|
Authored by: PolR on Tuesday, May 22 2012 @ 06:23 PM EDT |
> I believe that Compatibility per se doesn't matter if
> programs written before Android can't run on Android
> phones because they weren't written to run on
> Android phones (and may access hardware that isn't found
> on cell phones).
This is excessive. Some libraries would run. Also it is much easier to port a
program over a reasonably compatible implementation of the same program than to
rewrite from scratch in a different language. Compatibility matters even when it
is not perfect.
[ Reply to This | Parent | # ]
|
|
Authored by: Anonymous on Tuesday, May 22 2012 @ 06:28 PM EDT |
Your summary just broke down the whole case to me. Thanks and
many thanks to PJ.
Am curious why the court did not call expert of its own to
make some of this things clear to the jury. Is this not
allowed in the US?
I strongly believe that some cases should not go to jury.
It's been too technical for the jury.[ Reply to This | Parent | # ]
|
|
Authored by: Anonymous on Tuesday, May 22 2012 @ 09:16 PM EDT |
For what I gather and I never wandered through such dense
fog:
* The patent in dispute is about JIT-compiling.
(if it were not, it can be buried in prior art)
* JIT compiling presupposes a stack-oriented target.
* Dalvik is not a stack-oriented instruction set.
---
(1) The idea of attempting execution of something that
contains placeholders (symbols), only to use some of the
information available at that point of execution to resolve
the symbols, is the reason why this runtime translation
needs to happen at all. Maybe the patent is so specific that
even in this case Soraclen went some other route.
(2) Such a late-binding (association of symbol to value) is
useful if we can get enough context information to make
decisions at runtime, that would not be possible (or as
optimal) in a static review of the code. Only the stack of
a stack-oriented machine would provide the proper
granularity. The stack of a register machine is at the
function-invocation level, that of a stack-based machine
records all operations in progress at the individual
instruction level. Further, that information is
conveniently very localized, systematically organized and
"easy" for an optimizer to use.
(3) A Dalvik machine won't, at runtime, have the stack
information available, so the kind of optimization done by
this is out of reach. Dalvik uses a separate compiling pass
to effect its optimizations, rather than runtime. To discuss
which one is better is a religious issue; with available
silicon, Google decided that the Dalvik approach was better.
----
So not only does Dalvik not practice the patent, the
environment the patent expects, in order to be useful, is
absent from the Dalvik translation chain.[ Reply to This | Parent | # ]
|
- No - Authored by: Anonymous on Tuesday, May 22 2012 @ 10:06 PM EDT
- No? - Authored by: Anonymous on Wednesday, May 23 2012 @ 03:20 AM EDT
|
|
|
|