 0004717-4728).)
Page 17
Later
in the same memo, Mr. Jones cites "remove size/performance issues
(working set and install size)" as one of three top product requests
corporations
made
to Microsoft in connection with Internet Explorer 4.0.
c. An organization may wish to simplify the standard PC setup to the
greatest possible extent, for ease of configuration, use, documentation, and
so
forth.
For example, Microsoft's James Allchin alluded frequently in his March
1998
deposition to the importance of simplicity in computer configurations:
Reduce concepts is something that we've been trying to drive
because of simplicity. It's just too complicated for end users
and we want to simplify the system. (J. Allchin Deposition,
3/19/98, 20:9-12.)
My
experience is that corporate technology managers do indeed value simplicity.
Section VI.A of this testimony discusses standardization benefits in more
detail.
d. An organization may wish to minimize the support costs associated
with
having "extra" undesired software on the typical PC. Those costs include
answering user questions, troubleshooting software problems, training
technicians,
and so
on.
24. For organizations, it is not always or even generally true that "any browser
will do" just because it comes bundled with a particular operating system. Managers
have expressed a variety of reasons why this is the case.
a. Perhaps most notably, an organization with multiple operating
systems already in use may prefer a particular application because it can work
on
Page 18
each of
those operating systems, enabling users throughout the organization,
regardless of the operating system installed on their PCs, to use an
organization
standardized product to access the Internet and/or the organization's Intranet.
In
my
interviews over the past two months, some managers (including those at
Informix, Ford, Federal Express, Boeing, and Morgan Stanley/Dean Witter)
have
stated
that their organizations deploy a variety of operating systems and hardware
platforms, and therefore prefer a browser having greater cross-platform
availability
and
compatibility.
A Morgan
Stanley/Dean Witter manager listed several reasons that his
organization places great value on cross-platform applications: less
training
expense, less testing expense, more effective internal communications via
e-mail,
less
application testing, more efficient user services, and lower Help Desk costs.
Scott Vesey of
Boeing makes the following comments:
Q: One of the
things that you testified about before was the fact that at
Boeing there are PCs and workstations that run not only
Windows
but various forms of Unix and in some cases the Mac
operating
system. Is that right?
A: Yes, that's
true.
Q: Okay. Is it
important in your evaluation and in Boeing's ultimate
choice of browsers that the browser chosen be able to run across
all
of those platforms?
A: We've seen
that as being something that we believe to be
advantageous. (S. Vesey Deposition, 88:13-22.)
Microsoft executives recognize this preference. As Chris Jones stated in
one
internal memo:
Page 19
Cross platform adoption continues to be a major adoption
hurdle. Many corporations and ISPs will not deploy our
products until the cross platform products have shipped. Sim
[simultaneous] shipping a lower featured cross platform
product is better than shipping later with more features
. . . . . . .
What would success look like? Sim ship
(<=4 weeks) IE +
key components on key platforms with a clearly articulated x-
plat [sic] common, competitive feature set/user interface
where appropriate/admin and deployment. (Exhibit 364, C.
Jones memo to IE Project Team (MS7 004717-28).)
b. Other managers (for example, at Federal Express and Chrysler) have
told
me that they consider the strategic direction of the browser vendor with regard
to
various Internet standards, such as HTML (the language used to format Web
pages)
and Java (a programming language that lets programs run inside a browser
window), in deciding which browser to adopt.
c. Some companies develop their own custom software that only works
with a
particular browser, and compatibility with that custom software may
provide an ongoing motivation to use that particular browser. (Browsers differ
in
the
precise specifications for evolving standards such as HTML and Java. Also,
Internet Explorer supports a category of programs that go under the name
"ActiveX," a category which Navigator does not support without extra
third-party
software.)
d. Organizations wish to protect PCs against unauthorized programs
such
as viruses and software "Trojan horses." The Internet is rife with such
Page 20
security risks. Different browsers use different methods to provide security,
and
have
different security-related characteristics. For example, ActiveX programs
allowed by Internet Explorer can present a security risk that is different from
the
risk
associated with Java programs allowed by Netscape Navigator, because
ActiveX and Java are two different kinds of programs.
25. Each of these factors -- wishing to have no browser for some employees;
wanting to standardize on a browser across operating system platforms; considerations of
browser standards and strategic directions; regard for compatibility with other custom
applications; and the need to choose an appropriate security method -- provides an
independent reason why organization managers generally do not want to have their
selection of browser products linked to their operating system decisions.
B. Receiving PCs and operating systems with no
browser preinstalled is
the preferred mechanism for uncoupling the operating system and
browser decision; purchasing PCs with an easily removable browser
is
the next best thing.
26. Generally, the preferable way to achieve a no-browser desktop (if one is
desired for any or all of the reasons set forth in paragraph 23) or to preserve the
maximum flexibility in browser choices (if such choices are desired for any or all of the
Page 21
reasons set forth in paragraphs 24 and 25) is to receive computers that have no browser to
start with. (See J. Kies Deposition, 9/11/98, 39:3-9; S. Vesey Deposition, 33:4-18.)
27. If a browser comes on a computer already (e.g. with an operating system)
and the browser can be easily removed, then removing the browser is the next best thing,
although doing so is not without cost. Some additional information on how organizations
buy and configure PCs is relevant to understanding how different organizations achieve
this result:
a. Some small and medium-size organizations often buy PCs
preconfigured with the software that they desire. That is, these companies
specify
an
operating system and a list of applications, and the PC vendor loads that
software ahead of time. If the operating system comes with an application that
the
organization does not want, then it must take steps to remove that application
on
every
PC that comes in the door. If, further, the operating system vendor does not
facilitate the removal of the application, the organization may have to
expend
considerable time and energy determining how to perform the removal.
b. Other small and medium-size organizations buy PCs, erase their hard
drives,
and install operating system and application software themselves (for
example, from network servers configured specially for this purpose).
Again,
however, if the operating system installation also installs an undesired
application,
the
organization must do extra work to remove that application.
Page 22
c. Larger organizations often create their own "hard disk image" or
"standard build" by installing the software they want onto a test PC,
configuring
and
customizing that software as they want, and then providing that master hard
disk
(or a tape or CD replica thereof) to the PC supplier. The PC supplier then
copies
that hard disk image onto each PC supplied to the organization, effectively
"cloning" the customized test PC's hard disk onto all other PCs that the
organization buys.
28. Once an organization decides to remove a browser or other application
product (assuming the preferred option of purchasing PCs without the product installed is
not available), the next logical question is what suffices as "removal."
a. Consistent with the discussion above of how software is generally
viewed and defined by computer users, and as was discussed at length in
the
hearing on January 13 and 14 of this year in the United States' consent
decree
case,
Microsoft itself refers in numerous documents and articles available to end
users,
and in Windows 95 itself (specifically, the "Add/Remove Programs" utility),
to the
"removal" of Internet Explorer as consisting of removing of the means of
accessing the program.
b. Corporate technical support managers have also frequently told me
that
inhibiting the user-accessible means of access to a software product (e.g., an
icon
on the "desktop" screen of the user interface, or entries in menus of program
options) has the same effect, from the support cost standpoint, of removing
a
Page 23
program in its entirety. Indeed, in the case of Internet Explorer, as discussed
in
greater detail in paragraph 40a, many organizations have done exactly this.
This is
a
significant consideration because reducing technical support costs is an
imperative for many, if not all, technology managers. Annual support
costs
typically amount to more than the purchase price of a computer, as studies by
the
Gartner Group and other respected industry analysts have repeatedly shown.
My
own
experience and interviews corroborate this finding. For example, Federal
Express' Managing Director of Technology told me that "Support costs are
a
larger
part of the equation above and beyond purchase price." The Director of
Information Technology for US Steel Group told me that "Support is a
big
problem." Because removing the user-accessible means of using a browser
product makes the product disappear from the perspective of the user,
support
costs
are significantly reduced, and a primary objective of removing the browser is
achieved. Managers with whom I have spoken (for example, at Morgan
Stanley/Dean Witter, GE Supply, and Motorola) have corroborated that this is
the
case.
c. Nonetheless, if a browser or other application product is designed in
such a
way that its removal leaves a significant amount of code behind, then the
organization may suffer from unwanted consequences (such as extra disk
and
memory resource consumption) if only the means of access are removed.
This,
along
with the costs of removal, is a significant reason why corporate technical
Page 24
managers may prefer that an undesired application never be installed in the
first
place.
I next discuss the technical aspects of bundling products which give rise to
this
desire.
V. TECHNICAL ASPECTS OF
BUNDLING
A. Microsoft has designed Internet Explorer so that
some of its subroutines
reside in the same file libraries as other subroutines that are necessary
for the operating system to function.
29. One of the decisions that software designers make is how to combine the
atomic units of code, called subroutines or functions, to make up files (or
"libraries") on
disk. Software designers have great flexibility in this regard. They can create a so-called
"monolithic" program that consists of a single, large file; they can create a highly
modular program that uses a hundred different library files (called DLLs, for
Dynamic
Linked Library) to contain a thousand different subroutines; or they can choose any
intermediate degree between these two extremes.
30. All current versions of the Windows operating system use a large number
of
files, containing an even larger number of subroutines. A software designer with source
code access may choose to place an application subroutine into a file that contains
operating system subroutines. As is discussed at greater length in the testimony of Dr.
Edward Felten, Microsoft, for example, has chosen to design Internet Explorer so that
some of the code that it uses co-resides in the same library files as other code needed for
Windows 98 or even Windows 95 to run. As Boeing's Scott Vesey testified:
Page 25
So, you know, many applications do make changes in the
Windows system subdirectory. Fairly few of them make the
kinds of modifications that Internet Explorer 4 makes. (S.
Vesey Deposition, 153:21-23.)
Microsoft is perhaps in a unique position to make modifications of this nature, as it has
full access to the operating system source code.
B. Designing an application so that its code coexists
in some of the same
files required by the operating system does not necessarily deliver
any
benefit, and may in fact present disadvantages, to the
customer.
31. The commingling (or "integration," as it is sometimes loosely referred to)
of application and operating system code in this manner is not technically difficult, nor
does it necessarily deliver a benefit to the customer. Dr. Dertouzos, for example, noted in
his deposition that:
It is relatively easy for a software manufacturer, be they
Microsoft or Netscape or anybody else, to, in effect, combine
what the browsers and operating systems do, but in a
superficial or perfunctory manner, which would make it look
like it's one happy front end, but in effect, it is the union, the
addition of the two other worlds. (M. Dertouzos Deposition,
103:17-24.)
32. In fact, commingling application and operating system code may be
disadvantageous in a variety of ways. I summarize the actual and potential disadvantages
in the subparagraphs below. These drawbacks often constitute additional reasons (above
and beyond those discussed above) why organizations often prefer to purchase and install
operating system and application products, including browsers, separately.
Page 26
a. The likelihood of an application failure affecting the operating
system
may increase when code is shared between the two. For example,
Microsoft has pointed out (H. Partovi Deposition, 89:21-24) that the failure of
a
browser window could crash the Windows 98 desktop, due to the fact that
the
same
file (EXPLORER.EXE) manages both the browser window and the desktop.
I have
noticed in my own experiments with Windows 98 that the failure of an
Internet Explorer window can cause the entire desktop to malfunction, bearing
out
Mr.
Partovi's comments.
b. An application that modifies operating system files could create
(and,
in the case of Internet Explorer, has been documented in some cases to
create)
conflicts with other applications and with company-developed applications.
This
concern was a primary reason, for example, that Boeing selected Netscape for
its
fourth-generation browser instead of Internet Explorer (Exhibit 637, 18-Month
Tactical Plan, 7/14/98 (TBC 000409-411).).
c. User confusion can result from combining application code and
operating system code, for example, as Microsoft has done with Windows 98
and
the
"single Explorer." As Boeing's Scott Vesey stated in his September 30
deposition: "I have found in some cases that interface to be confusing to the
end
user.
They often don't understand the context of what they're looking at." (S.
Vesey
Deposition, 63:23 - 64:1.) That is, whether the computer files being viewed
are
local, or on the Internet, something that is clear when one uses a "Windows
Page 27
Explorer" to view local PC resources and a separate "Internet Explorer" to
view
Internet resources. A November 1997 internal Microsoft memo from Brian
Hall
quotes
participants in Internet Explorer user focus groups as saying:
…why do we need to see local files through our web
browser? It's like a whole other version of windows explorer
in a web browser. Need one or the other, don't need both."
(Exhibit 218, B. Hall message re: browser focus group
summary, 11/24/97 (MS7 006352-6354.)
Just as some users might find it useful to adopt a "single paradigm" for viewing
information, others -- as Mr. Vesey and Mr. Hall have learned -- do not.
d. The required hardware resources can increase significantly when an
operating system integrates application software. For example, Windows
98
requires a great deal more disk space and significantly more memory than
Windows 95, largely due to Internet Explorer software. The fact that Windows
98
uses
some Internet Explorer software to provide an operating system service such
as a
help system may not represent a net benefit for organizations that do not use
Internet Explorer as the standard browser. The improvements offered by
HTML
Help
(and it is not clear to me, based on my own testing, that HTML Help is on
balance "better" than its predecessor, WinHelp, given that HTML Help is
slower,
less
reliable, and lacks an important "hotkey" feature) may not be justified by the
significant increase in resource requirements.
e. It may become more difficult to enforce security when an operating
system
integrates application software. For example, Microsoft provides ways to
Page 28
limit
what users can do in Windows 95 and 98, but these methods become less and
less
airtight the more Microsoft integrates application software into its operating
system. In Windows 98, one can restrict access to running a particular
browser
program directly by using Windows' "system policies" security feature, but
users
can
still activate the browser indirectly, for example, through a "help" file that
links
to the Web. Securing the system against users running programs that
management doesn't want them to run becomes more difficult as
application
software is folded into the operating system. As Boeing's Scott Vesey explains:
ActiveX is the ability to have the Web browser run what
essentially amounts to local workstation instructions without
user intervention. In other words, components can be placed
on a Web page, downloaded, installed, and executed without
user interaction…The problem with that dependency is that
you can't rely on training 170,000 people to make accurate
security decisions on what components should be accessed
and what components shouldn't. (S. Vesey Deposition, 41:21
- 42:6.)
f. Operating
system vendors that integrate applications feel new
pressure to align, or synchronize, the release of both kinds of products and
therefore may delay an operating system update in order to include an
application
update. This delay can be disadvantageous to customers eagerly awaiting
non-
application updates to the operating system. As Microsoft's James Allchin put
it
in his
March 1998 deposition:
But we were under massive pressure and IE3 wasn't
quite ready. So there was a different set of opinions
about should we satisfy the OEMs and ship that [the
Page 29
OSR2 version of Windows 95] now or should we just
wait and align it with IE3 instead of shipping yet
another OPK once IE3 was done. So instead of
shipping multiple of them, this was the idea that we
would sync them up, which is what eventually
happened. (J. Allchin 3/19/98 Deposition, 147:8-15.)
g. Organizations often wish to update applications without making any
changes to their operating system software. The main reason for this
preference is
that
changing operating system software has a greater potential for creating
problems than changing a single application does, inasmuch as all applications
rely
upon
the operating system. The commingling of operating system and application
code
in the updated application can cause unwanted problems with other
applications still residing on the system, or confusion among users now
confronted
with
changes to the operating system. Microsoft has apparently recognized this
need
with regard to Internet Explorer 5. As Chris Jones testified:
The key is to make our customers to say, hey, I just want to
go get the latest browsing technologies. I want my start menu
and task bar to remain the same. (C. Jones Deposition,
113:12-14.)
33. In addition to these disadvantages, there is an overarching drawback of the
commingling of applications (including browsers) and operating system products: it can
skew users' decision-making about which application to select. To the extent this
disadvantage has taken and will take the form of influencing organizations to select
Internet Explorer as their standard browser when they might not otherwise do so based on
independent evaluation of their options, I discuss it in greater detail in paragraphs 43-45.
Page 30
34. Even if organizations continue to select their applications of choice,
notwithstanding the commingling of code, the commingling imposes otherwise-unneeded
costs on that choice. These costs include not only those above, but also the costs of
seeking to circumvent the commingling by removing unneeded application software,
devoting time and personnel to modifying operating system configurations, or using older
versions of the operating system.
C. The fact that an application has code that
co-resides with operating
system code does not imply that the application vendor cannot provide
a
removal or "uninstall" feature, such as an entry in the "Add/Remove
Programs" utility in Windows 95 and Windows 98.
35. An operating system vendor can mitigate at least some of the potential
disadvantages of integration by providing customers the option of uninstalling the
integrated application (or by offering non-integrated options to begin with). The
commingling of application and operating system code does not preclude the developer of
the product from providing customers the ability to remove a bundled application.
a. In Windows 95 and Windows 98, users generally add or remove
software via the "Add/Remove Programs" feature within the Windows
Control
Panel
– a sort of master program for controlling the Windows environment. Using
the
"Add/Remove Programs" feature may perform a variety of different actions,
from
the "mere" deletion of a desktop icon (which, as discussed above, may not be
"mere"
at all) to the complete removal of all the code associated with an
application. The specific actions associated with the removal of each
particular
Page 31
application are up to the vendor of the application. In any case, the net effect
is, at
minimum, that the program so removed is no longer accessible or visible to
the
user.
b. As I have discussed previously in testimony before the Court, in
Windows 95, Microsoft did not preconfigure the Add/Remove Programs utility
to
enable
users to remove Internet Explorer 3.0. However, doing so would have been
possible, as is indicated by Microsoft's publication of the methods
consumers
could
use to achieve this result (specifically, separately installing Internet Explorer
3.02,
which made an entry for Internet Explorer appear in the list of removable
programs in the utility). Microsoft achieved this result even though IE3
commingles application and operating system code.
c. In the final OEM Service Release ("OSR") of Windows 95,
Microsoft bundled Internet Explorer 4. Once again, it did not include an entry
for
Internet Explorer in the Add/Remove Programs utility, but could have done
so.
d. The "Add/Remove Programs" feature is not the only way Microsoft
has
provided for adding and removing software. The Network control panel, for
example, allows the user to add and remove networking components –
again,
presumably because the demand for these networking components varies
from
customer to customer, and because the forced inclusion of all possible
networking
components would make Windows slower and more memory-hungry than it
needs
to
be.
Page 32
VI. ADVERSE EFFECTS ON
CUSTOMERS OF IE'S INCLUSION IN WINDOWS
A. Standardizing hardware and software is
important to most
organizational computer customers for a long list of reasons.
36. Organizations generally want the PCs that their employees use to have
standard hardware and software configurations throughout the organization, to the extent
feasible given the differing needs of different departments and user classes (e.g.,
design
engineers, bookkeepers, and in-house attorneys), and given the historical evolution of
computing resources within the organization. (Typically, "standard" hardware and
software means products that the organization supports through licensing, testing,
installation, user training, documentation, troubleshooting, and maintenance, and that the
organization either recommends or insists upon.) A design engineer may need some
specialized software (e.g., high-powered workstation operating systems capable of
running technical or scientific applications) that a bookkeeping clerk would not need,
certainly, but an organization would typically try to set up all design engineer PCs the
same way and all bookkeeping PCs the same way. Further, an organization would
typically try to standardize any application software that both design engineers and
bookkeeping clerks use -- such as word processing, Web browsing, or spreadsheet
software.
37. Many organizations presently must contend with some degree of variability
in hardware, operating systems, and applications. There are a number of reasons:
Page 33
organizations sometimes merge with other organizations that have used different
computer hardware or software; even within organizations, individual departments may
have not coordinated their computer hardware and software purchasing policies; and in
large organizations, updating software for every user simultaneously is often impractical.
The fact that such variability exists does not contradict the fact that organizations
generally attempt to standardize computer configurations and product mixes to the extent
that they can. Most information technology managers I have discussed these issues with
in my consulting and seminar assignments view the drive towards standardization as an
ongoing effort.
38. The reasons that organizations try to standardize hardware and software
configurations to the maximum practical extent are many and persuasive. In fact,
standardization has become something of a holy grail for many information services
managers -- greatly desired and justifying extraordinary effort because of the potential
rewards. Standardization provides the following benefits:
a. Reduced desktop configuration costs. Companies often create
standard "builds" for their PCs, that is, standard sets of software including
both
operating system and applications. These standard builds need to work
reliably
and
efficiently, and organizations usually test them thoroughly. Organizations also
often
customize each application to work best for the kinds of users and business
situations that are specific to that organization. (For example, customizing
a
browser program may mean redefining the "home page" that it accesses when
the
Page 34
user
clicks a button that looks like a house.) The more different applications in the
standard build, the more time, effort, and money is required to test and
customize
those
applications.
b. Reduced end-user training costs. The fewer different
applications
and
operating systems, the less end-user training is required. It is true that
browsers, as a type of software product, work more similarly to each other
than
many
other applications and therefore may require less training than, say, database
programs. However, different browsers do many things differently. For
example,
users
set browser preferences – such as how Web pages should print, how to
restrict viewing certain kinds of Internet sites, or whether the browser
should
display
pictures in addition to text – differently depending on whether they are
using
one browser or another.
c. Reduced technical staff training costs. This is a significant cost
for
organizations that provide their own internal computer technical support
services.
Troubleshooting an application requires a knowledge of how it works
internally,
and
how to configure its settings, not just how a user works with it.
Troubleshooting an operating system requires an even greater degree of
technical
knowledge. Therefore, despite the fact that users need relatively little training
in
the
basics of using different browsers for simple navigation, technical staff require
significantly more training to be able to support different browsers.
Page 35
d. Reduced third-party support costs. This is a significant cost for
organizations that hire out computer technical support services. Such
service
bureaus typically charge higher fees the more applications they are asked
to
support.
e. Reduced troubleshooting costs. A well-accepted maxim of
computer
troubleshooting is that the simpler the software setup is, the quicker
troubleshooting can be. Conversely, the more applications reside on a PC,
the
more
complex the software setup is, and the greater the costs that the organization
must
incur when troubleshooting takes more time and resources. As Boeing's Scott
Vesey
stated in his September 30 deposition: "I believe that we save support cost
and
are able to improve the quality of support to users based on setting a single
product standard." (S. Vesey Deposition, 42:21-23.)
f. Reduced
end-user confusion. Generally speaking, the fewer different
programs on a PC, the less is the likelihood of end-user confusion over how
to
productively operate the PC. As Microsoft's Joe Belfiore described:
And as some number of items on the screen, whatever screen
it is, increases beyond some number – and it depends on the
screen and the context – then it becomes more of a detriment
to a person's ability to use the PC than a positive." (J.
Belfiore Deposition, 203:24 - 204:3.)
g. Increased employee mobility. To the extent that applications are
standardized across the PCs in an organization, an employee can change PCs
(as,
Page 36
for
example, may be required after assuming a new job) with a minimum amount
of
relearning.
h. Increased employee productivity. The fewer different
applications
and
operating systems in use within an organization, the less time employees must
spend
learning software products and the more time they can spend using those
products to do their jobs.
i. Reduced
computer resource set. The "resource set" is the collection
of
computer resources, such as disk space, memory, and processor "horsepower,"
necessary to support a given collection of software products. A smaller number
of
applications reduces the required computer resource set, which can not only
save
hardware dollars but also improve the computer's speed.
j. Reduced
potential for incompatibilities between programs in the
standard build. The more software you load onto a PC, the greater the
potential for
conflicts and incompatibilities that can reduce the PC's reliability. A PC with
one
program on it should almost never "crash" or freeze up. PCs that have dozens
of
programs on them have more frequent problems.
k. Reduced potential for incompatibilities between programs in the
standard build and programs developed by users. Sometimes, users develop
their
own
applications that (if useful and well designed) may spread throughout an
organization. The development of intranets – internal company networks based
on
Internet technologies – has accelerated this trend. The greater the degree
of
Page 37
software standardization, the greater the likelihood that such
user-developed
applications can work properly throughout the organization.
l. Reduced
costs due to volume purchasing agreements. If an
organization buys 1000 copies of spreadsheet application A from vendor 1,
instead
of 500
copies of spreadsheet application X from vendor 2 and 500 copies of
spreadsheet application Y from vendor 3, then the organization can likely
negotiate a better per-unit price. Organizations may consider this issue even
for
applications that are presently "free," because managers understand that what
is
free
today may not be free tomorrow. Indeed, internal Microsoft documents
suggest that this may be cause for concern, even with regard to the
previously-
announced "forever free" Internet Explorer. (See Exhibit 725, B. Chase memo
to
P.
Maritz re: "idea" [for possible configuration of Windows 98], 7/19/97, MS7
005460-62 ("starts getting people to think about everything won't be free"
and
"what
do corp and other biz customers do? we are essentially telling them that they
get it
for free and then have to pay if they want the Unified Explorer later.")
m. Reduced costs due to licensing agreements. Many software licensing
agreements assess costs based on number of installed copies of an
application.
(The
director of information technology at US Steel Group voiced this as a
concern, and it is a common issue I hear about in my seminar and
consulting
work.)
Page 38
B. Organizations that want a browser application
typically prefer to
standardize on a single browser.
39. Following from the general and typically strong desire for standard
configurations, organizations that want a browser typically prefer to standardize on one
particular browser. (For example, Ford, Chrysler, John Deere, GE Supply, Motorola,
Informix, Playboy, Florida Department of Revenue, Boeing, US Steel Group, and Morgan
Stanley/Dean Witter.) Some organizations support a dual browser standard, but in my
experience these are in the minority and represent special cases. Indeed, even
organizations that must use two or more browsers (for example, to test how their
organizational Web sites will look when viewed by users of different browsers), tend to
standardize on a single browser for general organizational use. The most common and
compelling specific reasons for standardizing on a single browser include the following:
a. Reduced end-user confusion. As I have discovered in my testing
of
Windows 98 and as Microsoft employees have testified (for example, H.
Partovi
Deposition, 73:10 - 75:4), and as Dr. Felten describes at more length in his
testimony, whether Internet Explorer (which the user cannot conveniently
uninstall) or a different browser selected by the user is opened when a user tries
to
access
the Internet using any of the numerous means included in Windows 98 for
doing
so depends on which means of access is used; some do not respect the user's
choice
of default browser. The typical computer user is unlikely to understand the
technical reason for that distinction and is likely to be confused by the fact
that
Page 39
some
actions run Internet Explorer, even though the user has chosen some other
browser as the preferred browser. Corporate technology managers have
confirmed
to me
that this specific issue is a matter of concern to them. As a Boeing memo of
July
14, 1998 indicates:
Having two web browsers on the desktop will confuse users.
This confusion will be worse if some web applications require
Internet Explorer while others require Netscape. (Exhibit
637, 189-Month Tactical Plan, 7/14/98 (TBC 000409-411.)
Similarly, in a conversation I had on September 2, 1998, the technology chief
at
Playboy stated that "The worst scenario would be if some operations invoked
IE,
some
Netscape."
b. Security Concerns. Another source of user confusion and
organizational concern about having multiple browsers may be that
different
browsers use different security models and present different messages to
users
when
a potential security risk is detected.
c. Reduced computer resource set. This is a significant issue with
regard
to Windows 98's forced use of Internet Explorer software regardless of the
customer's choice of an alternative browser. As a July 1997 Microsoft
internal
memo
states: "Windows 98 is inescapably most appealing to the IE user…in
many
scenarios, the 16MB Nav[igator] user will have a much slower experience
with
98 than 95." (Exhibit 725, J. Roberts memo to B. Gates et al. (MS7 005460-
62).)
Page 40
d. Reduced potential for incompatibilities between programs in the
standard build and programs developed by users. As Boeing's Scott Vesey
stated
in his
September 30 deposition:
If we have multiple browsers, we run into a potential problem
whereby we will have applications that work best if viewed
by a specific browser product. And that is something that we
seek to avoid. . . .
I would say probably it's likely that if I told people tomorrow
that yes, Internet Explorer is standard, we would suddenly
have a lot more sites with specific dependencies which would
trigger more people to have more duplicate browser software
installed, increasing complexity, increasing the amount of
support requirement that we had, and lowering eventually the
quality of the support that we can provide, all of which costs
us more money. (S. Vesey Deposition, 77:19-22; 135:1-8.)
e. Desire to provide uniform access to organizational Intranets.
Web
pages
on an internal intranet may look different when viewed by different
browsers. As Packard Bell/NEC's John Kies testified:
Q: What is the
basis for your understanding that it would
entail more support to have two browsers on there?
A: Well,
generally a company, like they're developing an
intranet for their own in-house information, will
understand which browser they're using or the
company is using to access that intranet. And if they
had standardized on [Netscape] Communicator and --
but some of their end users were using Internet
Explorer, the screens that came up might not work the
same. So although the products are very similar, in
some cases Java applets or other items that are tuned or
customized to one of the two primary browsers don't
work. So that would be one example. (J. Kies
Deposition, 33:4-18.
Page 41
f. Reduced
development costs for custom applications. Organizations
that
write their own programs to work with Web browsers (such as, for example,
an
intranet-based expense report system) keep their development costs down if
they
only have to write those programs to work with one browser. (This point was
made
to me by senior technology managers at GE Supply and Morgan Stanley/
Dean
Witter.)
C. Another consequence of valuing standardization
is that organizations
value the ability to remove applications that they do not want or
use.
40. Following from the desire for standard configurations, information
technology managers and support technicians typically value the ability to remove
software products that they do not need or use. (This point is supported by conversations
with Citibank, American Stores, US Steel Group, Morgan Stanley/Dean Witter, GE
Supply, Playboy, and Boeing, which, as I discuss below, have all removed Internet
Explorer in one form or another from the PCs they deploy. Even organizations such as
American Stores and GE Supply, which moved away or are moving away from Navigator
to standardize on Internet Explorer, appreciate the ability to completely remove Navigator
from existing PCs and from the "standard build." Organizations particularly value the
removability of software products that could create issues of testing, training, ongoing
user support, security, compatibility, and/or computer resource requirements, if not
removed, as is the case with a Web browser. (Organizations may not generally remove
software products or components that create none, or few, of these issues.)
Page 42
a. For example, organizations that do not want Internet Explorer have
accomplished its removal from Windows 95 in various ways, most common
of
which
is removing end-user access (e.g., desktop icon, menu entries, executable
file
(iexplore.exe), etc.). Motorola, Informix, GE Supply, and Morgan Stanley/
Dean
Witter, for example, have taken this approach. With certain versions of
Internet Explorer, Microsoft has provided technical notes on how to
accomplish
removal of such access. In response to perceived corporate demand,
computer
companies such as Dell and Packard Bell/NEC have removed the Internet
Explorer
icon
and other end-user access to Internet Explorer from PCs they sell with
Windows 95. (J. Kies Deposition, 9/11/98, 24:1-10.) This approach, however,
has
costs
associated with it.
b. Another possible approach is to install the original retail version of
Windows 95, which includes no Internet Explorer software, as Boeing has
done.
(S.
Vesey Deposition, 34:3-13.) However, in that case, the organization would
either
forego or incur additional costs of trying to reassemble various technological
advances (such as the more efficient FAT32 file system) provided by later
versions
of
Windows 95.
D. The forced inclusion of Internet Explorer with
Windows 98 or Windows
NT 5.0 forces organizations to either forego new technology, incur
the
costs of supporting two browsers or removing the unwanted one, or
alter their choice of browsers.
Page 43
41. The inclusion of Internet Explorer with Windows 98 in such a way that
Internet Explorer cannot easily be removed can cause organizations who do not want
Internet Explorer to question the wisdom of adopting Windows 98. Indeed, some
companies are resisting, or electing not to use, Windows 98 largely or in part because it
would force them to have a two-browser desktop (for example, Chrysler, where the
Manager of Performance and Cost Management stated that two browsers would increase
support costs). These companies' decisions to forego Windows 98 is indicative of some
of the disadvantages of the commingling of Internet Explorer and Windows discussed
above.
42. However, despite such drawbacks, many customers (depending on their
size
or profile) feel strong pressure to use Windows 98 for various reasons, including the
following:
a. The organization's customers, suppliers, or clients are likely to use
it, and
business reasons exist to use the same software that customers, suppliers, or
clients
use.
b. Windows 98 brings new technological benefits, such as a more
efficient file system; support for new types of hardware, such as
high-capacity
optical
discs; support for specific new hardware devices, such as printers and
network cards; better control of power-saving features on both desktop
and
portable PCs; Year 2000 compliance; and a number of new and enhanced
"housekeeping" utilities that Microsoft claims (and I concur) can reduce
support
Page 44
costs.
(Microsoft's OSR2 version of Windows 95 brings some of the same
benefits, but, unlike Windows 98, has not generally been available to customers
as
an
upgrade for existing PCs.)
c. The organization relies on hardware that is being discontinued by the
manufacturer and replaced by new hardware that doesn't work with Windows
95
but
that does work with Windows 98. (For example, Boeing experienced this
situation with a network card used in portable computers. I have also often
heard
similar
concerns from seminar students and consulting clients with regard to the
transition from Windows 3.1 to Windows 95.) At some point in the life cycle
of
an
outdated operating system, computer hardware manufacturers tend not to
devote
resources towards making their newest products compatible with that
outdated operating system. Considering that manufacturers also tend to retire
older
products more rapidly than in other industries, one can see how hardware
issues
can
become a persuasive inducement to upgrade an operating system – or even a
requirement to do so, once older hardware products start breaking and must
be
replaced by similar models in current production that demand the newer
operating
system.
43. If an organization feels pressure to use Windows 98 for one or more of the
possible reasons listed in the previous paragraph, and given that (a) Windows 98 includes
various end-user-accessible means to run Internet Explorer, (b) Windows 98 does not
provide a vehicle to remove those means, (c) Windows 98 does not provide a vehicle for
Page 45
preventing the installation of, or removing after the fact, the Internet Explorer browser
software (as opposed to simply the means of access to the software), (d) Microsoft
designed certain features in Windows 98 to use browser software specific to Internet
Explorer, and (e) Microsoft designed Windows 98 so that it does not always honor the
default browser choice (all points on which I understand that Dr. Felten is testifying at
greater length), then the organization must either use Internet Explorer as its standard
browser, or incur various costs by using a competitive browser as its standard browser,
due to the fact that the organization now has two browsers on the desktop. Therefore, that
organization's browser decision is influenced by factors other than the merits of the
product.
44. Particularly with the release of Windows 98 and the impending release of
Windows NT 5.0, some companies are considering changing the standard company
browser to IE primarily or in part in order to avoid the costs of having two browsers on
users' PCs.
Boeing is a case in point:
The main reasons for moving to Internet Explorer 5.0
in Q2/3 of 1999 of the 18-month tactical plan are: We
do not have a choice. . . . The integration between
Internet Explorer and the desktop operating system
cannot be fully disabled. . . . Our only choice is
whether we will install two browsers or just install
Internet Explorer. (Exhibit 637, 18-Month Tactical
Plan, 7/14/98, TBC 000409-411.)
American Airlines is moving to
Windows 98 and at the same time switching from
Netscape Navigator to Internet Explorer, in part because it sees some
potential
Page 46
benefit
to the integration of the browser, but also in part because it recognizes that
it has
no choice but to accept this integration, which it recognizes cannot be
undone.
At John Deere, the Manager of
Technology Integration commented that although
Navigator is presently the standard browser, the company would probably
switch
to
Internet Explorer when it moves to NT 5.0, because he expects Internet
Explorer to be bundled with NT 5.0 and because the company only wants
one
browser.
45. Early indications from Microsoft and the trade press are that Windows NT
5.0 will integrate Internet Explorer in much the same manner that Windows 98 does. (See
Exhibit 630, Microsoft website titled "Windows NT Workstation 5.0 Beta 1,"
www.microsoft.com/workstation/basics/ntw5/ ntw5overview.asp ("Windows NT
Workstation 5.0 and Windows 98 will continue to have the same Web integrated
graphical user interface.").) It is reasonable to infer that the organizations that express
concern today about Internet Explorer's non-removability from Windows 98 are likely to
express the same concern in the future about Internet Explorer's non-removability from
Windows NT 5.0. Organizations that do not presently concern themselves with the non-
removability issue because they plan to bypass Windows 98 and wait for NT 5.0 may
well become concerned with the issue when they begin evaluating NT 5.0 in earnest.
VII. CONCLUSION
46. Organizations continue to view Internet Explorer and Windows as distinct
products -- one an application, the other an operating system. Many organizations have
sought to keep these products separate by removing Internet Explorer from Windows 95,
Page 47
or by standardizing on an Internet Explorer-free version of Windows 95. Both
mechanisms impose costs on the organizations that undertake them. However, although
some organizations are willing and prepared to incur such costs, in Windows 98 (and
apparently in Windows NT 5.0), these choices are not readily available.
47. It is clear that the commingling of Internet Explorer and Windows
provides
few real-world benefits, and several significant real-world costs and risks, for corporate
customers that do not standardize on Internet Explorer. Microsoft's failure to provide
a
mechanism for disentangling Internet Explorer from Windows 98 may compel such
customers to incur these costs and risks.
.
APPENDIX A - OPERATING SYSTEMS DEPLOYED
BY INTERVIEWEES
| ORGANIZATION | NUMBER
OF DESKTOP PCs4 (APPROXIMATE) | DESKTOP OPERATING SYSTEMS
DEPLOYED |
|---|
| American
Stores | 4,000-5,000 | Most Windows 95, some Windows 3.1 | | Boeing | 180,000-200,000 [GLENN - THIS # IS FROM BOEING
SUBPOENA RESPONSE] | Approximately 80% Windows 95; 15% UNIX; 5%
Windows NT | | Chrysler | 24,000 to
29,000 | Largely Windows 95 | | Citibank | 102,000 | Approximately two-thirds Windows 3.1;
one-third Windows NT 4.0 Workstation; small number Windows 95 | | ConAgra | 7,000 | All Windows 95 or Windows
3.1 | | Federal Express | 32,000 | Standardized on
Windows 95; some Windows 3.1 and some Windows 98. Also HP-UX servers. | | Florida Department of Revenue | 4700 | Windows 95,
Windows NT Workstation, Windows 3.1, MS-DOS. | | Ford | 115,000 | Mostly Windows 95 | | GE Supply | 2,200 | Windows 95 | | Informix | 38,000 | Largely Windows 95; approximately 2000
UNIX | | J.C. Penney | 20,000 | 45% Windows
NT; 30% Windows 95; remainder other operating systems | | John
Deere | 25,000 | [not reported] | | Liberty
Corporation | 2,500 | Windows NT and Windows 95 | | Morgan Stanley/Dean Witter | 14,000 | Windows NT;
Windows 95; some Sun Solaris | | Motorola | 250 in unit
interviewed | Windows 95 | | Playboy | 475
Intel-based; 125 Apple | Windows 95; Macintosh | | Sabre
Group | 12,000 | Windows 95; Windows 3.1; Wndows NT | | US Steel Group | 7,000 | Largely Windows 95; Windows
NT |
FOOTNOTES
1 The pool of
interviewees was selected by the DOJ from the following sources: (1) organizations described by
Dell Computer Corporation as having requested the removal of all or part of Internet Explorer
from Dell personal computers; (2) organizations which had voluntarily expressed interest in the
DOJ's investigation and this litigation; and (3) a random sampling of Fortune 100 companies. To
ensure that my questions and analysis was not biased by the manner in which any particular
interviewee was selected, DOJ has not informed me which interviewee has been selected from
which category. I have not relied solely on these interviews for any conclusion or opinion stated
in this testimony. All of the points made by interviewees have been corroborated by all or some
of my conversations over the years with corporate technical support personnel; Microsoft
documents; my personal experience with Windows and Internet Explorer; and the deposition
testimony of other witnesses in this litigation.
2 To the extent that
organizations may assess operating system and application needs by reference to factors irrelevant
to individual computer users, the aspects of my testimony that focus on these factors pertains only
to the former category. For example, while both classes of user need operating system software
to run any computer, organizations often require the ability to restrict the functions that novice
users can perform to prevent data loss from a novice's mistake. Organizations further often
require the ability to restrict the functions that any users can perform, in order to
safeguard data, maintain security, and increase productivity. Organizations also value the ability
to standardize their operating system software and, for that matter, their application software, for
easier testing, troubleshooting, training, and support. Such organizational concerns typically do
not matter as much, if at all, to home computer users. This said, many of the principles relevant
to organizations' selection of operating system and application products, including browsers,
certainly apply to home users.
3 Of the organizations
whose managers I interviewed (or whose comments were summarized for me), most utilize
Microsoft's Windows operating system products for the large bulk of their employees' desktop
computing needs. The information I received on this score is summarized in Appendix A to my
testimony.
4 As used here, the
term desktop PCs encompasses both general purpose PCs (i.e., those used for office use,
including word processing, spreadsheet, and other productivity applications) and "workstation"
PCs (i.e., those used for more high-powered engineering or scientific applications).
|