I suspect novel grammatical constructs in English would fail part one of the patent-eligibility test: does the idea fall within one of the four permissible classes, those being process, machine, manufacture, or composition of matter. See http://www.uspto.gov/web/offices/pac/mpep/s2106.html. If it was deemed to fall within one of those categories, then it'd still have to qualify as either being non-abstract or, if abstract, having some further technical advance.

> does the idea fall within one of the four permissible classes, those being process, machine, manufacture, or composition of matter.

It would be a process in exactly the way that software is. Obviously the process can be carried out by a human (as can software), but never mind that, we'll just tack on a prior art typewriter or computer input device to make it a machine, right?

This whole mess is a result of the Federal Circuit reading everything too broadly. "Process" traditionally meant a manufacturing process or the like. If I'm not mistaken the only "software" patent the Supreme Court has found valid covered a manufacturing process.

And that's where software patents should live, if anywhere. Covering novel machines that happen to require some computation rather than novel computations that happen to be done by a machine.

The test such patents would fail is utility, which is mostly understood to be industrial, technical or commercial utility. This is the same reason recipes are generally not patentable -- they are, after all a "process of manufacturing something" -- except when they have "utility". A common example is a recipe for cattle feed that can reduce illnesses in cows, which clearly has utility in farming.

That test is unintelligible. Anything has industrial or commercial utility if you can find someone to buy it. A recipe for chicken soup clearly has such utility for Campbells because it's how they make their product. How is that supposed to be distinguished from someone making Playdough or artificial sweeteners, which presumably are patentable if novel and non-obvious?

You just mentioned examples that are very likely patentable. The difference is, you cannot try to patent a recipe if your only "utility" is "it helps my restaurant run better" or "it can help my book sell better". The test is fuzzy, sure, but it's an important grounds for rejection since any invention must be novel, non-obvious and useful. You could try to patent recipes and novel structures, but whether the PTO or the courts will buy your claim to usefulness is another matter.

> You just mentioned examples that are very likely patentable.

That's what I mean. How is the "recipe" for Playdough supposed to be different than a specific recipe for chicken noodle soup?

> The difference is, you cannot try to patent a recipe if your only "utility" is "it helps my restaurant run better" or "it can help my book sell better".

That's the utility of any commercial process. To make the product more desirable to customers. The utility the of process to make Playdough is that it helps you sell the product as compared with some other recipe that turns to stone on the retailer's shelf after a week.

I am not talking about the current state of patent law in the US. I am talking about the patentability of software in general. Not as the US implements it, but as it could ever be implemented, by anyone.