NETFLIX INC v. DIVX LLC (2026)

DivX, LLC, owns U.S. Patent No. 10,225,588, which claims systems and methods for streaming partly encrypted media content. After DivX sued Netflix, Inc., alleging infringement of the patent, Netflix successfully petitioned the Patent and Trademark Office (PTO) to institute an inter partes review (IPR) of all claims of the patent, whose subject matter Netflix asserted would have been obvious over specified prior-art references. Now before us is a final written decision of the PTO's Patent Trial and Appeal Board in which the Board panel's majority, over a dissent, adopted a disputed claim construction and on that basis rejected Netflix's obviousness challenge. Netflix, Inc. v. DivX, LLC, IPR 2020-00558, 2024 WL 734765, at *4–10, *12–14 (P.T.A.B. Feb. 22, 2024) (2024 Decision). On Netflix's appeal, we reverse the Board's claim construction, vacate the Board's decision, and remand for further proceedings consistent with the claim construction we adopt.

I

A

To stream media content on a user's playback device, individual media streams, including audio, visual, and subtitle streams, are sent from a server to the user's device. ’588 patent, col. 1, lines 45–58. Those streams can be stored in buffers on the user's device in sufficient quantity to enable uninterrupted playback—without the device having to pause playback to await receipt of more content from the server. Id., col. 1, lines 48–52. Individual streams of content can be encoded at several different bitrates (amounts of data transmitted per unit of time), producing “alternative streams” of the same content. Id., col. 1, lines 64–67; see id., col. 1, lines 55–56.

Adaptive bitrate streaming (ABS) is a form of media streaming in which, based on detected current streaming conditions (notably, network bandwidth), the quality of the streamed media to be played on the device is adjusted to match those conditions by selecting the appropriate stream from among alternative streams. Id., col. 1, lines 59–64. For example, a device using ABS may select a lower quality media stream (using less data) when network bandwidth is poor (forcing lower bitrate transmission) so that it has enough content to continue playback (albeit at lower quality) without interruption. See id.

To reduce unauthorized access or copying, streams of media content can be “protected” by techniques such as Digital Rights Management (DRM). Id., col. 2, lines 47–49. Encryption/decryption is one such technique, requiring that cryptographic information be provided to the playback device. Id., col. 2, lines 49–62. The patent indicates that one way to reduce resources needed for encryption/decryption processes (while still effectively reducing unauthorized access or copying) is to encrypt only parts of streamed media, such as “portions of frames of video,” and to furnish the playback device information about which portions are encrypted as well as “common” decryption information (e.g., encryption keys), not unique to each encrypted portion. See, e.g., id., col. 3, lines 13–30; col. 4, lines 39–58; claim 1, col. 27, lines 37–40.

In ABS systems, protected source media, owned by the provider, is usually stored on a server as a top-level index file. Id., col. 2, lines 12–17. Top-level index files identify alternative media streams for different audio and video data depending on the streaming conditions (e.g., a high-quality stream and a low-quality stream), and the alternative streams are typically located and stored within container files. Id. One system of container files, described in many embodiments of the patent, is evocatively named the “Matroska” file system, based on a standard known in the field. See, e.g., id., col. 2, lines 35–46; col. 20, lines 27–32; id. pp. 8–11 (“other publications” listing, citing many Matroska documents).

As the parties before us do not dispute, claim 1 of the patent itself makes clear that “encryption information” must “identif[y] encrypted portions of frames of video.” Id., col. 27, lines 54–55. The patent refers to several types of “encryption information,” such as “DRMInfo,” which the Board implicitly acknowledged was a type of encryption information. See id., col. 4, lines 42–46; 2024 Decision, at *6.

Claim 1 is representative for present purposes. We use the annotations adopted by the parties in their briefing:

1. [a] A playback device for playing protected content from a plurality of alternative streams, comprising:

[b] a set of one or more processors; and

a non-volatile storage containing an application for causing the set of one or more processors to perform the steps of:

[c] obtaining a top level index file identifying a plurality of alternative streams of protected video, [d] wherein each of the alternative streams of protected video includes partially encrypted video frames [e] that are encrypted using a set of common keys comprising at least one key, [f] and wherein the partially encrypted video frames contain encrypted portions and unencrypted portions of data;

[g] obtaining a copy of the set of common keys;

[h] detecting streaming conditions for the playback device;

[i] selecting a stream from the plurality of alternative streams of protected video based on the detected streaming conditions;

[j] receiving a container index that provides byte ranges for portions of the selected stream of protected video within an associated container file;

[k] requesting portions of the selected stream of protected video based on the provided byte ranges;

[l] locating encryption information that identifies encrypted portions of frames of video within the requested portions of the selected stream of protected video;

[m] decrypting each encrypted portion of the frames of video identified within the located encryption information using the set of common keys; and

[n] playing back the decrypted frames of video obtained from the requested portions of the selected stream of protected video.

’588 patent, col. 27, lines 30–63 (emphasis added).

B

DivX sued Netflix in 2019, alleging infringement of the ’588 patent. In February 2020, Netflix petitioned for an inter partes review of all the patent's claims (1–24) under 35 U.S.C. §§ 311–19, asserting obviousness over U.S. Patent Application Publication No. 2011/0096828 (Chen) in view of U.S. Patent Application Publication No. 2007/0083467 (Lindahl) and U.S. Patent No. 8,683,066 (Hurst). J.A. 11020–21. In August 2020, the Board instituted review. J.A. 11259. In its institution decision, the Board rejected as “too restrictive” DivX's proposed construction of limitation [l], which DivX argued meant that the “encryption information” itself be located “within the requested portions of the selected stream of protected video,” concluding that the better construction was instead that the encryption information (wherever it was located) simply had to identify encrypted portions of frames that themselves were “within the requested portions of the selected stream of protected video.” J.A. 11242–45. In the Board's view, at that stage, the encrypted-portion-containing frames of video (to be decrypted) must be within the requested portions of the selected stream, but the encryption information (identifying the encrypted portions) need not be. See id.

In August 2021, the Board issued a final written decision. See Netflix, Inc. v. DivX, LLC, IPR 2020-00558, 2021 WL 3729361, at *11 (P.T.A.B. Aug. 23, 2021). In that decision, the Board agreed with Netflix that a relevant artisan would have been motivated to combine the prior-art references, and that DivX's construction of limitation [l] was unpersuasive, but concluded that a relevant artisan would not have reasonably expected success in combining the prior art in view of certain disclosures in the prior art references. See id. at *3, *6, *10–11. On that basis, which did not depend on the resolution of the claim-construction dispute, the Board determined that Netflix had not shown unpatentability of any of the claims. Id., at *11. Netflix appealed, and in March 2023, we vacated the decision, concluding that the Board's analysis of reasonable expectation of success was erroneous, and without ruling on the claim construction of limitation [l], we remanded the matter to the Board. See Netflix, Inc. v. DivX, LLC, No. 2022-1083, 2023 WL 2298768, at *4–6 (Fed. Cir. Mar. 1, 2023).

On remand, the Board panel (by a divided vote) concluded again that Netflix had not shown unpatentability of the challenged claims, 2024 Decision, at *14, but this time, the Board did so by accepting the claim construction of limitation [l] that DivX had pressed and that the Board had previously rejected. The majority concluded that limitation [l] requires “that the encryption information is located within the requested portions of the selected stream of protected video.” Id., at *5 (emphasis in original). Based on that construction of limitation [l], the panel majority concluded that Netflix had not shown limitation [l] to be met by the asserted combination of prior art and so had not established obviousness. Id., at *14. One panel member dissented, rejecting the majority's claim construction and concluding that Netflix had shown unpatentability of all claims under the correct claim construction. Id., at *14–30.

Netflix timely appealed the Board's new final written decision. We have jurisdiction to review Netflix's appeal under 28 U.S.C. § 1295(a)(4)(A).

II

Netflix argues that the Board incorrectly construed limitation [l], whereas DivX argues to the contrary. Netflix argues, and indeed shows, that, under its construction, limitation [l] is indisputably taught by Lindahl and, hence, the asserted prior-art combination, Netflix Opening Br. at 63–66, and DivX does not disagree. Accordingly, the only issue requiring decision is the proper claim construction of limitation [l].

We review the Board's claim construction in this matter de novo, as the required analysis here relies entirely on intrinsic evidence and general grammatical, language-interpretation principles that are not matters for findings by the factfinder. See Intel Corp. v. Qualcomm Inc., 21 F.4th 801, 808 (Fed. Cir. 2021); SIMO Holdings Inc. v. Hong Kong uCloudlink Network Technology Ltd., 983 F.3d 1367, 1374, 1377 (Fed. Cir. 2021). We consider the ordinary meaning of the language specifically at issue, other claim language providing context, the specification, and aspects of prosecution history. See Intel, 21 F.4th at 809; World Class Technology Corp. v. Ormco Corp., 769 F.3d 1120, 1123 (Fed. Cir. 2014); Phillips v. AWH Corp., 415 F.3d 1303, 1312–17 (Fed. Cir. 2005) (en banc). We conclude that the Board's construction of limitation [l] was erroneous. We agree with Netflix about the proper claim construction—under which limitation [l] is indisputably taught by the asserted prior-art combination. We therefore vacate the Board's decision and remand for further proceedings.

A

The disputed limitation requires “locating encryption information that identifies encrypted portions of frames of video within the requested portions of the selected stream of protected video.” ’588 patent, col. 27, lines 54–57. One thing is plain about this language, considered on its own (before considering background interpretive principles, other claim elements, and the specification): the language is susceptible of two different interpretations, each syntactically and semantically available. Considering the phrase on its own, the modifier “within the requested portions of the selected stream of protected video” could modify either “encrypted portions of frames of video” or “encryption information,” i.e., could specify the location of either the encrypted portions or the encryption information. Syntactically, a modifier with two distinct possible modificands is susceptible of modifying either one. And semantically, in the present example, considering the phrase on its own (indeed, even in broader context), neither of the possibilities can be excluded as obviously senseless or implausible.

But the process for arriving at a proper understanding of even the phrase itself, in its ordinary linguistic meaning, requires one important step beyond identifying the words and their arrangement in the phrase. Critically, we look to “precepts of English grammar” to determine the ordinary meaning of the phrase as a whole. In re Hyatt, 708 F.2d 712, 714 (Fed. Cir. 1983). Not surprisingly, a general interpretive principle that applies to this very situation has long been established, precisely because the linguistic situation occurs frequently and therefore a default principle is needed.

The principle is that (where commas or other textual signals are not used) the modifier is presumptively understood to be tied to the nearest available semantically plausible modificand. See, e.g., Nearest-Reasonable-Referent Canon, Black’s Law Dictionary 1240 (11th ed. 2019) (“The doctrine that when the syntax in a legal instrument involves something other than a parallel series of nouns or verbs, a prepositive or postpositive modifier normally applies only to the nearest reasonable referent.”); Antonin Scalia & Bryan A. Garner, Reading Law: The Interpretation of Legal Texts 152 (2012); Rule of Last Antecedent, Black’s Law Dictionary 1449–50 (9th ed. 2009); Last Antecedent Rule, Webster’s Dictionary of Law 281 (1996). The principle also appears in normative guides to writing, which reflect (and reinforce) how language is most likely to be understood. See W. Strunk & E.B. White, The Elements of Style 28, 30 (4th ed. 2000); Chicago Manual of Style § 5.175 (16th ed. 2010), p. 248. And the principle appears as well in judicial decisions. See, e.g., Lockhart v. United States, 577 U.S. 347, 351–52, 136 S.Ct. 958, 194 L.Ed.2d 48 (2016); Finisar Corp. v. DirecTV Group, Inc., 523 F.3d 1323, 1336 (Fed. Cir. 2008); Anhydrides & Chemicals Inc. v. United States, 130 F.3d 1481, 1483 (Fed. Cir. 1997).

This principle clearly calls for Netflix's construction as the proper default interpretation—only the “encrypted portions of frames of video” (not the “encryption information”) need be “within the requested portions of the selected stream of protected video.” DivX offers a number of language-phrase examples that it says run counter to the principle supplying a default, presumptive interpretation here. See DivX Br. at 20–21, 31. But DivX's examples do not undermine that principle, as, for one reason or another, they do not involve a modifier in a phrase containing two syntactically appropriate and semantically sensible modificands. See Netflix Reply Br. at 9–11. They therefore do nothing to undermine the applicable grammar precept at issue.

Because the precept establishes a presumptive interpretation, the analysis cannot stop there. We must look to see if there are good indications that a contrary interpretation is the better one. As explained next, we see no such persuasive indications.

B

1

The context provided by the claim itself reinforces, rather than undermines, the interpretation called for by the general-interpretation principle. This interpretation makes good sense of the relationship between limitation [l] and the limitations that follow. The action specified in limitation [l] is finding the information (“encryption information”) that “identifies” encrypted portions of frames of video “within the requested portions” so that the actions specified in the next two limitations can occur. In limitation [m], the action specified is “decrypting” “each encrypted portion of the frames of video” just identified by the action specified in limitation [l]. And the action specified in limitation [n] is the culmination: “playing back the decrypted frames of video obtained from the requested portions of the selected stream of protected video.” ’588 patent, col. 27, lines 54–62.¹

Netflix's interpretation thus makes a straightforward and coherent whole out of the final three limitations, which have the focus of the parties’ arguments about the interpretive choice at issue. Nothing about those three limitations requires or even suggests that the trio of actions called for require that the “encryption information”—which simply identifies the video data needing decryption—must be found in “the requested portions of the selected stream of protected video.” In particular, Netflix's interpretation does not make limitation [l] superfluous. It preserves that limitation's demand that the processor locates encrypted portions of frames of video within the specifically requested portions of the selected streams, as opposed to within any portion of the selected stream of protected video. Id., col. 27, lines 54–57.

Limitations within claim 1 that precede limitation [l] play a much smaller role in the analysis here. At least one such limitation is worth noting as claim-language context. Netflix argues, based on the specification and prosecution history, that encryption information need not be located in the requested portions of the selected video stream, but can be located elsewhere—and the elsewhere featured by Netflix is “a top level index file” that limitation [c] expressly requires the playback device to obtain. Id., col. 27, lines 36–43; see Netflix Br. at 51–53. That limitation supplies an anchor in the claim for Netflix's argument that the invention as a whole does not require the specific location of encryption information that DivX insists is an essential part of the invention.

2

The specification does not overcome the presumptive meaning of this limitation driven by the well-recognized background principle of interpretation we have identified. Netflix does not dispute that several embodiments described in the specification involve the location of encryption information within the requested portions of the selected stream of protected video. See 2024 Decision, at *6 (citing ’588 patent, col. 9, lines 23–29; id., col. 13, lines 57–63; col. 16, lines 33–52; col. 22, lines 14–17; col. 25, lines 15–21). But we agree with Netflix that the specification does not limit the invention to that arrangement—that, indeed, it positively indicates that the invention covers (though it does not elaborate on) embodiments in which encryption information is not so located. It could, in fact, be located in a top-level index file. That is enough to mean that the specification does not overcome the presumptive interpretation Netflix presses.

For example, while the specification highlights embodiments implementing a specially formatted Matroska container file, the specification also states that “many embodiments ․ utilize a conventional Matroska container.” ’588 patent, col. 20, lines 30–32 (emphasis added); see col. 2, lines 35–46; id., col. 20, lines 27–32. DivX's expert recognized that the “conventional,” unmodified Matroska container cannot contain necessary encryption information, meaning that there is no possibility that encryption information could be located “within the requested portions of the selected stream” in an embodiment utilizing a conventional Matroska container file. See J.A. 9461 ¶ 90. In discussing Matroska files, the specification pervasively discusses a “top level index file,” including, to cite just one example, at ’588 patent, col. 20, lines 39–41. The specification thus contemplates a location of the encryption information at issue different from where DivX says it must be.

The dissenter in the Board highlighted another relevant passage in the specification. That passage suggests that some encryption information does not need to come from the container files themselves—instead, the user's device can use the information from either Uniform Resource Identifiers or container file headers to request byte ranges from the server. ’588 patent, col. 24, line 63, through col. 25, line 2; 2024 Decision, at *16.

In addition, Netflix points to several other specification passages that, in describing embodiments, either are silent about the location of the encryption information or specify that the location of the encryption information is within the “alternative streams of protected video” but not more specifically (as DivX's construction requires) “within the requested portions of the selected streams.” Netflix Opening Br. at 48; see ’588 patent, col. 3, lines 1–12 (not specifying a location for the encryption information); id., col. 3, lines 13–30; col. 5, lines 4–23; col. 7, lines 8–25 (describing DRM information, a type of encryption information, as being located within the alternative streams of protected video, but not, more specifically, within the requested portions of the streams).

In this circumstance, it is not enough, to support DivX's construction, that the most fully elaborated embodiments describe a location of encryption information (in particular, DRMInfo) within the requested portions of the video. The specification's description of those embodiments does not indicate that this feature is required in the invention as a whole. And the specification, as noted, refers (though with less detailed elaboration) to embodiments that do not call for the location of encryption information to which DivX would limit claim 1. Here, “the language of the written description is sufficient to put a reader on notice of the different” possible locations of encryption information. Pitney Bowes, Inc. v. Hewlett-Packard Co., 182 F.3d 1298, 1311 (Fed. Cir. 1999). At a minimum, the specification does not furnish a basis for overcoming the presumptive interpretation of the claim language established by the grammar precept we have discussed.

3

The prosecution history provides one additional bit of support for reading the specification as confirming that, as Netflix urges, the invention is not restricted to the location of encryption information DivX says is required by claim 1. U.S. Patent No. 9,621,522 is the grandparent of the ’588 patent, and it shares the ’588 patent’s specification and similarly claims ABS streaming systems and methods using partial encryption techniques. Claim 1 of the ’522 patent calls for “DRM Information” (a type of encryption information) to be located within a “top level index file” (“identif[ying] a plurality of alternative streams of protected video” stored in container files), which is different from being located within the “requested portions of selected streams.” ’522 patent, col. 3, lines 14–15, id., col. 9, lines 30–32; id., col. 27, lines 31–41. DivX's filing and the PTO's issuance of that claim, which requires written-description support, provide additional reason to conclude that the specification (shared by the ’522 and ’588 patents) contemplates encryption information being located outside the requested portions of selected streams.

III

For the foregoing reasons, we reverse the Board's construction of limitation [l] and conclude that it is the “encrypted portions” of frames of video, and not the “encryption information,” that must be located “within the requested portions of the selected stream of protected video.” We hold, too, that limitation [l], so construed, is taught by the asserted prior art. We vacate the Board's decision and remand for further proceedings consistent with this opinion.

Costs awarded to Netflix.

REVERSED, VACATED, AND REMANDED

FOOTNOTES

1.   Limitation [m] reads: “decrypting each encrypted portion of the frames of video identified within the located encryption information using the set of common keys.” ’588 patent, col. 27, lines 58–60. We read “within” in this phrase to mean what would be more clearly expressed if the word were “by”: What is being decrypted are the encrypted video portions identified by the “encryption information.”

Taranto, Circuit Judge.