WTF is “software patent”, Mikey ?

Did they teach you some intro computer science ?

Like Turing equivalence etc. ?

You know… software is equivalent to hardware…
This is apparently the First Law of computing you should learn…

Do you know what Verilog is ?

How about ASIC or FPGA ?

Get familiar with the subject first, dude..
This is the First Law of journalism, which you haven’t learned either…

Well, you have to keep in mind the distinction between mathematics and science. The truth, falsity, or undecidability of mathematics does not depend on external reality. Science does depend on external reality, but it also depends on imperfectness of perception. The equations physicists (and engineers) write are not reality– they are merely an imperfect model of reality. Immanuel Kant’s old Phenomena/Neumena distinction in short. A mechanical engineer will write equations in Newtonian physics. That doesn’t mean that he denies Einstein– it merely means that he does not expect mechanical parts to move fast enough for the difference between Einstein and Newton to matter. An electrical engineer would of course take a somewhat different view.

Now, if you take a complex digital circuit, say a microprocessor, it is built up out of logic gates. A logic gate is a transistor circuit designed to perform an elementary logic operations. The “inventive step” from the logic gate to the microprocessor is a mathematical, or formal logical process. It may very well, as Angry Dude (#2) points out, take the form of a Verilog file, or or a “gatemap” file for a FPGA, and both of those would qualify as programs. This does not, however, apply to the internal design of the logic gate itself. The internal design of the logic gate itself is concerned not with bits, but with physical quantities such as volts, amperes, ohms, coulombs, farads, etc., There is the breakpoint between software and hardware.

However, we can take this forwards rather than backwards. Traditionally, the expense and complexity of debugging digital circuits has been because bug fixes involved going and etching batches of new chips or circuit boards. Obviously, that is no longer the case when you have FPGA’s to use. To allow an analogy, EEPROM’s were not always EEPROM’s. The first ROM’s were “mask-programmable,” customized during manufacture, and those were followed by PROM’s, which could be written once, blowing fuses, and EPROM’s, which could be erased with ultraviolet light, and finally EEPROM’s and Flash chips. The philosophical status of firmware did not change because it was becoming easier to write and rewrite. The more rigid types of fixed manufacturing have dropped out of circulation. An electronic device is manufactured with EEPROM’s, not because the manufacturer actually expects to go out in the field and reprogram the things, but because he has access to highly economical EEPROM programming machines, and an old-style mask-programmed ROM would be an expensive special order. In the same way, stamping out CD’s does not make sense anymore, when there are these cheap CD-DVD-burning machines. According to this logic, a microprocessor/chipset can reasonably incorporate a certain proportion of FPGA logic, to be configured as needed by the program running on the microprocessor. Beyond the components of the standard computer, you just would not have ASIC’s anymore, because the comparatively small volumes would make them uneconomic.

There are a number of firms which are called “fabless” chip manufacturers, meaning they don’t actually produce chips. Rather, they merely produce Verilog code, which they employ other firms to make up into chips. Fabless manufacturers are in fact proprietary software developers. NVIDIA would be an example, and so was ATI before it got absorbed by AMD. I believe the major wireless chip makers fall in this category. By the standards of conventional software, chip logic is not really very complex, not compared to things like operating system kernels. Such firms are in fact prime candidates for being overtaken by open source, or rather, “Open Cores.” A lot of these firms are in fact making money doing things which are at approximately the level of Google Summer-of-Code projects. They were artificially protected by the traditional chip-etching regime, with its very high cost of “recompiling.” I have found historical instances of early computer programmers (back in the late 1950’s) who were working under a regime where recompiling took about one to two days. Their access to the computer was effectively by mail order, so to speak They would send a program off to the computer within half an hour or so of the start of the work day (9:30), and get the results back within half an hour of closing time (4:30). Depending how quick they were, the programmers might or might not have the revised version done fast enough to send it off the next day. Naturally, their productivity was only about a twentieth of the productivity of someone with his own personal computer. That seems a roughly analogous situation to the fabless chip manufacturers.

Thus, if software patents are invalid for word processors, then they are also invalid for things like processor instruction sets and memory models.