Open Source Patents

From StoneHome

As always, <acronym title="I Am Not A Lawyer">IANAL</acronym>. Most of my legal clue comes from a distillation of the better of SlashDot and GrokLaw posts and a few scattered entry-level college classes, making me about as educated as a well-trained Markhov chain. Nonetheless, occasionally those chains pump out interesting things. Here's today's scary thought.

What if the answer to the current patent problems really is the Open Source method?

Oh god, not this again

No, really. I've heard these stupid arguments a billion times too, and just like you, I've never heard one that made any sense to me. But, well, this one sort of does, though it has some significant problems.

Lately I've been looking at opening up a bunch of source I wrote for the GameBoy Advance which never sold, given that the DS is as of this writing one day from release, and that I sort of believe that the platform will be dead-ish by not this but next christmas, and utterly dead the christmas following, which will probably see the last few weirdo holdouts releasing games that didn't get done on time or which were written on tiny budgets (it's like looking at my game in a crystal ball...)

As a result of looking at giving all this crap away, I've had strange license clauses bumping around my head like "you may use this freely, but you must put a note in the manual and the help file, and you have to tell me so that I can put it in my resume." More formal than that, ofcoz, but hey, this is teh wiki, and so I don't have to adhere to any decorum.

It occurs to me that something similar would work well for software patents. We'd have to have a year-long bounty to enumerate established algorithms, and we could probably formally eliminate everything in NIST's algorithm book. Still, I have a few little mechanisms I'd establish free, defensive patents on, if it didn't cost me twenty bucks, and then I could put a clause in my patent mentioning that I had to be in the manual and you had to "pay" me with an e-mail notifying me for each major version, to put in my resume.

Twenty bucks? You whiner.

No, shut up. See, I have about twenty of these things that I'd defensively patent, except that comes to (IIRC) about $440. OTOH, many algorithm designers like their things to be free. It occurs to me that if we had a system whereby someone could patent a software method for free, software patents would blossom, everything under the sun would get indexed in the year-long pre-emptive bounty, then genuinely new methods could be clearly documented, attributed, and patented.

What are you, mad?

Yeah, yeah. "Software patents are the devil!," raises the hue and cry, pitchforks and torches waving maldy in the stale air surrounding my moat. Whutevar. Patents have a bunch of very important aspects, such as establishing clear lineage of mechanism. Maybe more importantly, a pre-emptive patent can stave off abusive patents, and can prevent legitimate patents from falling into the hands of abusive creditors and clearing houses.

But this will be exploited!

Not nessecarily. If the intent of the mechanism is to facilitate people's free release of their own work from within the existing patent system, then it can easily be suggested that the amendment be made to the system that a special channel should exist for people whose patent establishes immediate and permanent release of the mechanism into the public domain. Certain restrictions should be allowed, such as requiring annotation in manuals, but nothing about money, licensing, or restricting child developments of this development.

In an awful and scary way, this is very similar to the MIT License.

The MIT license? You #include <ircFlame.h>!

Yeah, a bunch of people are going to tell me that this needs to be, or at least to allow, viral GPL behavior: you may not build upon my invention unless all of its children are free. We can't allow that here.

See, in the software world that's perfectly rational. You don't like it? Write your own! Nothing wrong with that: you can set terms on what you give away, because someone (given sufficient experience, anyway) can always just reimplement it if they don't like your terms.

The problem is, that's not how patents work. Patents cover a mechanism, not an implementation; it covers the concept rather than your particular instance of a concept. In terms of patents, some things just don't have an alternate method; this would effectively choke the huge early speculative investment in development on things related to that method which corporations are so often otherwise able to bring to bear. Say what you like about corporate research, but corporate research has been able to tackle problems on a regular basis the scale of what when done by government are usually considered massive public works. The space industry is a key example; everyone remembers the Apollo program, whose NASA research budget has since been absolutely dwarfed by corporate interests.

We cannot put economic chokeholds on branches of scientific development, even if the original author wants to. It's just not the same as with implementations. Imagine if LEDs had been viral-GPL style patented; it took tens of millions of dollars of research to reduce them to commercial viability, and if the corporation which did that had to give the results away - and LED manufacturing is very easy if you know how - then no corporation would have ever done the research in the first place, and we might just be seeing them due to amateur experiments as a new thing today.

It is laudable to release your work into the public domain, but irresponsible to hobble it so that it is unavailable to potentially its largest sponsor. Free patents would have to be one hundred percent free. No viral licensing. Period.

And who'll pay for all this?

Well, this is something of an issue, but I suspect it'd be quite a surmountable one. New algorithms are rare, and very many existing algorithms are well-known and well-documented. As long as there was a sufficient-length negation period (I'm guessing a year, based on all of my zero experience in the legal and patent fields) where do-gooders spend time clearly documenting and outlining a method in patent form, and as long as some sufficient basis (again, the earlier NIST algorithm list as an example) was used to prevent gaps of omission, then the general state of the art should be reasonably comprehensively represented. If this were to attain community support like the patent bounty movement did (albeit briefly,) then much of the heavy lifting - patent generation - could be done by the community, in order to prevent abusive patents like the incredibly overflogged horse of the Amazon One-Click Patent.

Problematically, there's some less heavy lifting which the community cannot do - namely, validating the patents - and this is going to generate an incredible one-time bulk to catch up the missing algorithm patents over years gone by. Also problematically, the patent bureau is already massively overworked and their budgets continue to be cut by the shady misvotemeister who's stolen office a few times and may a couple'a more. Thing is, WikiPedia was able to put together $50k in donations in under two weeks, with only one corporate sponsor in excess of $50 in the first six days of the drive (after which I got sick of reading the list,) and that from a company whose sole reason to exist is repackaging WikiPedia content.

Now, just looking at some of the numbers on the old patent bounty sites makes it deadly clear what kind of numbers underlie critical patents. Many corporations may invest money towards such a bureau as a good business move. Others may do it for the PR - especially the big guys like IBM, Sun and Novell who are currently playing big hands full of anti-patent-abuse PR cards on certain very fundamental patents. Moreover, critical patent donation to this system would legally enforce that the patents never be abused in their dying days by a company on the ropes, which would be a major PR move for these companies and a fairly large PR fault if they didn't actually commit; the system being there at all allows the public to judge with crystal clarity just how willing companies are to saw their short teeth.

What would it cost to run such a system? Probably a good two hundred thousand just to establish the new workplace, do specialty hiring, buy PCs and pay a year's rent. There probably needs to be at least a six month precursor period where how exactly to document and differentiate algorithms is clearly established, and where a bunch of tutorial documents and auto-writer forms are created to give the well-meaning professional with little legal experience to participate; that'll be another two hundred thousand dollars for three lawyers and two programmers over half a year. Given that the bulk of the work will be in the negation phase, which just means checking references and documentation, if the group has good access to a legal library, and given that a "this is already documented here" note requires little effort to check and file, they should be able to tear through a large pile of submissions relatively quickly with relatively little manpower. Things which need serious consideration, such as unclear points of differentiation, would be pushed to the end of the queue; as there became fewer more difficult questions to solve, much of the inexperienced staff could be released in favor of the legally acute.

As those difficult matters were solved, we would then need to wait for the moratorium to end, at which point there will be another, far smaller one-time bulk of material to go through, where people released algorithms they'd been hoarding or been too lazy or cheap to document. There shouldn't be a commercial gold rush on these like there would be under current patent situations, as the value of these patents is by definition zero from a purely monetary perspective. Still, commercial entities would likely provide a significant bulk of new entries due to PR maneuvers, finance defense and

But what's the point?

Well, there are two. One is the defensive patent status which I detailed above.