`571-272-7822
`
`
`
`
`
`
`
`
`Paper No. 62
`Filed: September 28, 2017
`
`
`
`
`UNITED STATES PATENT AND TRADEMARK OFFICE
`____________
`
`BEFORE THE PATENT TRIAL AND APPEAL BOARD
`____________
`
`HEWLETT-PACKARD ENTERPRISE CO.; HP ENTERPRISE
`SERVICES, LLC; and TERADATA OPERATIONS, INC.,
`Petitioner,
`
`v.
`
`REALTIME DATA LLC d/b/a IXO,
`Patent Owner.
`____________
`
`Case IPR2016-00783
`Patent 6,597,812 B1
`____________
`
`
`Before GEORGIANNA W. BRADEN, JASON J. CHUNG, and
`SCOTT C. MOORE, Administrative Patent Judges.
`
`CHUNG, Administrative Patent Judge.
`
`
`FINAL WRITTEN DECISION
`Inter Partes Review
`35 U.S.C. § 318(a) and 37 C.F.R. § 42.73
`
`
`
`
`
`
`
`IPR2016-00783
`Patent 6,597,812 B1
`
`
`I.
`
`INTRODUCTION
`
`Hewlett-Packard Enterprise Company, HP Enterprise Services, LLC,
`and Teradata Operations, Inc. (collectively, “Petitioner”), filed a Petition to
`institute an inter partes review of claims 1–4, 8, 14–17, 21, and 28 of U.S.
`Patent No. 6,597,812 B1 (“the ’812 patent”). Paper 1 (“Pet.”). Realtime
`Data LLC (“Patent Owner”), filed a Preliminary Response pursuant to
`35 U.S.C. § 313. Paper 12 (“Prelim. Resp.”).
`Upon consideration of the Petition and the Preliminary Response, on
`October 5, 2016, we instituted inter partes review of claims 1–4, 8, 14–17,
`21, and 28 (“instituted claims”), pursuant to 35 U.S.C. § 314. Paper 19
`(“Dec.”).
`Subsequent to institution, Patent Owner filed a Patent Owner
`Response. Paper 29 (“PO Resp.”). Petitioner filed a Reply to Patent
`Owner’s Response. Paper 37 (“Reply”). An oral hearing was held on June
`30, 2017 and a transcript of the oral hearing is available in the record. Paper
`59 (“Tr.”).
`We issue this Final Written Decision pursuant to 35 U.S.C. § 318(a)
`and 37 C.F.R. § 42.73. For the reasons discussed herein, Petitioner has
`shown by a preponderance of the evidence that claims 1–4, 8, 14–17, 21, and
`28 of the ’812 patent are unpatentable. See 35 U.S.C. § 316(e).
`
`Related Matters
`
`A.
`Petitioner and Patent Owner inform us that the ’812 patent is involved
`in multiple suits in the U.S. District Court for the Eastern District of Texas.
`Pet. 1; Paper 9, 1–2; Paper 10, 2–3; Paper 58, 4–5. Patent Owner also
`informs us that the ’812 patent is involved in a suit in the U.S. District Court
`
`2
`
`
`
`IPR2016-00783
`Patent 6,597,812 B1
`
`for the Northern District of California. Paper 9, 2; Paper 10, 2–3; Paper 58,
`4–5.
`
`The Instituted Grounds
`
`B.
`We instituted the following grounds of unpatentability:
`References1
`Basis
`Instituted Claims
`O’Brien2 and Nelson3
`§ 103(a)4 1–4, 8, and 28
`O’Brien, Nelson, and Welch5 § 103(a) 14–17 and 21
`
`The ’812 Patent
`
`C.
`The ’812 patent describes systems and methods “for providing
`lossless data compression and decompression.” Ex. 1001, Abs. The ’812
`patent further describes “characteristics of run-length encoding, parametric
`dictionary encoding, and bit packing to comprise an encoding/decoding
`process.” Id. Figure 1 of the ’812 patent is reproduced below.
`
`
`1 Petitioner also relies upon the Declarations of Dr. Charles D. Creusere.
`Ex. 1005
`2 U.S. Patent No. 4,929,946; issued May 29, 1990, (Ex. 1002, “O’Brien”)
`3 MARK NELSON, THE DATA COMPRESSION BOOK (1992), (Ex. 1003,
`“Nelson”)
`4 The Leahy-Smith America Invents Act (“AIA”), Pub. L. No. 112-29,
`125 Stat. 284, 287–88 (2011), revised 35 U.S.C. § 103, effective March 16,
`2013. The ’812 patent was issued prior to the effective date of the AIA.
`Thus, we apply the pre-AIA version of § 103.
`5 U.S. Patent No. 4,558,302; issued Dec. 10, 1985, (Ex. 1004, “Welch”)
`
`3
`
`
`
`IPR2016-00783
`Patent 6,597,812 B1
`
`
`
`
`Figure 1 of the ’812 patent, reproduced above, is a detailed block
`diagram of a system for combining run-length encoding with dictionary
`encoding. Ex. 1001, 5:14–23. Input buffer 11 temporarily buffers an input
`data stream, and encoder 12 compresses the input data stream. Id. at 4:66–
`5:2. Encoder 12 implements a combination of run-length encoder 13 and
`dictionary encoder 14. Id. at 5:14–22. More specifically, encoder 12
`identifies any run-length sequence in the data stream and outputs one or
`more code words from dictionary 15 to represent the run-length sequence.
`Id. at 5:31–37. Dictionary encoder 14 builds a character string comprising
`two or more characters that does not comprise a run-length sequence,
`searches dictionary 15 for a code word corresponding to the character string,
`and then outputs the code word representing the character string. Id. at
`5:38–42.
`
`4
`
`
`
`IPR2016-00783
`Patent 6,597,812 B1
`
`
`The Instituted Claims
`
`D.
`We instituted inter partes review of claims 1–4, 8, 14–17, 21, and 28.
`Claim 1 is illustrative and reproduced below:
`1. A method for compressing input data comprising a plurality
`of data blocks, the method comprising the steps of:
`detecting if the input data comprises a run-length sequence of
`data blocks;
`outputting an encoded run-length sequence, if a run-length
`sequence of data blocks is detected;
`maintaining a dictionary comprising a plurality of code words,
`wherein each code word in the dictionary is associated with a
`unique data block string;
`building a data block string from at least one data block in the
`input data that is not part of a run-length sequence;
`searching for a code word in the dictionary having a unique data
`block string associated therewith that matches the built data
`block string; and
`outputting the code word representing the built data block string.
`Ex. 1001, 16:53–17:2.
`
`II. ANALYSIS
`Principles of Law
`A.
`A claim is unpatentable under 35 U.S.C. § 103(a) if “the differences
`between the subject matter sought to be patented and the prior art are such
`that the subject matter as a whole would have been obvious at the time the
`invention was made to a person having ordinary skill in the art to which said
`subject matter pertains.” KSR Int’l Co. v. Teleflex Inc., 550 U.S. 398, 406
`(2007). The question of obviousness is resolved on the basis of underlying
`factual determinations, including: (1) the scope and content of the prior art;
`
`5
`
`
`
`IPR2016-00783
`Patent 6,597,812 B1
`
`(2) any differences between the claimed subject matter and the prior art;
`(3) the level of skill in the art; and (4) objective evidence of nonobviousness,
`i.e., secondary considerations. See Graham v. John Deere Co. of Kansas
`City, 383 U.S. 1, 17–18 (1966).
`A determination of whether a patent claim is invalid as obvious under
`§ 103 requires consideration of all four Graham factors, and it is error to
`reach a conclusion of obviousness until all those factors are considered.”
`Apple v. Samsung Elecs. Co., Ltd., 839 F.3d 1034, 1048 (Fed. Cir. 2016) (en
`banc) (citations omitted). “This requirement is in recognition of the fact that
`each of the Graham factors helps inform the ultimate obviousness
`determination.” Id.
`“In an [inter partes review], the petitioner has the burden from the
`onset to show with particularity why the patent it challenges is
`unpatentable.” Harmonic Inc. v. Avid Tech., Inc., 815 F.3d 1356, 1363 (Fed.
`Cir. 2016) (citing 35 U.S.C. § 312(a)(3) (requiring inter partes review
`petitions to identify “with particularity . . . the evidence that supports the
`grounds for the challenge to each claim”)). This burden of persuasion never
`shifts to Patent Owner. See Dynamic Drinkware, LLC v. Nat’l Graphics,
`Inc., 800 F.3d 1375, 1378 (Fed. Cir. 2015) (discussing the burden of proof in
`inter partes review). Furthermore, Petitioner cannot satisfy its burden of
`proving obviousness by employing “mere conclusory statements.” In re
`Magnum Oil Tools Int’l, Ltd., 829 F.3d 1364, 1380 (Fed. Cir. 2016).
`Thus, to prevail in an inter partes review, Petitioner must explain how
`the proposed combinations of prior art would have rendered the challenged
`claims unpatentable. At this final stage, we determine whether a
`preponderance of the evidence of record shows that the challenged claims
`
`6
`
`
`
`IPR2016-00783
`Patent 6,597,812 B1
`
`would have been obvious over the proposed combinations of prior art.
`We analyze the instituted grounds of unpatentability in accordance
`with the above-stated principles.
`
`Level of Ordinary Skill in the Art
`
`B.
`The parties do not disagree as to the level of skill in the art. See
`generally Pet., PO Resp., Reply. We find that the level of ordinary skill in
`the art is reflected by the prior art of record. See Okajima v. Bourdeau,
`261 F.3d 1350, 1355 (Fed. Cir. 2001); In re GPAC Inc., 57 F.3d 1573, 1579
`(Fed. Cir. 1995); In re Oelrich, 579 F.2d 86, 91 (CCPA 1978).
`
`Claim Construction
`
`C.
`In the Decision to Institute, we did not construe any terms. Dec. 5.
`Patent Owner argues a construction for “maintaining a dictionary” (claims 1
`and 14) and “consecutively outputting a first control word indicating a run-
`length sequence, a code word in the dictionary . . . that corresponds to the
`input data block, and a word corresponding to the number of successive data
`blocks that are similar to the input data block” (claims 3 and 16). PO Resp.
`23–33. In response, Petitioner argues for a construction of “maintaining a
`dictionary” (claims 1 and 14) and “consecutively” (claims 3 and 16). Reply
`1–9.
`
`Consecutively (Claims 3 and 16)
`
`1.
`Patent Owner argues “consecutively outputting a first control word
`indicating a run-length sequence, a code word in the dictionary . . . that
`corresponds to the input data block, and a word corresponding to the number
`of successive data blocks that are similar to the input data block” (claims 3
`
`7
`
`
`
`IPR2016-00783
`Patent 6,597,812 B1
`
`and 16) should be construed as “the control word indicating a run-length
`sequence is outputted before the other two words in the run-length
`sequence.” PO Resp. 27–32. At the outset, Patent Owner mischaracterizes
`Petitioner’s argument to construe “consecutively” (see Pet. 31) as a request
`to construe “consecutively outputting a first control word indicating a run-
`length sequence, a code word in the dictionary . . . that corresponds to the
`input data block, and a word corresponding to the number of successive data
`blocks that are similar to the input data block” (claims 3 and 16) to mean
`“following one another in uninterrupted order; successive.” PO Resp. 27.
`We note, however, that Petitioner did not argue for a construction of
`“consecutively outputting a first control word indicating a run-length
`sequence, a code word in the dictionary . . . that corresponds to the input
`data block, and a word corresponding to the number of successive data
`blocks that are similar to the input data block” (claims 3 and 16). Pet. 31;
`Reply 8. Rather, Petitioner argues for a construction of just the term
`“consecutively” (claims 3 and 16). Pet. 31; Reply 8.
`Regarding the construction of the phrase “consecutively outputting a
`first control word indicating a run-length sequence, a code word in the
`dictionary . . . that corresponds to the input data block, and a word
`corresponding to the number of successive data blocks that are similar to the
`input data block” (claims 3 and 16), Patent Owner argues the use of the term
`“first” recited in claims 3 and 16 necessitates a particular chronological
`order – i.e., the control word is output first. PO Resp. 28–30 (citing
`Ex. 2007 ¶¶ 101–107, 152). In addition, Patent Owner argues Petitioner is
`attempting to read out “first” from claims 3 and 16, thereby rendering the
`term “first” superfluous. PO Resp. 29.
`
`8
`
`
`
`IPR2016-00783
`Patent 6,597,812 B1
`
`Patent Owner also argues its proposed construction is supported by a
`preferred embodiment discussed in the Specification of the ’812 patent that
`describes the control code being output first from the dictionary, followed by
`the code word for the character that is stored, which is then followed by the
`number of consecutive characters that were found in the input stream. PO
`Resp. 29–32 (citing Ex. 1001, 8:33–39, Fig. 4A).
`We disagree with Patent Owner. In particular, we disagree with
`Patent Owner that the word “first” necessitates a particular chronological
`order in which “a first control code word indicating a run-length sequence”
`(claims 3 and 16) is output before both “a code word in the dictionary
`having a unique data block string associated therewith that corresponds to
`the input data block” (claims 3 and 16) and “a word corresponding to the
`number of successive data blocks that are similar to the input data block”
`(claims 3 and 16) are output. More specifically, we disagree with Patent
`Owner because the word “first” recited in claims 3 and 16 modifies the
`claimed “control word” rather than modifying the claimed “outputting,”
`while “consecutively” modifies “outputting.” Ex. 1001, 17:9–15, 18:40–48.
`That is, claims 3 and 16 recite “first control code word” and “consecutively
`outputting.” Id. Because “first” modifies “control word” rather than
`modifying “outputting,” we are not persuaded by Patent Owner that “first”
`has any bearing on an alleged chronological order of outputs in claims 3 and
`16.
`
`Moreover, we disagree with Patent Owner’s argument that the
`Specification of the ’812 patent limits claims 3 and 16 to the chronological
`order of “words” described in the “preferred embodiment.” Ex. 1001, 6:14–
`33. Limiting a claim to what is described in a preferred embodiment is
`
`9
`
`
`
`IPR2016-00783
`Patent 6,597,812 B1
`
`seldom correct. Hill-Rom Servs., Inc. v. Stryker Corp., 755 F.3d 1367, 1372
`(Fed. Cir. 2014) (“Even when the specification describes only a single
`embodiment, the claims of the patent will not be read restrictively unless the
`patentee has demonstrated a clear intention to limit the claim scope using
`‘words or expressions of manifest exclusion or restriction.’”).
`We instead adopt Petitioner’s definition of “consecutively” to mean
`“following one another in uninterrupted order; successive” (Pet. 31; Reply 8)
`because it is consistent with the plain and ordinary meaning and the
`Specification.
`
`Remaining Terms
`
`2.
`We determine that no other terms require express construction for
`purposes of this Decision. See Vivid Techs., Inc. v. Am. Sci. & Eng’g, Inc.,
`200 F.3d 795, 803 (Fed. Cir. 1999) (only those claim terms or phrases that
`are in controversy need to be construed, and only to the extent necessary to
`resolve the controversy).
`
`D.
`
`Overview of O’Brien (Ex. 1002)
`
`Alleged Obviousness of Independent Claims 1 and 14
`1.
`O’Brien teaches adaptive data compression to compress efficiently a
`user data file. Ex. 1002, Abs. O’Brien teaches that “[r]uns of three or more
`repeated bytes are encoded using a predetermined set of reserved reference
`values to indicate that the preceding character was repeated the number of
`times specified by the repeat code.” Id. at 3:67–4:2. O’Brien further teaches
`the adaptive data compression operates in a way such that strings are built a
`character at a time, which means “a previously defined string plus the next
`
`10
`
`
`
`IPR2016-00783
`Patent 6,597,812 B1
`
`user data byte shall define a new string and is assigned the next previously
`undefined reference value.” Id. at 12:23–31. As a result, strings become
`longer and data compression becomes more efficient as more data bytes are
`examined. Id. at 12:31–33. String definition occurs by combining the last
`used reference value with the next user data byte; this resultant string is then
`used to search the string table to determine if this string was defined
`previously. Id. at 12:34–36. If the string was defined previously, the next
`subsequent data byte is concatenated to the reference value of the string that
`has been found to form a new string table search pattern. Id. at 12:36–41.
`The search is repeated iteratively until a string is found that has not been
`defined previously. Id. at 12:43–44. Once the undefined string is found, the
`last used defined string reference is placed in the output compressed data
`stream and the next consecutive unused reference value is assigned to this
`undefined string. Id. at 12:44–48.
`
`Overview of Nelson (Ex. 1003)
`
`2.
`Nelson teaches source code for a complete version of Lempel-Ziv-
`Welch (“LZW”) compression and decompression. Ex. 1003, 306. Nelson
`teaches using a dictionary to include definitions of symbols, code,
`characters, etc. Id. at 308. In addition, Nelson teaches adding definitions to
`the dictionary if a definition is not present in the dictionary. Id.
`
`O’Brien Teaches Maintaining a Dictionary (Claims 1 and 14)
`
`3.
`The parties’ dispute focuses on whether O’Brien teaches “maintaining
`a dictionary” as recited in claims 1 and 14. Pet. 40–42, 55–57; PO Resp.
`35–40. Because Petitioner has the burden of proof (see 35 U.S.C. § 316(e),
`37 C.F.R. § 42.20(c)), we begin with Petitioner’s arguments.
`
`11
`
`
`
`IPR2016-00783
`Patent 6,597,812 B1
`
`Petitioner argues O’Brien teaches string compression, which is a
`dictionary algorithm. Pet. 41 (citing Ex. 1005 ¶ 84). Petitioner further
`argues that a person having ordinary skill in the art would have considered
`O’Brien’s “reference values” to be dictionary indices. Id. And, Petitioner
`argues “[t]he combination of character reference values mapped to
`characters and string reference values mapped to character strings stored in
`the string table is an example of a ‘dictionary.’ (Run length reference values
`are also part of the ‘dictionary,’ as control code words.)” Id. Petitioner
`argues O’Brien creates, populates, and uses the data structures and logic
`associated with its reference values and string table, and therefore that
`O’Brien teaches “maintaining a dictionary” as recited in claim 1 and 14. Id.
`at 41–42, 55–57.
`In response, Patent Owner argues that O’Brien’s segmentation
`approach fails to teach maintaining a dictionary because O’Brien generates a
`new dictionary for each new segment. PO Resp. 33–40. In particular,
`Patent Owner argues O’Brien divides its input data into segments of a pre-
`determined size; O’Brien then encodes each segment independently from the
`other segments and discards the dictionary after each segment is encoded
`and assigns new reference values for each segment. Id. at 34 (citing
`Ex. 1002, Abstract; Ex. 2009, 43:14–20).
`Patent Owner argues because O’Brien discards the dictionary and
`assigns new reference values for each segment, O’Brien fails to teach
`“maintaining a dictionary” pursuant to Patent Owner’s proposed
`construction. PO Resp. 34–35. Moreover, Patent Owner also argues
`O’Brien’s BEGIN and END variables teaches the generating and discarding
`
`12
`
`
`
`IPR2016-00783
`Patent 6,597,812 B1
`
`of dictionaries, which Dr. Creusere acknowledges. Id. at 35–37 (citing
`Ex. 2009, 55:4–9).
`Patent Owner further argues O’Brien’s adaptive algorithm builds a
`dictionary by matching input strings to strings that were defined previously
`in the dictionary until a match is not found; then, O’Brien’s algorithm adds
`the unmatched string to an unused reference value in the dictionary. PO
`Resp. 38–39. Patent Owner argues O’Brien’s adaptive dictionary
`compression algorithm fails to teach “maintaining a dictionary” because it
`never makes a determination of whether to retain the dictionary during the
`course of compression of the input data stream and instead, discards its
`dictionary after encoding each segment and generates a new dictionary. Id.
`at 39–40 (citing Ex. 2007 ¶¶ 117–119; Ex. 2009, 55:4–9). We disagree with
`Patent Owner.
`At the outset, we note that Petitioner, Patent Owner, and their
`respective declarants all agree that O’Brien’s encoder is a type of dictionary
`encoder. See e.g., Pet. 41; PO Resp. 16–18; Ex. 1005 ¶¶ 32, 84; Ex. 2007
`¶¶ 74, 79, 80, 82, 85. Because O’Brien teaches a “dictionary,” we now turn
`to whether O’Brien teaches the larger phrase, “maintaining a dictionary” as
`recited in claims 1 and 14. To determine whether O’Brien teaches
`“maintaining a dictionary,” we turn to dependent claims 4 and 17 (Ex. 1001,
`17:16–22, 18:48–55), which depend from claims 1 and 14, respectively.
`Dependent claim 4 recites “maintaining a dictionary comprises the
`step of: dynamically generating a new code word corresponding to a built
`data block string, if the built data block string does not match a unique data
`block string in the dictionary; and adding the new code word in the
`
`13
`
`
`
`IPR2016-00783
`Patent 6,597,812 B1
`
`dictionary” (emphasis added). Id. at 17:16–22. Dependent claim 17 recites
`similar features. Id. at 18:48–55.
`We are persuaded that a person having ordinary skill in the art would
`have considered O’Brien’s “reference values” to be dictionary indices,
`because O’Brien’s string compression includes the signature features of
`LZ78 (i.e., Lempel and Ziv’s paper in 1978) and LZW, which are dictionary
`algorithms. Pet. 41. Moreover, O’Brien teaches an adaptive algorithm that
`builds a dictionary by combining strings and matching these combined
`strings to strings that were defined previously in the dictionary until a match
`is not found. Id. at 49 (citing Ex. 1004, 12:43–48; Ex. 1005 ¶ 101); see id. at
`55. At that juncture, O’Brien’s algorithm adds the unmatched combined
`string to an unused reference value in the dictionary, which we find to teach
`“maintaining a dictionary comprises the step of: dynamically generating a
`new code word corresponding to a built data block string, if the built data
`block string does not match a unique data block string in the dictionary; and
`adding the new code word in the dictionary” (emphasis added) as recited in
`claim 4 (and similarly recited in claim 17). Pet. 49 (citing Ex. 1004, 12:43–
`48; Ex. 1005 ¶ 101); see id. at 55.
`Accordingly, because O’Brien teaches “dynamically generating a new
`code word corresponding to a built data block string, if the built data block
`string does not match a unique data block string in the dictionary; and
`adding the new code word in the dictionary” recited in dependent claims 4
`and 17, we are persuaded that Petitioner has established by a preponderance
`of the evidence that O’Brien teaches “maintaining a dictionary” (claims 1
`and 14).
`
`14
`
`
`
`IPR2016-00783
`Patent 6,597,812 B1
`
`4.
`
`O’Brien’s Dictionary Encoder is Similar to a LZ78 and LZW
`Dictionary Encoder
`
`Petitioner argues O’Brien combines run-length encoding with LZ78 or
`LZW dictionary encoding. Pet. 10–11 (citing Ex. 1005 ¶¶ 32–40). In
`particular, Petitioner argues that O’Brien teaches “an example of an LZW
`variation of the LZ78 dictionary encoder.” Id. at 13 (citing Ex. 1005 ¶¶ 35–
`40). That is, O’Brien first initializes a dictionary with code words for all
`possible characters such that, at the beginning of each data segment,
`reference value encoder 304 is provided with two variables, END and
`BEGIN that correspond to the largest and smallest individual character
`codes in the data segment. Id. at 13–14 (citing Ex. 1002, 10:19–25).
`According to Petitioner, O’Brien’s END and BEGIN variables define the
`range of reference values, or code words that represent single characters and
`are referred to as character reference values. Pet. 14 (citing Ex. 1002,
`10:22–28). Petitioner also argues that O’Brien’s setting a range of character
`reference values and associated single characters “is analogous to the
`initialization of the LZW dictionary with all possible single characters.” Id.
`(citing Ex. 1005 ¶ 36).
`Petitioner then explains that O’Brien compression technique uses
`signature features of an LZ78 dictionary encoder. Pet. 13–15. In particular,
`Petitioner explains that
`O’Brien’s reference value encoder uses the signature LZ78
`features of 1) reading the next input character; 2) adding the next
`input character to the current prefix string to build a new,
`combined string; 3) searching the dictionary to see if the new
`string is found; 4) if so, continuing by updating the prefix string,
`building the new string by adding one character at a time, and
`continuing to search the dictionary until the new string is not
`
`15
`
`
`
`IPR2016-00783
`Patent 6,597,812 B1
`
`found; 5) when the new string is not found in the dictionary,
`outputting the reference value for the prefix string and adding the
`new string to the dictionary. As in LZW, O’Brien’s encoder also
`includes features of initialization of code words (character
`reference values) for individual characters and setting the prefix
`string (referenced by last RV) to be the last input character when
`a combined string is not found in the string table. As such, it is
`a classic LZW dictionary encoder.
`
`Pet. 15–16 (citing Ex. 1005 ¶ 40) (emphases added).
`Moreover, Petitioner presents a table comparing the similarities
`between LZ78 compression, LZW compression, O’Brien’s compression, and
`that of the ’812 patent, and contrasting LZ77 (i.e., Lempel and Ziv’s paper in
`1977) compression with the aforementioned compression techniques. Reply
`14. Petitioner’s table is reproduced below.
`
`Petitioner’s table, above, illustrates three characteristics: “sliding
`window dictionary”; “initializes with all possible single character strings”;
`and “builds dictionary by adding new strings.” Petitioner argues its table
`illustrates that LZ77 is unlike O’Brien in that LZ77 has a sliding window
`
`
`
`16
`
`
`
`IPR2016-00783
`Patent 6,597,812 B1
`
`dictionary, whereas O’Brien does not. Reply 14–15, 17. In addition,
`Petitioner explains the table illustrates that LZ77 does not initialize its
`dictionary with all possible single character strings, whereas O’Brien does.
`Id. at 15–17. Petitioner also points out that its table illustrates LZ77 does
`not build its dictionary by adding new strings, whereas O’Brien builds its
`dictionary by adding new strings. Id. at 16–17.
`Furthermore, Petitioner argues its table illustrates that LZ77 is unlike
`the ’812 patent in that LZ77 has a sliding window dictionary, whereas the
`’812 patent does not. Id. at 14–15, 17. In addition, Petitioner explains the
`table illustrates that LZ77 does not initialize its dictionary with all possible
`single character strings, whereas the ’812 patent does. Id. at 15–18.
`Petitioner also points out that its table illustrates LZ77 does not build its
`dictionary by adding new strings, whereas the ’812 patent does build its
`dictionary by adding new strings. Id. at 16–17.
`Patent Owner argues that O’Brien’s dictionary encoder is “more akin
`to an LZ77 dictionary encoder.” PO Resp. 16 (citing Ex. 2007 ¶¶ 82–83,
`85–86). In particular, Patent Owner argues that rather than reinitializing the
`dictionary when the dictionary is full and maintaining a dictionary when the
`dictionary is not full, O’Brien’s system, instead, partitions the input data into
`segments and uses a separate dictionary for each segment, which is similar
`to the functioning of an LZ77 dictionary encoder. Id. at 17 (citing Ex. 2007
`¶¶ 74–75). According to Patent Owner, O’Brien’s segmenting is a basic
`principle of its operation and can prevent its dictionary from becoming full,
`which ensures the dictionary does not impact negatively the encoding speed
`that data compression requires. Id. at 17–18.
`
`17
`
`
`
`IPR2016-00783
`Patent 6,597,812 B1
`
`Moreover, Patent Owner argues its declarant, Mr. Laub, states that Dr.
`Creusere’s classification of O’Brien’s dictionary encoder as both an LZ78
`and an LZW dictionary encoder is flawed for several reasons. PO Resp. 18–
`21. First, Patent Owner argues that Petitioner is conflating LZ78 and LZW
`because LZW initializes the dictionary with all 256 possible single
`characters, whereas LZ78 starts with an essentially empty dictionary and
`builds the dictionary out of previously seen symbols in the input data. Id. at
`14–15, 19 (citing Ex. 2007 ¶ 78).
`Second, Patent Owner argues that O’Brien is neither an LZ78
`dictionary encoder nor an LZW dictionary encoder. PO Resp. 19. Patent
`Owner argues O’Brien is not an LZ78 dictionary encoder because O’Brien
`does not use a single dictionary for the entire input data, whereas an LZ78
`dictionary encoder can use the dictionary for an entire input data stream. Id.
`at 14, 19–20. Patent Owner further argues that O’Brien is not an LZ78
`dictionary encoder because O’Brien segments its data similar to a text
`window, whereas the LZ78 dictionary encoder abandons the text window.
`Id. at 14, 19.
`Patent Owner also argues O’Brien is not an LZW dictionary encoder
`because O’Brien does not initialize a dictionary to include all 256 possible
`single characters, whereas an LZW dictionary encoder does initialize a
`dictionary. Id. at 15, 20. And, Patent Owner argues O’Brien is not an LZW
`dictionary encoder because O’Brien does not use a single dictionary for the
`entire input data, whereas an LZW encoder can use the dictionary for the
`entire input data stream. Id. at 14, 20.
`Third, Patent Owner argues that Petitioner’s interpretation of O’Brien
`is wrong because O’Brien looks at short pieces of input data; much like an
`
`18
`
`
`
`IPR2016-00783
`Patent 6,597,812 B1
`
`LZ77 dictionary encoder’s sliding window. Id. at 13–14, 20–21. According
`to Patent Owner, both O’Brien and an LZ77 dictionary encoder limit their
`dictionaries to the contents of those short pieces. Id. at 20–21.
`Patent Owner also argues that third parties having no interest in the
`present proceeding and Mr. Laub’s opinion characterize O’Brien’s encoder
`as an LZ77 dictionary encoder rather than as an LZ78 or an LZW dictionary
`encoder. PO Resp. 21–22 (citing Ex. 2010, 2; Ex. 2011, 1). These third
`parties and Mr. Laub characterize U.S. Patent No. 4,988,998 (“the ’998
`patent”), which is also issued to O’Brien, as an LZ77 compression
`technique. Id. at 21–22 (citing Ex. 2007 ¶ 84). We disagree with Patent
`Owner.
`At the outset, the instituted claims of the ’812 patent lack any implicit
`or explicit recitation of the type of dictionary compression. Instead, the
`instituted claims set forth the particular actions the claimed compressor
`requires. That said, whether O’Brien is more akin to LZ77, LZ78, or LZW
`is irrelevant and not dispositive to the present case. The important factor is
`whether O’Brien’s actions to compress the input data are the same as the
`actions required by claim 1. We believe the Petition shows that O’Brien’s
`actions are the same actions required by claim 1.
`Nonetheless, we analyze the parties’ argument regarding what type of
`compression is most similar to O’Brien’s type of compression because
`resolving the parties’ issue is helpful in determining whether O’Brien’s
`compression type is similar to Nelson’s compression type, as we will discuss
`infra in §§ II.D.5., II.D.6., II.D.7., and II.D.8.
`Although both an LZ77 dictionary encoder and O’Brien look at short
`pieces of input data (see PO Resp. 20–21), we agree with Petitioner that
`
`19
`
`
`
`IPR2016-00783
`Patent 6,597,812 B1
`
`there are numerous significant departures between an LZ77 dictionary
`encoder and O’Brien (see Reply 14). First, LZ77 is unlike O’Brien in that
`LZ77 has a sliding window dictionary, whereas O’Brien does not. Reply
`14–15, 17. Second, LZ77 does not initialize its dictionary with all possible
`single character strings, whereas O’Brien does. Id. at 15–17. Third, LZ77
`does not build its dictionary by adding new strings, whereas O’Brien builds
`its dictionary by adding new strings. Id. at 16–17. Fourth, an LZ77
`dictionary encoder has a fixed size dictionary, whereas O’Brien does not.
`Id. at 18. Fifth, LZ77 does not process fixed sized segments of input data,
`whereas O’Brien processes mostly fixed sized segments. Id.
`Moreover, Mr. Laub’s statement that O’Brien is more akin to an LZ77
`dictionary encoder rather than an LZ78 or LZW dictionary encoder because
`both an LZ77 dictionary encoder and O’Brien look at short pieces of data is
`flawed. This flaw exists because the sliding window of an LZ77 dictionary
`encoder is a window of already processed input data that operates as a
`dictionary, whereas O’Brien’s segments are unprocessed chunks of input
`data. Id.
`Patent Owner and Mr. Laub’s reliance on Exhibit 2010 and 2011 (see
`PO Resp. 21–22; Ex. 2007 ¶¶ 83–84) is misplaced because those Exhibits
`discuss a different patent rather than the ’812 patent at issue in this
`proceeding. Exhibit 2010 is a PCT application and Exhibit 2011 is a blog,
`and both state that the ’998 patent issued to O’Brien uses LZ77 compression
`(see Ex. 2010, 2; Ex. 2011, 1); however, the O’Brien patent at issue in this
`proceeding is the ’812 patent – a differ