signed to be. For while software is plastic, as the evidence in this case suggests, it is too simple to assume
(as the treatise does) that any combination is possible. It might be that an update to a program is so sig-
nificant, and the range of different installations of a program so vast, that it would be more efficient to
perform the update at the factory, so to speak, rather than permitting the consumer the choice. One can’t
know in the abstract. So to evaluate whether this is true, a court would have to investigate the character of
the two products — to examine, as this Court has, the nature of the interaction between the two prod-
ucts, and the character of the process that combines, or might combine, the two products. But to require
courts to engage in that inquiry would defeat the rule as a “screening” device. To apply ¶1746b would re-
quire an extensive factual inquiry into what “could be” done, contrary to the purpose of ¶1746b and con-
sideration (3) above that the inquiry be relatively simple.
These considerations suggest that at least the second disjunct of the “New Product Rationale” is
not usefully applied in the context of software products. To the extent that it turns on the technology of
bundling, it either requires a judicial review that is too extensive, or it creates an incentive for designers to
architect their software to fit an antitrust rule. Both results counsel against embracing this “special reason”
for treating two products as a “single product.”
The second “special reason” outlined by ¶1746 of the Areeda Treatise for treating two items as a
“single product” under the “New Product Rationale” (¶1746a) does not turn upon how the two products
are or could be bundled. Here the question is whether “no relevant analogue of bundling or unbundling
exists because the defendant’s bundle causes the items to operate together in a way that had not been tried
not a “single product” but was instead an impermissible tie of MS-DOS and Windows 3.x. 1999 SUPPLEMENT,
¶1746.1, at 489. The same claim was recently made against Microsoft in the Calderalitigation. Caldera, supra. In
that case, because there was a dispute about whether in fact there was technological integration between the then
current versions of Windows and MS-DOS, the District Court permitted the issue to go to trial. According to
deposition testimony, and contrary to the assumption of the Court of Appeals in Microsoft II, in fact there was no
“shared code” between the two products. But in both that case, and under the consent decree case, this analysis
misses the first step of the Areeda test — whether it was a competitive market practice to bundle a GUI interface
with the operating system. Under that test, it would seem the dominant and emerging practice of other OS firms
was to bundle the GUI interface (Macintosh, and OS/2), and that the same bundle by Microsoft should be consid-
ered a “single product” — again, whether or not technologically the package shared code or was “truly integrated.”
tronics, 187 F. Supp. 545 (E.D. Pa 1960), relied upon by the Supreme Court in Jefferson Parish, where
Judge Van Dusen concluded that in a nascent market, the alleged bundling was justified, but at an un-
specified point later in the course of the case, it was no longer justified. Id. at 560. The aim of the test is
to protect this innovative new design, at least so long as it is new.32
The applicability of this test to the facts of this case might be questioned, because obviously,
browsers and operating systems have been functioning together for almost a decade. But as Microsoft
claims, Microsoft has bundled its browser technology in a way that has not been tried before — for
browsers, at least. Specifically, by choosing to componentize its browser functionality, and thereby expose
a larger range of APIs to other applications, Microsoft has caused its browser product to operate with its
operating system product in a “new way.” Thus to the extent there is a relevant “New Product Rationale”
for software products, it would appear that the analysis of ¶1746a is the better aspect of that rationale to
consider Windows 95/98 under.
But in the context of software products, this disjunct suffers from a flaw that is analogous to the
flaw with ¶1746b as applied to software. For again, because of the character of software, it is always pos-
sible to bundle software such that it operates in a “new way.” AREEDA
TREATISE, ¶1746a; 1999
SUPPLEMENT, at 494. This consideration led the Court of Appeals to look to other factors that might
limit the scope of its interpretation of “integrated products” — ways, that is, of distinguishing integrated
products that would reflect genuine improvements from new products that were merely pretextual. The
exclusion of mere “bolting” and the embrace of designs that bring “some good” are two ways to filter new
product designs to assure that the test does not immunize bundles too broadly.
But these limitations just lead the court down a path that has no obvious stopping point. Critics
of the Court of Appeals’ standard believe it is not extensive enough to capture strategic bundling behavior.
Other Courts have therefore suggested a more extensive inquiry into the character of the “integration.”
32Microsoft argues that “[n]o court has ever sustained such a challenge to a single integrated product.” Df. CL., at 2.
That is not completely correct. In Jerrold Electronics, while the court held that a television transmission system was
properly considered a “single-product” during its initial period of operation, it did sustain a challenge of that same
“single-product” after the introductory period passed. 187 F. Supp., at 560.
Caldera, supra, at 1319. And to the extent that the inquiry becomes more invasive, the less the test looks
like a “screening” test, and the more it looks like a full blown analysis into the merits of the design.
As I described above, these considerations have led the author of the most recent supplement to
the Areeda Treatise to recommend the finding of a “single product” “in all cases in which the code for the
two programs is interspersed such that the purchaser cannot readily separate them.” 1999 SUPPLEMENT,
¶1746.1b, at 494. As I said, if the Court embraced this view, then the Court would conclude that Win-
dows 95/98 is a “single product” for purposes of Section 1 tying, and Microsoft would prevail.
In my view, however, the evidence as this Court has found it suggests a more straightforward ap-
proach to the question of new software products — one that does not lose “most of the advantages of i n-
tegration” or create the incentive for software producers to “comply with” the rule “by interspersing code.”
Id. But one that does account for the special factors about software that Microsoft advances.
Under this approach, two software products combined in a “new way” would be considered a
“single product” for purposes of antitrust tying law — regardless of how they were linked (whether by
code or by contract) — but this conclusion would be presumptive only. The presumption could be de-
feated if the plaintiff can show that the particular bundle at stake raises the risk of a particular anticom-
petitive harm. If, in other words, it is a bundle of the kind of products likely to cause an anticompetitive
harm, then the presumption finding a “single product” would be rebutted.
What the risks that would defeat this presumption are I will describe in the next section. But I
want first to highlight the reasons why a test different from the current version of the Areeda test seems
especially appropriate in light of the facts that this Court has found.
As the treatise acknowledges, the effect of the “New Product Rationale” will be to create an in-
centive for software producers to modify their code to fit within the terms of the rule. Id.This, in my
view, is a significant weakness of the rule. So long as there is a risk of strategic bundling, the law should
not create an incentive for the bundler to hide that bundle in code. As this Court has found, such pretex-
tual bundling is both inefficient and costly. ¶173, 174. The law has no reason to create the incentive for
Instead, the “New Product Rationale” should be neutral between “new combinations” that are ef-
fected by contract or effected by code. If there are two software products that could be combined to oper-
ate together “in a new way,” then so long as there is no risk of strategic bundling, the law should allow the
innovator to decide how the two products are more efficiently combined. The aim of any antitrust inquiry
should be whether the particular bundle is a strategic bundle, aiming at anticompetitive ends, not whether
the bundle achieves it interlinkage through contract or software.
The better solution is to keep the court’s focus above the hood, so to speak, and probe the risk of
anticompetitive bundling of software through different techniques. That is the aim of the modification to
the “New Product Rationale” that I believe this Court could adopt. Under this modification, software
manufacturers who show that they have bundled two software products together in a “new way” should
get the benefit of this “New Product Rationale” whether their bundle achieves this benefit through “tec h-
nological integration” or not. But the benefit of this rule — treating two products as one — should be
presumptive only, subject to being defeated by the other considerations that I describe below.
The consequence of this rule would be that software bundling generally would constitute a single
product, so long as the defendant can establish that the bundling operates together in a “new way.” This
would mean that in the general case, a software producer would be free to add functionality to its software
products, whether or not that inclusion meant other software producers would be disadvantaged. As the
history of this case shows, there are many examples of improvements to the operating system that have
eliminated the demand for other products in the software industry.
But this Court has implicitly accepted this consequence by dismissing the States’ Section 2 claim
of “monopoly leveraging” against Microsoft. United Statesv. Microsoft Corp., 1998 WL 614485 (D.D.C.
1998), at *27. As the Court said there, in the ordinary case, antitrust law has little reason to fear this form
of bundling, even if the consequence is that competitors are driven out of business. The general rule
should be a lenient one, to encourage innovation in the design of software products; the exception should
be where there is good reason to fear that the bundle will create anticompetitive harm.
Under this approach to the “New Product Rationale,” then, Microsoft’s operating system and
browser functionality should be treated presumptively as a “single product” for purposes of antitrust tying
law, unless an independent reason exists why this type of bundle raises special anticompetitive concerns. I
consider one such concern in the next section.
The third rationale relevant to this case is the “Same Product Rationale.” AREEDA
TREATISE, at
¶1747. Like the “New Product Rationale,” this test involves two inquiries, though only one directly sup-
ports the rationale. The core of the rationale recommends that if the defendant is bundling “complete
substitutes,” then the two products should be considered “one product” for purposes of tying law. But the
rationale is quick to make clear that if the defendant is bundling “partial substitutes,” then the “same
product rationale” does not apply. If the products are partial substitutes, then they should be considered
“two products” for purposes of tying law. AREEDA
TREATISE, at 232.
The reason for the difference lies in the differences in competitive harm likely to arise from the
bundle. “[A] tie of partial substitutes might leverage market power and spread inefficiency from one ‘mar-
AREEDA
TREATISE, ¶1747, at 232. Thus, while complete substitutes bundled together are typically not
harmful to the competitive process, “bundles of partial substitutes are moredangerous than the typical tie.”
AREEDA
TREATISE, at ¶1747. Bundles of partial substitutes, Professor Hovenkamp notes, can “sabotage
a nascent technology that might compete with the tying product but for its foreclosure from the market.”
1999 SUPPLEMENT, ¶1746.1d, at 495.
This insight into the dangers of “partial substitutes” provides a check on the scope of the “New
Product Rationale.” Based on the facts that this Court has found, in light of the Supreme Court’s injunc-
tion that the “single product” inquiry must be “based on whether there is a possibility that the economic
effect of the arrangement is that condemned by the rule against tying,” Jefferson Parish, 466 U.S., at 21,
this Court could condition the “New Product Rationale” as applied to software products upon whether
the product was a “partial substitute” for the tying product. If it is a “partial substitute” that the plaintiff
can show will likely have an anticompetitive impact, then the bundle should not be considered a “single
product” for purposes of antitrust tying.
This rationale has a direct application to the Court’s findings in this case. The theory of the gov-
ernment’s case, and the core of the findings that this Court has made, is that Microsoft exercised its
power to protect what has been termed the “applications barrier to entry.” ¶31, 36-44. By maintaining
control over the APIs that independent software vendors write to, Microsoft is in a position to maintain
the dependency the market has on its Windows operating system. ¶68.
One aspect of Microsoft’s campaign, this Court has found, has been to assure that the browser
technology that Windows users deploy is a browser technology that it controls. ¶227. In this way, it can
assure that the browser doesn’t become a competitor to the Windows API set, or at least that, to the ex-
tent the browser exposes APIs, they are APIs that Microsoft controls as well. ¶79. By bundling a browser
that itself is a partial substitute for the Windows platform, but which Microsoft nonetheless controls, Mi-
crosoft is able, this Court has found, to avoid losing developers to a foreign API set.
In this way, “browser technology” has the potential to be a partial substitute for the Windows op-
erating system. It is not itself a complete substitute — as this Court has found, browsers do not expose
anywhere close to the number of APIs that an operating system does. ¶77. But to the extent that they
expose any significant set of APIs, sufficient to induce developers to write applications to the browser
APIs, they threaten to reduce the market power of the underlying tying product — Windows.
Applying the rationale of the partial substitutes test to the facts of this case, the Court could
therefore conclude that the presumption of the New Product Rationale — that the browser product and
operating system should be considered a “single product” — has been rebutted, and the two software
products of an operating system and browser functionality should be considered as “two products” for
purposes of antitrust tying.
While application of this test, given the current design of Microsoft’s operating system, should
entail a finding of “separate products,” as a general matter software developers should be able to avoid the
risk of this “partial substitute” rationale simply by designing their software so as to give consumers an op-
tion not to take the partial substitute. Like the free but optional browsers discussed in the CMP test
above, if the product were easily removable there would be no antitrust forcing. This was apparently
IBM’s strategy with software bundling, see Innovation Data Processingv. IBM Corp., 585 F. Supp. 1470,
1476 (D. N.J. 1984) (two software products bundled together not considered tied where customer had the
option to remove one, and cancel the license for that component), though the court in that case did, in
dicta, suggest that that the removability did not matter to the test. Again, however, this case arose before
Finally, nothing would require that the presumption in the “New Product Rationale” as applied
to software could be rebutted only in the case of partial substitutes. Obviously, there may be other objec-
tive and significant competitive threats that merit similar treatment.34But this Court need not write a
Treatise. It is enough to apply a reasonable rule in light of the facts that this Court has found.
TEST
APPLIED
If Microsoft II does not control, then the Court should decide the Section 1 tying claim under the
standards of Jefferson Parish. While case law before Jefferson Parishsuggested that the “separate product”
question may be affected by “integration,” Jefferson Parishmakes the “separate product” analysis turn upon
“separate demand.” Applying the “separate demand” test to the facts as this Court has found them, the
Court should conclude that the government has established “separate products” for purposes of Section 1
It is also appropriate, however, for the Court to suggest in addition a test to apply to software ties
that reflects the experience of this case. Based on the framework of the Areeda Treatise, I have suggested
33 The Areeda Treatise points out the pricing problem with a solution that permits a finding of a “single product” if
the defendant permits opt-out. See AREEDA TREATISE, ¶1743b, at 194-95. Since the product at issue here is zero-
priced, this Court need not resolve that question in this case.
34 For example, if the tied product raises rivals’ costs sufficiently to make competition in the tying market impossible,
then again a court may have reason to find the presumption that these two products are one rebutted. See, e.g., Pl.
CL., at 64; AREEDA TREATISE, ¶1729b; Thomas Krattenmaker & Steven Salop, Anticompetitive Exclusion: Raising
Rivals’ Costs to Achieve Power Over Price, 96 YALE L.J. 209, 230-48 (1986); Meese, Bundling, supra note 13, at 108-
10.
an analysis that reflects the findings this Court has made, and the rightful concerns about innovation that
Microsoft has raised. I have argued:
A “Software Product” should be identified as “functionality separately valued by
consumers.” This Court has therefore found that “browser functionality” can be
considered a “software product.”
The “Competitive Market Practices” test indicates that while it is not improper
to bundle an optional browser with an operating system, there is no precedent
among competitors to validate tilting the operating system towards a particular
browser. Thus, the CMP test would provide no reason to treat this kind of
“browser functionality” combined with operating systems as a single product.
Under a modified “New Product Rationale,” however, there would be a reason to
treat the browser functionality and operating system presumptively as a single
product.
But given the competitive threat posed by the bundling of “partial substitutes,”
that presumption would be rebutted, as the alleged tied product, browser func-
tionality, operates in this market as a partial substitute for an operating system.
“Browser technology” and “operating system” would therefore be considered two products for
purposes of antitrust tying.35
35 I have not addressed whether this modified “New Product Rationale” would entail a “rule of reason” analysis or a
modification of the “per se” rule, but I do not believe the test entails a “rule of reason” analysis. While the “separate
product” inquiry may be more extensive under a rule of reason, AREEDA TREATISE, ¶¶1742b, 1742c, and hence
more in line with the modified “New Product Rationale,” the aim of the modified rule is to narrow the over-
inclusiveness of the rule, and may therefore be an appropriate balance in the context of the per se rule as well. In any
case, the Supreme Court has indicated that the ultimate aim under both standards is the same. See NCAA v. Bd. of
Regents, University of Oklahoma, 468 U.S. 85, 103-04 n.26 (1984).
CONCLUSION
For the foregoing reasons, in the manner described, this Court may conclude either that the gov-
ernment has succeeded in establishing a “separate product” for purposes of “tying” under Section 1 of the
Sherman Act, or that it has not.
_________________________
Lawrence Lessig, pro hac vice
Harvard Law School
1525 Massachusetts Ave
Cambridge, MA 02138
(617) 495-8399
I hereby certify that on this 1st day of February, 2000, I caused a true and correct copy of the
foregoing Brief of Amicus Curiae to be mailed by U.S. First Class Mail to:
A. Douglas Melamed, Esq.
Antitrust Division
U.S. Department of Justice
950 Pennsylvania Avenue, N.W.
Washington, D.C. 20530
John L. Warden
Sullivan & Cromwell
125 Broad Street
New York, NY 10004
(212)-558-4000
Richard L. Schwartz, Esq.
Deputy Chief, Antitrust Bureau
New York State Attorney General’s Office
120 Broadway, Suite 2601
New York, New York 10271
Kevin J. O’Connor, Esq.
Office of the Attorney General of Wisconsin
P.O. Box 7857
123 West Washington Avenue
Madison, Wisconsin 53703-7957
Fax: (608) 267-2223
Christine Rosso, Esq.
Chief, Antitrust Bureau
Illinois Attorney General’s Office
100 West Randolph Street, 13th Floor
Chicago, Illinois 60601
Fax: (312) 814-2549
______________________
Lawrence Lessig