I just assumed that anyone who had worked in software for any length of time was aware of the current minefield that is software development. Every field I've worked in (proxies and security, scientific computing, large-scale cluster computing, web service, web-based GUIs, virtualization) have had at least a few seriously dangerous and over-broad patents that loom over the whole field and make everyone a little nervous or forces workarounds or closing ones eyes and pretending to not know and hoping you don't get sued.

It is simply impossible to write non-trivial software without violating some patent or other. If you're lucky, it will be patents held by "good" companies that won't sue you.

The amount of harm done by software patents is vast, and the amount of good is pretty hard for me to spot.

Anyway, I've sent off my email in support of excluding software from patentability, though I have my doubts as to its effectiveness.

The FSF's suggested text is:

Software patents hurt individuals by taking away our ability to control the devices that now exert such strong influence on our personal freedoms, including how we interact with each other. Now that computers are near-ubiquitous, it's easier than ever for an individual to create or modify software to perform the specific tasks they want done -- and more important than ever that they be able to do so. But a single software patent can put up an insurmountable, and unjustifiable, legal hurdle for many would-be developers.

The Supreme Court of the United States has never ruled in favor of the patentability of software. Their decision in Bilski v. Kappos further demonstrates that they expect the boundaries of patent eligibility to be drawn more narrowly than they commonly were at the case's outset. The primary point of the decision is that the machine-or-transformation test should not be the sole test for drawing those boundaries. The USPTO can, and should, exclude software from patent eligibility on other legal grounds: because software consists only of mathematics, which is not patentable, and the combination of such software with a general-purpose computer is obvious.

But if (like most of us) you write A program, then its a program, not an equation or a theorem or anything else mathematical.

Really? http://en.wikipedia.org/wiki/Curry-Howard_isomorphism

How about a disk driver? a network protocol? A game where you shoot a bunny with a flamethrower?

Some Computer Scientists occasionally gain benefit from applying math to certain algorithmic software. Is that the only kind that is being patented?

Undoubtedly there will be many comments emphasizing the bad effects of software patents, and that for policy reasons they should not be accepted. While I agree with this argument, I see other fundamental problems with applying patents to software, and write to discourage their use.

I instead want to stress that software is not a machine, but instructions, an algorithmic, step-by-step, description of processes. This core nature of software means that even when the process being described by the software is patented, this should not hinder the distribution of the software. A patent application is itself much the same thing -- a description of the process to be covered. It would be utterly ludicrous to forbid the distribution of an approved patent application for violating the patent. The description is not the process itself, nor is it a machine for performing the process. The same really holds true programs. Imagine taking a patent application and annotating with details of a particular way of implementing the patent. At no point does it make sense to forbid the distribution.

It is, of course, perfectly reasonable to have an otherwise patentable invention be partially implemented in software. I must argue though that what violates the patent is the whole machine (including the software, to be sure), and not the software itself, nor the general purpose computer itself. Either should be free to be built, used, sold, and otherwise distributed. This is not much different from parts in an invention being illegal to combine in a way that violates the patent, while perfectly legal to have apart or combine in other ways.

I will also note that in practice, many software patents have been granted that seems obvious. Combining something with a general purpose computer should never have been considered non-obvious, nor should doing something with a computer network as an intermediary be considered non obvious. Computers were built to perform algorithmic processes. Selecting a particular algorithm may not be obvious, but the use of a computer to execute it certainly is obvious. Computer networks were built to carry general information; making that type of information more specific to a certain use does not make the use of a general network less obvious. Many machines are adapted to a specific used, and must be readapted to be used in a different, but similar circumstance. Computers and computer networks are different. They are general purpose machines where the general really does encompass all specifics.

Microsoft, IBM, Oracle, etc. companies that make money from legitimate sources would be barely effected at all (and many tech companies who have huge patent portfolios have spoken in favor of ending software patents or making them less broad). And, startups could breathe a sigh of relief and stop worrying about being the subject of extortion the day they turn a big enough profit to be a target of patent trolls.

Besides, this is about how the USPTO grants software patents. It's not about revoking them or abolishing existing patents, though lawsuits to break some of the more pernicious ones probably would follow.

I'd need to see some evidence of this assertion. When I was spending a lot more time researching tech stocks than I do now (now I spend 0 time researching tech stocks), I never saw a patent portfolio as a significant portion of the value of the companies I was investing in. Revenue and growth is king in tech company stocks. Book value is pretty much considered a big goose egg in software companies because they tend to have so little of it compared to old blue chips that have factories, brick and mortar stores, or infrastructure that has real world value. I'd be absolutely stunned if marking down their entire patent portfolio had more than a 1% impact on their market value, because I suspect even less than 1% of their revenue comes from patents...and revenue and growth is how people value MSFT.

And what does "aggressively" mean? Making a patent unenforceable will have the same effect as invalidating it.

So they'll be making the same money, with less legal risk. If the market is silly enough to crash in response to that, I'll hustle out to buy as many shares in the crashed companies as I can.

The USPTO should consider economic ramifications of a decision to stop issuing software patents, or to invalidate patents. I'm not saying they should not go ahead and do it, but it would be irresponsible not to consider any potential negative effects.

This is an opportunity to tell the USPTO that software patents are not sensible, as they are currently being granted. It is not about "ending software patents" per se, but making them harder to obtain...and now is the best possible time to say something about it, because the supreme court has said, "Yeah, this whole software patent thing probably doesn't exactly make sense."

I was short on time, so it is longer than I hoped. However, I think it gets my core feelings across.

Seriously though, does anyone really believe that writing a letter will get something done? Change in Washington, especially for esoteric topics like this, has a cost, and that cost is tribute paid to lobbyists.

Perhaps not, but a large enough flood of letters from actual software developers may slightly nudge USPTO policy for the better until someone lobbies Congress to force the USPTO to change.