Software patents are patents not just on abstract ideas, but on *pure* abstract ideas. By pure I don’t mean “extremely” — I mean “independent of empirical facts or particulars,” like math.

The general purpose logic processor does *pure* logical operations; it follows that the instructions provided to that processor are likewise *pure* logic. What translates a pure logical algorithm or process to something of specific use is the devices you attach to the logic processor; it is in principle perfectly conceivable that the exact same set of logic instructions could control entirely different processes, just depending on what devices you attach. And while higher-level, human-readable code uses language that humans can relate to specific things or uses in the empirical world (variable names like “PartNo” or “EmployeeID” or “Chemical1Proportion,” or function names like “InitiateStirring” or “TurnOnBlender”), at bottom the compiled code is a pure logical abstraction.

The point is, the software is in principle not patentable, though it may be the case that a particular empirical process that that code is being used to control, is patentable. But what decides that is the empirical process as such, not the pure algorithm that the code expresses.

Note that interestingly, this ruling may also undermine the European “computer-implemented inventions” formula, since it disallows the machine-or-transformation test as solely determining the question, and then it specifically resorts to the abstract idea exclusion. Note also that the Court states that its ruling should not be construed as endorsing the Federal Circuit’s past interpretations of patentable subject matter under section 101.

So I see this ruling as simply excluding the easier resolution to the software patent issue (an approach not really on point and not really addressing the nature of the issue of software patents) that sees software patents becoming untenable along with business methods, as well as the notion that what makes something patentable is that it is a machine or a transformation.

What makes software unpatentable is not really directly related to either of those points, but rather to the fact that it is *pure* abstraction. You don’t have to say patents only cover machines and transformations to eliminate software patents. Instead, all you have to do is show that software is always abstract because the nature of the logic processor is inherently pure — it deals only in bits/numbers and logical and methematical operations and algorithms — and when one says those numbers and algorithms are meant to refer to something more, what you’re really saying is that there’s also an empirical process, alongside the pure code, which is the actual process which may be patentable or not.