|
|
|
| Authored by: PolR on Tuesday, May 22 2012 @ 09:24 PM EDT |
The term of art does not sufficiently define the construction. Even with this
understanding there are several possible constructions.
The patent is about resolving symbolic references dynamically, when the virtual
machines instructions are executed. We need a construction which allows for a
symbolic reference which is present in the bytecode during run-time.
I can see several possibilities. The main one is to insert the character string
inline, say right after the instruction opcode. The main alternative is to have
a pointer to a character string. Or we may have an index to a table where the
character string is located. In all of these three cases we have a reference to
a character string which must be resolved using the very kind of symbol tables
you describe. The three alternatives meet the term of art construction in some
way. But only the inline version does not involve some numeric reference of some
sort. If we allow the other interpretations a reference can be both numeric and
symbolic. Then where do we stop? How many indirections are we allowed before we
conclude the reference is not symbolic anymore? I think this is what the jury
was struggling with. Google's code allowed to reach a symbol if enough
indirections were followed and this symbol had to be resolved in a symbol table
exactly like the term of art requires.
I think the judge settled the issue today when he said a reference cannot be
both a number and a symbol. Then the string must be inline. But before he
decided this I think there was room for discussion.
[ Reply to This | Parent | # ]
|
|
|
|
|
