[EgoIncarnate]

This ruling doesn't really change anything with respect to CPU instructions. The fair use defense doesn't cover patents.

Patents are what are generally what is used by Intel, etc to protect (and license) new CPU instructions and provide protection for novel ideas/inventions for up to 20 years.

Copyright generally protects specific expressions/implementations of an idea and last up to 95 years for corporate patents, or 70 years + the lifetime of the author for individual patents.

For completeness there is also trademarks which cover names and logos which can last indefinitely, as long as they are in commercial use.

The text of a CPU instruction specification would be covered by copyright, the algorithm for implementing the instruction by a patent, and the branding (ex: MMX) by trademark.

[Ericson2314]

Say one black-box reimplements x86? Say one makes a "transformative work" with a number of extensions?

I fail to see why an ISA is fundamentally different than a standard library.

[gwd]

> I fail to see why an ISA is fundamentally different than a standard library.

As GP said, the difference is whether it's patented. If Sun had patented parts of the API (or algorithms necessary to implement it), then Oracle would have another weapon against Google even after Google was granted a fair-use defense.

[Ericson2314]

Sure I'm not disagreeing with the legal history, but on what merits is one patentable, and the other either fair use to reimplement or not even copywritable!!

I could understand Intel having a CPU patent for specific CPUs, but an specific ISA?!

A really interesting test case would be to implement an isomorphic encoding to x86 with same instruction widths and what-not such that it's trivial to convert binaries from one to the other, and modify compilers (especially the JIT ones).

[wtallis]

We shouldn't necessarily assume that the x86 patent war chests are legally sound. Rather, Intel and AMD have a mutually assured destruction cross-licensing arrangement and they need each other to stay viable to fend off antitrust regulators. But they would prefer not risking an unfavorable precedent by actually wielding their patents in court, so their deterrent operates more on the promise of protracted and expensive litigation, rather than on the promise that Intel would actually win against an upstart CPU vendor.