On the '520 patent the issue centered on the meaning of "simulating execution." Once again, Google's expert testimony, this time from Dr. Parr, was persuasive in establishing that Android does not simulate execution; rather, it relies upon pattern matching. Again, this pretty much shot holes in Oracle's arguments. As a result, the Court denied the motion with respect to the '520 patent, as well.
The Court took one final shot at Dr. Mitchell in its conclusion:
IN THE UNITED STATES DISTRICT COURT
ORACLE AMERICA, INC.,
FOR THE NORTHERN DISTRICT OF CALIFORNIA
No. C 10-03561 WHA
ORDER DENYING ORACLE’S
MOTION FOR JUDGMENT
AS A MATTER OF LAW RE
In this patent and copyright infringement action, Oracle moves for judgment as a matter
of law that Google infringed Patent Nos. RE38,104 and 6,061,520. For the reasons stated below,
the motion is DENIED.
The procedural and technological background of this action has already been discussed
(Dkt. Nos. 137, 433). This action began in 2010 with seven asserted patents. In May 2011, a
claim construction order was issued (Dkt. No. 137). Neither party filed a motion for
reconsideration of that order. In January 2012, at the request of the parties, a supplemental claim
construction order for additional terms was issued (Dkt. No. 704).
By the time of trial in April 2012, Oracle had reduced the number of asserted patents and
claims to two patents and eight claims. This was done as an incentive to receive an earlier trial
date and because the withdrawn patents had received “final rejections” in PTO reexaminations.
The issue of infringement was tried to a jury and Google dropped all affirmative defenses,
including invalidity. The parties stipulated that indirect infringement (inducement and
contributory infringement) would automatically follow from a finding of direct infringement
(Dkt. No. 1139). Because Oracle had not disclosed an infringement theory under the doctrine of
equivalents, the only issue for the jury to decide was whether Android literally infringed the
asserted claims in the ’104 and ’520 patents.
The parties also narrowed down the claim limitations in disputed (TX 1106). For
the ’104 patent, the dispute was whether Android met the limitation of having “instructions
containing one or more symbolic references.” Claim 11 is a representative example of the claim
An apparatus comprising:
(TX 1106) (for illustration, the disputed limitations are underlined). All asserted claims for
the ’104 patent contained the limitation that the “symbolic reference” must be in the instructions.
Google stipulated that Android contained the non-underlined limitations in the asserted claims.
For the ’520 patent, the dispute between the parties was whether Android satisfied the limitation
of “simulating execution . . . to identify the static initialization of the array.” Claim 1 is a
representative example of the claim language:
a memory containing intermediate form
object code constituted by a set of
instructions, certain of said instructions
containing one or more symbolic references;
a processor configured to execute said
instructions containing one or more
symbolic references by determining a
numerical reference corresponding to said
symbolic reference, storing said numerical
references, and obtaining data in accordance
to said numerical references.
A method in a data processing system for statically
initializing an array, comprising the steps of:
compiling source code containing the array
with static values to generate a class file
with a clinit method containing byte codes
to statically initialize the array to the static
(col. 9). All asserted claims for the ’520 patent contained the “simulating execution” limitation.
receiving the class file into a preloader;
simulating execution of the byte codes of the clinit
method against a memory without executing the
byte codes to identify the static initialization of the
array by the preloader;
storing into an output file an instruction requesting
the static initialization of the array; and
interpreting the instruction by a virtual machine to
perform the static initialization of the array.
After a week-long trial, followed by another week of jury deliberation, the jury rendered
a verdict of non-infringement for all asserted claims. Oracle moves to set aside the jury verdict
as a matter of law under Rule 50.
Judgment as a matter of law may be granted against a party if it “has been fully heard on
an issue and there is no legally sufficient evidentiary basis for a reasonable jury to find for that
party on that issue.” Rule 50(a). “Judgment as a matter of law may be granted only where, so
viewed, the evidence permits only one reasonable conclusion, and that conclusion is contrary to
the jury’s verdict.” Wallace v. City of San Diego, 479 F.3d 616, 624 (9th Cir. 2007).
The final charge to the jury was and remains the complete statement of the law governing
the trial issues unless a timely and proper objection to the instructions was made and the Court
now agrees that the objection has merit. The post-trial motions under Rule 50 must be evaluated,
therefore, in light of that statement of the law and not in light of new citations, new legal theories
and/or legal variations. Rule 50 is not an occasion for yet another round of summary judgment
based on new slants on the case law. Of course, an appellate court might later find that an
objection to an instruction was preserved, well-taken and not harmless and so require a new trial.
But until then, this Court is satisfied that the jury instructions were proper and both sides are
bound by the jury instructions as the exclusive statement of the governing law for the
Where a party does not object to the court’s jury instruction regarding the construction of
claim term, the verdict must be tested by the charge actually given under the ordinary meaning of
the language of the jury instruction. Cordis Corp. v. Boston Scientific Corp., 658 F.3d 1347,
1355–56 (Fed. Cir. 2011).
[P]arties cannot reserve issues of claim construction
for the stage of post-trial motions. When issues of
claim construction have not been properly raised in
connection with the jury instructions, it is improper
for the district court to adopt a new or more detailed
claim construction in connection with the JMOL
motion. On JMOL, the issue here should have been
limited to the question of whether substantial
evidence supported the verdict under the agreed
instruction. In other words, where the parties and
the district court elect to provide the jury only with
the claim language itself, and do not provide an
interpretation of the language in the light of the
specification and the prosecution history, it is too
late at the JMOL stage to argue for or adopt a new
and more detailed interpretation of the claim
language and test the jury verdict by that new and
more detailed interpretation. The verdict must be
tested by the charge actually given and by giving
the ordinary meaning of the language of the jury
Hewlett-Packard Co. v. Mustek Systems, Inc., 340 F.3d 1314, 1320–21 (Fed. Cir. 2003)
1. THE ’104 PATENT.
As discussed, the dispute with respect to infringement of the ’104 patent was whether
Android’s Dalvik bytecode instructions contained “symbolic references.” More specifically, the
question for the jury was whether the field indices and operands in the Dalvik bytecode
instruction stream were “symbolic references.”
A. Meaning of “Symbolic Reference.”
The jury instructions contained the Court’s May 2011 construction of the term “symbolic
reference,” which was defined as “a reference that identifies data by a name other than the
numeric memory location of the data, and that is resolved dynamically rather than statically”
(Dkt. No. 137 at 22; Dkt. No. 1153 at 5). In formulating this definition, the order did not adopt
either parties’ proposed definition. During the May 2011 Markman proceeding, Oracle had
argued that the term “symbolic reference” did not need construction and that the plain meaning
meant “a reference by name.”
Importantly, Oracle did not object to the use of the term “data” in the Court’s
construction or request additional clarification before the claim construction’s definition went
into the final jury instructions. During the charging conference, the Court specifically asked
whether Oracle wanted to construe the term “data.” Oracle declined.
COURT: No one has asked for any other claim
constructions. If you want me to put in a sentence
that would say as to anything else it has its plain
meaning as used in the patent, I can do that. But I
don’t — you can’t sandbag me at the last minute
and say, Judge, please give us a ruling on what
“data” means. That would take me some time to
sort out. So if — you know, this was the time to
ask me for it. I guess you could — for good cause
you could ask for it later. But, it would have to be
really good cause to say at the last minute you want
me to go research a legal point and do a claim
construction. But I hear you saying you’re happy
with the plain language.
(Tr. 3710–11). Oracle’s answer was an unequivocal “no.” Therefore, the ordinary meaning of
the term “data” in the jury instructions controls. See Cordis Corp., 658 F.3d at 1355–56.
ORACLE’S COUNSEL: We’re happy with the plain
language of the claim and the claim constructions.
Additionally, the jury asked a question regarding the meaning of symbolic reference
during their deliberations. The jury asked whether a reference can be both a numeric reference
and a symbolic reference. Over Oracle’s objections, the Court answered:
You look at whatever is in that little box and you
say, does that identify data by a name other than the
numeric memory location of the data? If the answer
to that is that it identifies a numeric memory
location of the data, then it is not a symbolic
reference; it's a numeric reference. And it doesn't
get transmogrified into a something else on account
of what happens downstream.
But you do have to consider what happens
downstream to find — to at least figure out whether
or not the item that is in that box is referring to the
numeric memory location of the data. You’ve got
to look at least that far downstream in order to see
what it’s referring to. But, once you do that — and
if you decide it’s a symbolic reference, great. If you
decide it’s a numeric reference, that’s great, too.
But it doesn’t then change because the computer
marches on to one instruction after the other, and so
(Tr. 4325–26). To sum up, the Court answered that a reference cannot be a “symbolic reference”
for one datum and a “numeric reference” for another datum within the meaning of the claim
This answer was appropriate and already inherent in the claim construction. Cordis
Corp., 658 F.3d at 1356. Based the patent specifications, a reference cannot be both a “symbolic
reference” and a “numeric reference” within the meaning of the claims. The patent expressly
stated that a reference was either a “symbolic reference or a numeric reference” (col. 2). The
patent described two different subroutines that were invoked, depending on whether the
reference was numeric or symbolic. There was nothing in the patent describing how to first
determine whether a reference was numeric or symbolic with regard to any particular datum.
The only reasonable understanding of the patent was that a reference would be recognized by an
interpreter as either numeric or symbolic but not both. This understanding was inherent in the
Court’s claim construction order, which stated that “[t]he ’104 patent teaches two different types
of data references: numeric references and symbolic references” (Dkt. No. 137 at 20).
Furthermore, the testimony at trial of Oracle’s own expert, Dr. John Mitchell, supports
this mutually exclusive characterization of a reference as either a numeric or symbolic reference.
In his opening expert report, Dr. Mitchell labeled the indexes in the Dalvik bytecode instructions
as “numeric references.” Realizing that this could be problematic to Oracle’s infringement
contentions, in his trial testimony, Dr. Mitchell testified that his report was mistaken and that
those indexes were instead symbolic references (see, e.g., Tr. 3490). This 180-degree change of
opinion proves the mutually exclusive characterization of a reference (within the meaning of the
patent claims): if a reference could be both a numeric and a symbolic reference, then
Dr. Mitchell would not have needed to repudiate his expert report at trial and instead could have
supplemented his report by testifying that the indexes were both numeric and
B. Sufficient Evidence to Support Jury’s Verdict.
There was sufficient evidence at trial to support the jury’s finding of non-infringement.
At trial, the Dalvik bytecode instruction was the focus of both parties’ arguments. In particular,
the parties focused on the IGET instruction. The IGET instruction retrieved a piece of data out
of an instance field object (Tr. 3221). The IGET instruction contained an operand —
field@CCCC — also known as the field index (Tr. 3221). Android resolved class, method, and
string indices in the same way as the field indices and both parties agreed that the field index
example was dispositive for all indices.
Oracle’s theory of infringement presented to the jury was that the field indices —
operands in Davlik instructions — were symbolic references. In defense, Google argued that the
same field indices were not symbolic references but were instead, numeric references, because
they were pointers to table entries from the Field ID tables (a table in the constant pool of
Android dex files). The jury sided with Google’s argument.
There was ample evidence to support the jury’s verdict of non-infringement. Google’s
expert Dr. August testified that the term, “data” from the Court’s claim construction of
“symbolic references” would have been understood to include data from Field ID tables in the
dex files. Dr. August illustrated with an example of Android’s IGET instruction “52 01”, which
instructed the computer to go to entry 01 in the Field ID table. Dr. August explained that the
operand (“01”) for the IGET instruction qualified as a numeric reference because it referred to
data by a numeric memory location: entry 01 in the Field IDs table. That is, the operand “01”
identified data (the value at entry 01 in the Field ID table) in the constant pool by the numeric
memory location, entry 01.
The premise being that the certain of
said instructions containing one or
more symbolic references.
And you’re saying that — well, what
is “01”? Is that, in your view, a
No. “01” is a numeric reference
because it gives you an actual
numeric memory location. Right
(Tr. 3865). His testimony took Android’s operand “01” out of the plain meaning of the Court’s
“symbolic reference” definition, and thus, the claim limitations of the ’104 patent. The jury was
reasonable in crediting Dr. Parr’s testimony.
here (indicating). This is the
There’s no resolution, no search.
Nothing – nothing expensive about
figuring out what that instruction is
referring to when it goes to the field
Oracle’s argued (and continues to argue) that “data” in the claim construction and claims
themselves referred narrowly to “the ultimate data to be obtained or used after symbolic
reference resolution is performed” (Dkt. No. 1189 at 1). This argument was reasonable rejected
by the jury. The values in the Field ID table in the Android dex file were “data” in the ordinary
sense of the word. The jury was reasonable to adopt this understanding of the term “data” and
reject Oracle’s expert Dr. Mitchell’s opinion that “data” was restricted to “ultimate data” in
Android. Google’s Dr. August persuasively explained that there was no limitation regarding the
type of data being referenced in the plain meaning of the Court’s claim construction. He
explained on cross-examination (Tr. 3955):
Even adopting Oracle’s belated effort to construe “data” to mean “the ultimate data to be
obtained or used after symbolic reference resolution is performed,” a reasonable jury could still
find that Android did not infringe. Dr. August testified that the field data in the instance object
referenced by the operand (Oracle’s definition of a “data” in Android) may itself be a reference
to other data. For example, that data may reference another string in the string data table.
Now, just to clean up a few other
things. You labeled this ‘data,’ but
this is actually the constant pool;
Constant pool is data.
What can be in this instance object
by way of data?
Well, it could be a symbolic
reference. It could be a numeric
(Tr. 4002–03). Thus, the jury was reasonable to believe that the operand “01” in Android’s
instruction stream did not identify the “ultimate” data within the meaning of the claim language.
* * *
Let’s talk about in Dalvik. Is the
data at these entries always nonreference
data, as we talked about in
your technical tutorial?
No, it’s not.
How do you know that?
Because I write programs. And you
can write programs that contain
references in instances of objects.
Oracle also argues Android’s operand “01” must be a symbolic reference because it was
resolved and only symbolic references had to be resolved. Even if this construction were
adopted, the jury was still reasonable to find that Android did not infringe. There was testimony
that Android’s field indices were not resolved. As Mr. Andy McFadden explained:
And are indexes in this process being
Why do you say no?
Well, resolution implies something is
unknown, something is ambiguous.
If you have an index, you know
exactly where you’re going. You
have the location. For symbols, you
don’t know where you're going and
you won’t know until you have
resolved the ambiguity. So it doesn’t
really make sense to say that you
“resolve an index.”
Therefore, Oracle’s motion for judgment of infringement of the ’104 patent is DENIED.
2. ’520 PATENT.
Oracle argues that Android SDK’s dx tool infringes claims 1 and 20 of the ’520 patent as
a matter of law. At trial, the only dispute with respect to infringement of the ’520 patent was
whether Android’s pattern-matching technique in its dx tool was encompassed in ’520 patent
“simulating execution” process limitation. There was no dispute that the dx tool used a
technique called “pattern matching” that looked for a repeating sequence of instructions used to
initialize static arrays (Tr. 3800–01). The jury was reasonable to credit the testimony of
Google’s expert, Professor Terence Parr, who testified that Android’s pattern-matching
technique was not simulating execution because it did not go through the steps required to
produce output showing what the static array would look like if the Java byte code instructions
A. Meaning of “Simulating Execution.”
Neither party requested construction of the term “simulating execution.” Neither party
sought to clarify the term in the jury instructions. Therefore, the verdict must be tested under the
ordinary meaning of “simulating execution.” Cordis Corp., 658 F.3d at 1355–56.
The purpose of the ’520 patent was to efficiently create an instruction for static
initialization of an array for the Java virtual machine, a stack-based machine. The specification
discussed static initialization only through simulating actual Java virtual machine operations:
manipulation of a stack by pushing, popping, and replacing values from the top of an operand
stack. Thus, the ordinary meaning of “simulating execution” is the simulation of actual Java
virtual machine operations. Google’s expert, Dr. Parr, also testified that this was the plain
meaning that would have been understood by those skilled in the art (Tr. 3794).
B. Sufficient Evidence to Support Jury’s Verdict.
It was undisputed that Android did not manipulate a stack to identify the initialization of
an array. At trial, Professor Parr confirmed this by looking at the source code and also
performing multiple experiments. As an experiment, he inserted an extra piece of bytecode that
put a zero into the zeroth entry in an array — where a zero already existed at the inception of the
array. Such an extraneous piece of code, when run through a program that simulated execution
of byte code instructions, would create an efficient instruction with no problem because the
useless instruction would end up having no effect on the eventual result (Tr. 3810). However, in
his experiment, the dx tool failed to create a proper instruction when it was fed that extraneous
code because it failed to find a particular sequence of instructions through pattern matching (Tr.
3809–10). The jury was reasonable to find that the dx tool code did not simulate execution of
the byte code within the meaning of the claim limitation.
In its Rule 50 motion, Oracle argues that the limitation of “simulating execution” must,
as a matter of law, encompass Android’s dx tool because the dx tool “examines the bytecode
without executing them.” This belated construction of “simulating execution” is rejected. The
jury was free to reject Oracle’s evidence that Android’s pattern matching was within the plain
meaning of “simulating execution.” The plain meaning of “simulating execution” does not
encompass pattern matching, and there was no mention of pattern matching in the ’520 patent
itself (Tr. 3521; TX 4011).
Therefore, Oracle’s motion for judgment of infringement of the ’520 patent is DENIED.
The foregoing is sufficient but it is worth adding that Oracle’s infringement case was
presented through Dr. Mitchell. A reasonable jury could have found his many “mistakes” in his
report merely to be convenient alterations to fix truthful admissions earlier made before he
realized the import of his admissions. For this reason, a reasonable jury could have rejected
every word of his testimony.
For the reasons stated, Oracle’s motion for judgment as a matter of law of infringement
IT IS SO ORDERED.
Dated: May 30, 2012.
/s/ William Alsup
UNITED STATES DISTRICT JUDGE