Authored by: Anonymous on Saturday, May 12 2012 @ 02:57 PM EDT |
27. A computer-implemented method comprising:
receiving a
program with a set of original instructions written in an intermediate form
code;
generating a set of new instructions for the program that contain
numeric references resulting from invocation of a routine to resolve any
symbolic data references in the set of original instructions; and
executing the program using the set of new instructions.
In a
Java VM, there are never any symbolic data references in the set of
original (Java bytecode) instructions. There are symbolic data references in a
Java class, but they are elsewhere (in the constant pool). A couple of Google's
witnesses testified that there aren't any in the Dalvik bytecode instructions
either.
39. A computer-implemented method comprising:
receiving a program formed of instructions written in an intermediate form code
compiled from source code;
analyzing each instruction to determine
whether it contains a symbolic field reference; and
executing the
program by performing an operation identified by each instruction, wherein data
from a storage location identified by a numeric reference is thereafter used for
the operation when the instruction contains a symbolic field reference, the
numeric reference having been resolved from the symbolic field
reference.
I think none of the instructions used in either a Java
VM or a Dalvik VM ever "contains a symbolic field reference".
I guess this
is what happens when you let lawyers run the asylum for a few decades. :P [ Reply to This | Parent | # ]
|
|
Authored by: Anonymous on Saturday, May 12 2012 @ 03:00 PM EDT |
I think when they were applying for this patent, they removed the layer of
details about the constant pool from the claims (trying to make the patent as
broad as they could get away with).
And they accidentally ended up with a
set of claims that don't even cover the invention they were trying to protect!
Now that's karma at work. [ Reply to This | Parent | # ]
|
|
Authored by: Anonymous on Saturday, May 12 2012 @ 03:22 PM EDT |
Okay, so I got an important detail wrong in the above post.
A Java VM
actually does overwrite the entries in the constant pool when it resolves them.
So that part matches the description in the patent.
See VM Spec Loading, Linking and Initializing for another description of how a
Java VM does things.
Where it differs from the patent, is just that in Java
bytecode, there are no symbolic references in the instructions themselves. In
claim 1, there should be something between steps b) and c) describing how they
find the proper symbolic reference that is referred to by the instruction (which
actually contains a numeric index of the entry in the constant pool where that
symbolic reference is).
And if Oracle's lawyers try to claim that the
constant pool entry "is part of" the instruction, Google can rebut that by
pointing out that constant pool entries can be shared by several different
instructions. (That's the other reason why constants are stored in a pool.. so
they can be shared by several different instructions.) [ Reply to This | Parent | # ]
|
- oops - Authored by: Anonymous on Saturday, May 12 2012 @ 03:26 PM EDT
|
Authored by: Anonymous on Saturday, May 12 2012 @ 04:14 PM EDT |
as is done in every OS ?
/Arthur[ Reply to This | Parent | # ]
|
|
Authored by: Anonymous on Saturday, May 12 2012 @ 05:10 PM EDT |
If Google is somehow found to infringe the '104 patent, how will damages be
calculated?
Is Oracle contending that Google has harmed some kind of "real"
Java-on-phones project that they would have done? Can they claim that if it
turns out that Oracle can't actually prove that it practices the '104 patent?
Or do the damages have nothing to do with that?[ Reply to This | Parent | # ]
|
|
Authored by: PJ on Saturday, May 12 2012 @ 07:12 PM EDT |
While you ponder, read what the experts witnesses
said about this, please. Then everyone: let me
know if this is correct, so I will know whether or
not to highlight it in the article.[ Reply to This | Parent | # ]
|
|