|
Authored by: Anonymous on Monday, May 07 2012 @ 07:42 PM EDT |
In the Android world, there are 4 major passes over the
code (some may
consist of multiple internal passes):
- On the developer's PC, the
SUN/ORACLE Java compiler
(Google accepts no other compiler for Android 2.x)
converts
each .java source code file to Java SE byte code in
a
.class file, which includes symbolic references
to items in other in
APIs, in other files and possibly even
some in the same file.
- On the
developer's PC, the Google dx
tool converts the .class
files into .dex
files with Dalvik byte code, possibly combining
multiple
.class files in the process. Each resulting
.dex
file still contains symbolic references to APIs, other
.dex files and
possibly even the same .dex
file, but the number of references may be
different than in
the .class files.
- On the phone, the Google
dexopt tool
converts all the symbolic references to pointers in
accordance with where the various APIs etc. are loaded on
that particular
phone with that particular firmware build
and other installed
software.
- On the phone, the Dalvik virtual machine actually
runs the code with the pointers.
Google is arguing that the
patent only applies when step
3 and 4 are combined into one step, like it
allegedly is in
the Sun Java SE VM, but not when those are separate steps
like
on Dalvik.
It is worth noting that the 4 step process above is
identical in all but names to the process when creating and
using a native
code DLL on pre-1993 versions of Microsoft
Windows and on any version of
Solaris (I am deliberately not
discussing younger OSes here, as patents are all
about age
and who did it more than 20 years before Android). Just
replace
.java with .c, .class
with .obj or
.o, .dex with
.exe or (no extension), then
replace
javac with cl.exe/cc,
dex with
link.exe/ld,
dexopt with the loader in
KERNEL.EXE/ld.so and Dalvik with native
CPU execution.
[ Reply to This | Parent | # ]
|
|
Authored by: Anonymous on Monday, May 07 2012 @ 08:19 PM EDT |
And the PTO Affirmed this patent?[ Reply to This | Parent | # ]
|
|
Authored by: Anonymous on Tuesday, May 08 2012 @ 02:18 AM EDT |
Where Gosling's innovation stands as a patentable innovation
over prior art is
in the interleaving of symbolic reference
resolution with code execution in a
stepwise, alternating,
realtime way during execution of the byte code. That
was
rather innovative (unless pcode did it? not sure).
And ... Dalvik
does not do it that way. It resolves
the references prior to real time
execution. So the
prior art applies, not to the patent, but to the application
of the patent here. I wonder if the jury can sort that out?
:). [ Reply to This | Parent | # ]
|
|
|
|
|