competitive harm are relatively infrequent, and the efficiency gains from tying have often been ignored.4
Thus while tying contracts have been subject to a form of per se review, skepticism about this per se
treatment has been longstanding.
One consequence of this skepticism has been that the “per se” nature of the tying rule is relatively
unique.
5Rather than condemn all tying contracts, the Supreme Court has limited the proscription to
those contracts where the seller possesses market power in the tying market, and substantial commerce in
the tied market is affected. Eastman Kodak, supra, 504 U.S., at 461-62.
A significant minority on the Court would go further. As articulated in her concurrence in Jeffer-
son Parish, 466 U.S., at 32, Justice O’Connor and three other justices have indicated that they would jet-
tison the per se rule, and condemn tying contracts only after applying the rule of reason.6
There is good reason to believe that this skepticism will be relevant in the context of alleged ties
of software products. Especially in a newly emerging market, it is sensible to hesitate before condemning
a market practice, at least until it has been shown convincingly that that practice is anticompetitive.7
Historically, antitrust law seems to have gone the other way round — condemning the practice first, and
only later coming to see that much of what it condemns is in fact competitively benign. But the hesitation
of the Court of Appeals, and the importance of the software market, will lead justices of the Supreme
Court to think carefully about how the law of tying should be extended to software products.
4 See, e.g., Fortner Enters. v. United States Steel Corp., 394 U.S. 495, 514 n.9 (1969) (White, J., dissenting) (procom-
petitive justifications); IX PHILLIP E. AREEDA, ANTITRUST LAW ¶1730 (1991); ROBERT BORK, THE ANTITRUST
PARADOX 380-81 (1978); HERBERT HOVENKAMP, FEDERAL ANTITRUST POLICY 403-04 (1999); RICHARD A.
POSNER, ANTITRUST LAW 171-84 (1976) (“prohibition against ties ought to be radically curtailed”); RICHARD A.
POSNER & FRANK H. EASTERBROOK, ANTITRUST 809-10 (2d ed. 1981); Frank H. Easterbrook, Vertical Arrange-
ments and The Rule of Reason, 53 ANTITRUST L. J. 135, 143-46 (1984); Benjamin Klein and Lester Saft, The Law
and Economics of Franchise Tying Contracts, 28 J. L. & ECON. 345 (1985); Alan Meese, Tying Meets the New Institu-
tional Economics, 146 U. PENN. L. REV. 1, 59-86 (1997); Keith Wollenberg, Note, An Economic Analysis of Tie-in
Sales: Re-examining the Leverage Theory, 39 STAN. L. REV. 737 (1987).
Nonetheless, the Supreme Court has indicated that it is “far too late in the history of our antitrust
jurisprudence to question the proposition that certain tying arrangements pose an unacceptable risk of
stifling competition and therefore are unreasonable ‘per se.’” Jefferson Parish, 466 U.S., at 9. And it is of
course not appropriate for this Court to decide this case by second guessing or predicting the Supreme
Court’s jurisprudence, at least with respect to a statutory matter. See, e.g., Maislin Industriesv. Primary
Steel, 497 U.S. 116, 130 (1990). It would nonetheless be prudent for this Court to decide the question
about tying software products both under the existing per se analysis and under an alternative analysis. In
the context of tying generally, and the tying of software products in particular, it is my view that there is a
significant probability that the Supreme Court will modify current doctrine.
The prevailing Supreme Court authority on the law of tying is Jefferson Parish, relied upon by the
Court in Eastman Kodak. As described in Eastman Kodak,
A tying arrangement is “an agreement by a party to sell one product but
only on the condition that the buyer also purchases a different (or tied)
product.” Such an arrangement violates §1 of the Sherman Act if the
seller has “appreciable economic power” in the tying product market, and
if the arrangement affects a substantial volume of commerce in the tied
market.
504 U.S., at 461-62 (citations omitted). In this case, Windows 95/98 is the alleged “tying product”; In-
ternet Explorer is the alleged “tied product.” The only question that I will address is whether the func-
tionalities associated with Internet Explorer should be considered a product separate from Windows 95
ASSUMING
MICROSOFT
II CONTROLS
The resolution of these questions turns on the status of the Court of Appeals’ decision in Micro-
soft II. That case arose as an appeal from a preliminary injunction by this Court to enforce the terms of a
consent decree entered into by the parties in 1994. Microsoft challenged that injunction on a number of
grounds, one being the failure to provide adequate notice. Microsoft II, 147 F.3d, at 943. The Court of
Appeals agreed, and proceeded to decide whether the government was likely to prevail on remand. This
required the court to interpret the 1994 consent decree. It is the Court of Appeals’ interpretation of that
decree that raises the ambiguity in this case.
At the core of the dispute was the interpretation of Section IV(E)(i) of the 1994 consent decree.
That section provided that:
Microsoft shall not enter into any License Agreement in which the terms
of that agreement are expressly or impliedly conditioned upon: the li-
censing of any other Covered Product, Operating System Software prod-
uct, or other product (provided, however, that this provision in and of
itself shall not be construed to prohibit Microsoft from developing inte-
grated products)….
147 F.3d, at 939. Windows 95 was a “covered product.” The government alleged that Microsoft condi-
tioned its license for Windows 95 on an OEM’s licensing of an “other product” — Internet Explorer. Mi-
crosoft argued that Internet Explorer and Windows 95 were “integrated products” and thus, because of
the proviso, exempt from the scope of the decree. 147 F.3d, at 946-48.8
The parties to the 1994 decree did not define “integrated products.” Nor was the intended pur-
pose of the proviso readily apparent. In the proceedings before me during my brief time as special master
in that case, Microsoft did offer a fair reading, in my view, of the purpose of that clause. As Microsoft
Section IV(E)(i) prohibits the tying of all three categories of products
unless the claim of tying is based on the creation of an integrated product
of which a[n] …“other product” is a component: Although the develop-
ment (read: creation) of such a product could be conceptualized as a tie,
what the proviso tells the reader is that it is not to be construed that way
for purposes of the Consent Decree—because the parties agreed that any
claims of this nature would only be addressed in a plenary antitrust action un-
der the Sherman Act.
Letter of January 21, 1998, from Richard Urowsky to Lawrence Lessig, 2 (final emphasis added). This
understanding is consistent with the government’s interpretation of the decree in the 1994 Tunney Act
proceedings. In response to a filed comment, the government stated that
Activity of this sort [i.e., integration] requires case by case analysis, and a
broad injunction against such behavior generally would not be consistent
with the public interest.
8 The proviso exempts “developing integrated products.” The Court, and the parties, assumed that “developing” was
the same as “developing and licensing,” so that if the proviso exempted developing, it also exempted licensing. There
is some authority in the European law context, within which the decree was drafted, for distinguishing the two, and
reading developing narrowly, see R&D Block Exemption Regulation No.418/85 [1985] O.J. L53/5 as amended on
Regulation 151/93 [1993] O.J. L21/8, but the distinction would not matter for the purposes of this case. (I am
grateful to Professor Steve Anderman for this point.)
United States v. Microsoft Corporation,Civ. No. 94-1564, Response of the United States to Public Com-
REG59426, 59428 (1994). Accordingly, rather than resolving claims about integrated
products in the context of the consent decree, both Microsoft and the government indicated that those
claims would be dealt with on a “case by case basis.” To the extent the decree was regulating “tying,” it
was intended, under this reading, to regulate ordinary contract-tying. It was not intended to regulate “in-
tegrated products” which “could be conceptualized as a tie.”
Thus the proviso was jurisdictional — allocating to the consent decree one class of tying claims,
and reserving to a “plenary antitrust action” a different class of tying claims. This difference makes perfect
sense of the policies at stake. Consent decrees enable expedited proceedings. They should only regulate
behavior that is unambiguously problematic. Ambiguously problematic behavior (sometimes benign,
sometimes not) should not be regulated by an expedited proceeding. Such behavior is best left to a context
where the nature of the behavior and any anticompetitive threat can be fully explored. The government in
the Tunney Act comments indicated that bundling through integration was ambiguously problematic. Id.
It neither indicated that all such bundling was benign; nor that no such bundling was benign. Rather it
said such behavior had to be evaluated on a “case-by-case basis.” Where the behavior was ambiguous,
then, the decree left it to a “plenary antitrust action.”
This background is necessary to understand the meaning of Microsoft II. The government, as well
as many commentators and at least one court, have criticized the Court of Appeals for a establishing an
overly restrictive interpretation of tying law’s application to software tying claims.9Those criticisms make
sense, however, only if one reads the Court of Appeals to be establishing the law of software tying for the
Circuit. But in light of the interpretation of the decree offered by Microsoft, and the understanding of the
proviso offered by the government, there is no reason to interpret the opinion in Microsoft IIas anything
more than an interpretation of this jurisdictional provision. For perfectly sensible reasons, the Court, on
9 See, e.g., Pl. Conclusions of Law, at 57-58 [Pl. CL.]; Michael Woodrow De Vries, United States v. Microsoft, 14
BERKELEY TECH. L.J. 303, 314-22 (1999); Norman W. Hawker, Consistently Wrong: The Single Product Issue and
the Tying Claims Against Microsoft, 35 CA. W. L. REV. 1 (1998) (applying Jefferson Parish analysis to software
claims); John E. Lopatka & William H. Page, Antitrust on Internet Time, 7 SUP. CT. ECON. REV. 157 (1999); Wil-
this view, gave the term “integrated products” a very broad meaning, thereby assuring that any claims
about “integrated products” will be adjudicated in a “plenary antitrust action.”
It would not follow on this reading of Microsoft II, however, that the Court intended to say any-
thing about how the “single product” test of Jefferson Parishshould be interpreted in the context of a “ple-
nary antitrust action.” Indeed, the Court was quite explicit that “[t]he antitrust question is … distinct.
The parties agree that the consent decree does not bar a challenge under the Sherman Act.” 147 F.3d, at
950 n.14. Thus, rather than offering a binding interpretation of Jefferson Parishas applied to software
products — an act that would have been clearly beyond the Court’s proper judicial role, given the posture
of the case — the Court said that its “task … [was] to discern the bargain that the parties struck.” 147
F.3d, at 946. Invoking well established Circuit precedent, the Court wrote that while the consent decree
was to be “consistent” with antitrust law, “an antitrust consent decree cannot be read as though its ani-
mating spirit were solely the antitrust laws.”10147 F.3d, at 946.
in which the decree could be “consistent” with antitrust law, and hence the interpretation of a decree
“consistent” with antitrust law, even though the scope of the “integrated products” proviso would not be
coextensive with the “single product” test of Jefferson Parish. A consent decree is a bargain between the
federal prosecutor and a potential defendant, which for reasons of separation of powers need not map an-
titrust law generally. Id.at 59-60. A defendant does not necessarily concede that the behavior regulated by
a decree is illegal; nor does the government, by implication, concede that all behavior not regulated by the
decree is legal. The decree must be “consistent” with antitrust law in the sense that it cannot exceed the
government’s prosecutorial power. But the government’s prosecutorial power, under constitutional princi-
10 Microsoft suggests that the Court of Appeals indicated it was deciding the case under tying law because, as Mi-
crosoft writes, the Court “referred to the provision of the Consent Decree at issue in that case as an ‘anti-tying pro-
vision.’” Df. CL., at 3. Actually, what the Court said was that “[b]oth Microsoft and the Department characterize §
IV(E) as an ‘anti-tying’ provision.” Microsoft II, 147 F.3d, at 946. In the next paragraph, the Court pointed out that
“Microsoft is clearly right that the decree does not embody either the entirety of the Sherman Act or even all ‘tying’
law under the Act.” Id. Thus the Court’s characterization of Microsoft’s position does not seem to support the con-
clusion that the Court was interpreting tying law generally.
ples of executive authority, is not the same as a court’s interpretation of the Sherman Act. There is no
authority in the D.C. Circuit for the proposition that a consent decree can be read to imply the contours
of antitrust law, and no reason to read an opinion interpreting a consent decree as interpreting the con-
tours of antitrust law either.11
It is of course this Court’s job to interpret the scope of Microsoft II. But when there are two un-
derstandings of a higher court’s opinion — one that reads it narrowly, as properly applying to the issues
raised before it, and sensibly resolving the claims properly presented, and another that reads it broadly, as
improperly reaching issues that were not before the court, and adopting a controversial interpretation of
federal law that was not even briefed by the parties — charity in interpretation at least should lead a lower
court to adopt the first rather than the second reading of that opinion. This seems especially true where
the text of the opinion itself expressly indicates its limited reach. 147 F.3d., at 950 n.14.
Nonetheless, a substantial question has been raised by the parties about whether the “framework”
used by the Court of Appeals to interpret the word “integrated” in the context of that decree should also
be used to decide whether Microsoft’s “browser functionality” is a “separate product” for purposes of anti-
trust law. Microsoft reads the Court of Appeals’ decision to so indicate, Microsoft Conclusions of Law at
6 [Df. CL.], and this Court’s opinion in the summary judgment stage of this case seemed to indicate
likewise. United Statesv. Microsoft Corp., 1998 WL 614485 (D.D.C. 1998), at *10-13. The Court of Ap-
peals’ opinion does not speak exclusively of “integrated products” — the language of the consent decree.
In a number of places, it uses “integrated products” and “single product” — language drawn from antitrust
tying law — interchangeably. E.g., 147 F.3d, at 946. It is thus a reasonable prediction that the Supreme
Court would modify existing tying law to fit its proposed standard for determining whether software
products should be considered “separate products” or a “single product.” For these reasons, it is certainly
possible for this Court to view itself bound by the Court of Appeals’ opinion, in spirit if not in form.
11 Microsoft argues, “[t]he Court of Appeals thus held that an integrated product constitutes a single product for
purposes of antitrust law so long as there are ‘facially plausible benefits to [the] integrated design.’” Microsoft Con-
clusions of Law [Df. CL.], at 6. But I am unsure how the Court of Appeals could “hold” as to a matter that was not
before it; nor how it could “hold” anything with respect to “antitrust law” when it had stated expressly that the issues
were “distinct.”
Therefore, for the purposes of this part of my analysis, I will assume that Microsoft IIset the standard for
tying law as it applies to software products generally.
The Court of Appeals said that two software products were “integrated” for purposes of the con-
sent decree if it was more efficient for them to be combined by Microsoft than for them to be combined
by OEMs or the consumer. The test had two parts. As the Court wrote,
We think that an “integrated product” is most reasonably understood as a
product that combines functionalities (which may also be marketed
separately and operated together) in a way that offers advantages unavail-
able if the functionalities are bought separately and combined by the pur-
chaser.
147 F.3d, at 947. And later,
[T]he combination offered by the manufacturer must be different from
what the purchaser could create from the separate products on his own.
Id. at 949. This, however, left open the question of when the relevant “combination” occurs — whether at
the design or installation stage. If it was the installation stage, then, as this Court has found, ¶190, at least
with some versions of Windows 95, there would be essentially no difference between the “advantages
[]available if the functionalities [were] bought separately and combined by the purchaser,” and the ad-
vantages available if the combination were made by Microsoft. At that stage, the code from both, and
hence functionality of both, would be the same.
The Court of Appeals, however, interpreted the relevant moment of “combination” to be at the
design stage of Windows’ development, not the installation stage. The relevant “combination,” the Court
wrote, “is the creation of the design that knits the two together” rather than the stage where the two
products are installed. “Consequently,” the Court concluded, “it seems clear that there is a reason why the
integration must take place at Microsoft’s level.” 147 F.3d, at 952.
That was the first part of the test for determining whether two products are “integrated.” The
second part required that the integration actually do some good. As the Court explained, “there should be
some technological value to integration.” 147 F.3d, at 949.
The concept of integration should exclude a case where the manufacturer
has done nothing more than to metaphorically “bolt” two products to-
gether, as would be true if Windows 95 were artificially rigged to crash if
IEXPLORER.EXE were deleted.
147 F.3d, at 949. The court was quick to note, however, that “do some good” does not mean “a net plus.”
147 F.3d, at 950. The review, the Court instructed, “must be narrow and deferential.” Id. Rather than
asking whether the integration was efficient, or a good idea, or sold more operating systems, the Court
said the “question is … whether there is a plausible claim that [the integration] brings some advantages.”
While the Court relied in its opinion upon the analysis of the Areeda Treatise, its rule is actually
more forgiving than the rule announced in the treatise.12The question the Areeda Treatise asks, after the
plaintiff establishes that two products are “separable,” is whether the products could bedesigned such that
the consumer could combine them just as well as the defendant. X PHILLIP
E. AREEDA, EINER
HOVENKAMP, ANTITRUST
LAW¶1743, at 192-94 (1996) [AREEDA
TREATISE]. If there is a reason why the consumer can’t bundle them as well as the defendant — for ex-
ample, if they are too cumbersome, or if the combination would need to occur at an early stage in the
product’s production — then the “New Product Rationale” that the Court relied upon would treat the
two “separable” items as a “single product” for purposes of antitrust tying law. But if there was no reason
why the product couldn’t be designed to allow the consumer the choice of whether to install the product
or not, while keeping the same efficiencies, then the product would be considered “two products” for pur-
poses of antitrust tying law.
The standard in Microsoft IIis different. The Court does not follow this two step analysis, asking
first whether products are “separable” and second, whether there is some good reason that they should,
nonetheless, not be separated. Instead, the court simply asks, consistent with the language of the consent
decree, are they “integrated”? Given the product as designed, is there a reason why the defendantshould
bundle rather than the consumer? The question the Court addressed was whether they werearchitected to
be separate products, not whether they could be.
This fact about the Court of Appeals’ approach is made clear in the examples it considers. For ex-
ample, the Court noted that “IE 4 technologies are used to upgrade some aspects of the operating system
unrelated to Web browsing,” and these features “cannot be included without also including browsing
functionality.” 147 F.3d, at 951. In saying these features “cannot be included,” the Court was not opining
about how software could be designed. It certainly understood that Microsoft could have designed its
products differently, to permit the IE 4.0 technologies to provide browsing functionality only, while pro-
viding users with a separate upgrade package to upgrade operating system functionality. See, e.g., 147
F.3d, at 951. But the Court was not holding Microsoft to the test of what could be designed. The court
expressly directed that it was not proposing that “in making this inquiry the court should embark on
product design assessment.” 147 F.3d, at 949. Because of their “limited competence” courts were not to
“second-guess[] the claimed benefitsof a particular design decision.” 147 F.3d, at 950 n.13 (emphasis
added). The perspective the Court adopted was to evaluate Windows 95/IE as they were designednot as
they could have been designed.
The standard in Microsoft IIis different from the test relied upon in the Areeda Treatise in a sec-
ond way as well. In making its comparison, the baseline the Court adopted was whether the design by
Microsoft provided “plausible benefits [] as compared to an operating system combined with a stand-
alone browser such as Netscape’s Navigator.” 147 F.3d, at 950. The issue was no longer whether consum-
ers could combine the stand-alone products (such as IE 4 and Windows 95) and produce the same func-
tionality; the Court had determined that the relevant “combination” was at the design stage. Indeed, the
Court expressly rejected the idea that “in installing IE 4 an OEM is combining two stand-alone prod-
ucts.” 147 F.3d, at 951. “[T]he products,” the Court instructed, “do not exist separately.” 147 F.3d, at
951-52. Thus the relevant comparison was not whether Windows 98 provides “some benefit” in function-
ality as compared with Windows 95 plus IE 5.0; the question was whether its benefit compared with “a
stand-alone browser such as Netscape’s Navigator.” 147 F.3d, at 950.
12 So, again, if one interprets the Court’s opinion as applying antitrust law generally, then its interpretation of the
Areeda Treatise at least was mistaken. See Einer Elhauge, The Court Failed My Test, Washington Times, A19, July
10, 1998. Professor Elhauge did not address Windows 98.
But, again, this is not the methodology of the Areeda test that the Court purported to follow, if
the court was interpreting tying law generally. The question the Areeda standard asks is whether combi-
nation by the consumer could be as good as the combination by the defendant, not whether the combina-
tion by the defendant is as good as a combination with another product. See AREEDA
TREATISE, ¶1746
at 229. Thus again, if this Court read the Court of Appeals to be applying the test that the Areeda Trea-
tise summarizes in order to explicate tying law generally, then the Court of Appeals was mistaken.
Between concluding (1) that the Court of Appeals was articulating a new standard for antitrust
tying law, but was mistaken in interpreting the doctrine that it purported to rely upon, and (2) concluding
that the Court of Appeals was interpreting “integrated products” in the context of a consent decree, using
analogous principles from a related area of law, the second would seem a fairer reading for a District
Court to adopt than the first. But if the Court believes that though mistaken in its interpretation of the
test it was relying upon, the Court of Appeals was setting the standard for antitrust tying law generally,
then the methodology of the Court of Appeals is in my view dispositive of the government’s Section 1
tying claims in this case, at least as they relate to tying achieved through technological design.13
If the relevant “combination” under Microsoft IIis at the design stage, then there can be no dis-
pute about whether Windows 95/98 and IE pass the first requirement of the Court of Appeals’ test —
that they are better combined by Microsoft because only Microsoft can design the code. This is true by
definition under the test proposed by the Court.
So too is it clear that the combination passes the second requirement of the Court of Appeals’ test
—the question whether the integration does “some good,” or whether there is a “plausible claim” that the
integration “brings some advantages” relative to a “stand-alone browser such as Netscape Navigator.” 147
F.3d, at 950. While this Court has found instances of the design of Windows 95/98 that serve no effi-
ciency function — indeed instances that hamper the efficiency of Windows 95/98, ¶¶ 173,174 — the
13 I am therefore not addressing the question whether IE 1.0 and IE 2.0 were improperly “tied” to Windows 95; that
bundle was effected through contract alone. ¶155. Nor am I addressing the question whether the vertical restraint
on OEMs not to run the Add/Remove program constitutes an improper restraint under the rule of reason. See Alan
![[made with GoClick]](imgs/goclick.gif)
