|
Authored by: Anonymous on Tuesday, May 15 2012 @ 09:46 PM EDT |
I think Van Nest summarized them accurately in his closing.
You can also
read them online, and form your own opinion:
US Patent RE38,104 (the '104
patent) covers a simple memoization optimization in VM instructions.
Unfortunately for Oracle, it requires the symbolic reference to be "contained
in" the instructions -- which means it not only doesn't cover the Dalvik VM, it
doesn't even cover a Java VM either! Its a completely worthless patent, I think
when they were applying for it the lawyers must have mangled its meaning a
little too much, and nobody noticed until this trial. (I noticed it a few days
ago and posted about it here, but I think Google's legal team must have noticed
it quite a while ago.. or perhaps it was Dr. August who pointed it out to
them?)
US Patent
6,061,520 (the '520 patent) covers an old kludge called "clinit
optimization". This is the one that requires simulation of the array
initialization bytecodes to work out their effects. But Dalvik doesn't use
simulation for its optimization of array initialization -- it uses a much
simpler technique, pattern matching, instead. [ Reply to This | Parent | # ]
|
|
Authored by: bugstomper on Wednesday, May 16 2012 @ 06:22 AM EDT |
I have read both patents, as well as the report of the re-examiner at the PTO
that rejected almost all of the claims on the basis of prior art submitted by
Google.
Your instincts on this are correct. Most of the '104 patent (the one about
symbolic references) was invalidated by the re-examiner based on prior art in
the classic 1971 textbook on compilers by Gries. Google was not allowed to
mention that in this trial because the rejections are still not final, being
subject to appeal by Oracle.
Yes, it does raise the question as to how the PTO could grant such a patent in
the first place.
[ Reply to This | Parent | # ]
|
|
Authored by: Anonymous on Wednesday, May 16 2012 @ 06:50 PM EDT |
I have prior art from the 1970's:
The macro assembler, link, and the operating system job loader in the RT-11
operating system performed resolution from symbolic names to numbers to
addresses in both a static form and a dynamic form.
The static form was for programs loaded at a fixed address (0) and the dynamic
was for programs loaded at arbitrary locations.
Dynamic loaded operating system components ("handlers") also had
hand-coded relocation using a standard format.
This is the reason software patents should be discontinued, there is almost
always prior art.
/s/ Jim Williams
jfw.rt11@gmail.com[ Reply to This | Parent | # ]
|
|
|
|
|