EFF Politely Asks PTO to Stop Issuing So Many Crappy Software Patents
eff.org
eff.org
Keep in mind that for a patent to be granted the invention must be:
1) Non-obvious to someone skilled in the art. 2) Novel (ie. not done before).
Patent attorneys and patent offices seem to only use a test for novelty. They tell me that if an invention is obvious, then somebody would have done it before. They have no specific test for non-obviousness apart from demonstrating prior art. They seemingly have no ability to refuse a patent by just claiming that it is obvious. To them, lack of prior art is evidence of non-obviousness.
It is as if they live in another world, where the only inventions and ideas possible are the ones previously patented.
I can only assume that this has come about because a test for non-obviousness is subjective. It's easier for an examiner to let it through and lets the courts sort out the mess later, at our expense.
It seems to me that this is the root cause of ridiculous patents. In technology, we're moving too quickly for this. We need a better solution.
So it either already exists, or makes sense to be patented?
This truly boggles the mind.
Then again - does it really? If you are a patent lawyer, your job depends on there being lots of patents. The definition "anything that hasn't been patented should be" seems to be the logical extreme to serve that equation.
...and for someone to have done it before, someone must have patented it...
I think that's the real rub of the patent conundrum. The patent office assumes that if something hasn't been patented yet, it must be non-obvious. The whole "skilled in the art" things sort of goes out the window.
It's easy. Every time someone files a patent, the PTO gains the filing fees. If they don't check too much, people file a lot and they win a lot. If they check, it means first more cost for them, then less patent filed hence less money.
Now, what happens to them, when they accept bogus patents ? Answer : nothing. The PTO is simply not accountable for that.
No surprise they have no incentive to do their work properly.
The current patent situation is a failure of the federal government to understand and legislate contemporary technology, not a misalignment of incentives.
1. http://www.uspto.gov/web/offices/ac/ido/oeip/taf/us_stat.htm
A patent is a deal we, the people of the US, make with an inventor: add significantly to the sum total of our technical knowledge, and in exchange, we'll give you a time-limited monopoly on the technique you invented. The problem is that there isn't anyone unambiguously charged with making sure this is a good deal for us: that the knowledge we're getting is worth the price we're paying. It's technically the PTO's responsibility, and they do make some effort, but the incentives given patent examiners don't encourage them to be hard-nosed about it.
Currently, I don't think patents about to be granted are reviewed by the examiners' supervisors or anything like that. Seems to me there should be an internal committee that reviews every patent about to be granted by junior examiners, and an occasional one of those about to be granted by senior examiners, to verify its quality. Repeatedly approving applications that then get rejected by this committee should slow an examiner's promotion progress.
It's the entire system that is built around the assumption that 'obviousness' is too difficult to measure objectively, and that protection is vastly more important than anything else, so we should always err on the side of protection when laying the boundary of what should be patentable and what not. In other words, better to have 1000 bad patents than running the risk of having 1 idea that should be patentable being rejected by accident.
No thought is given to the enormous damage this does to our industry.
Separately, so far as I know, there's no reason to think that the USPTO is immune from typical bureaucratic incentive misalignment. More applications do mean more work and the simplest way to deal with more work is to get more people to do it. More people would increase the bureaucratic fiefdom of USPTO decisionmakers (who are first in line in terms of deciding how to handle more work) and the increased fees coming from the increased applications are at least a good starting point when arguing for a budget increase next year. This is a weaker incentive chain than "the USPTO isa business funded by filing fees", but it is still a significant one.
[1] http://www.patentlyo.com/patent/2011/08/usptos-future-budget...
The PTO revenue is vastly greater than that, and the vast majority is diverted to other agencies.
The actual PTO revenue is 2.2 billion in 2011
There is an examination fee of 720, a search fee of 600, and an issue fee of 1780.
There are also often plenty of other fees :)
The actual fees for an average patent are closer to 10k per filing, sometimes a lot more.
19th century administrative standards is very simple and got bright lines; they don’t require a great deal of expensive investigation. The assumption of the 19th century patent system is that everything that should be patented already is patented already.
Like other 19th century administrative structures, the patent office assumes that all it can be expected to do is to look in its own files and to make rough binary determination - 0 or 1, novel or obvious or not - on the basis of the consultation of its own administrative record.
This mean that PTO's job is to only answer questions based on its own records. A invention is novel if the clerk can't find a similar invention in the office. A invention is non-obvious, if the clerk do not consider the invention to be a natural extension of any existing files.
PTO stills live in that system while everything else has evolved forward during the 200 years that has passed, including what type inventions that are patentable. This mean that a large number of invention never got filed when they got discovered. If a patent application then get sent in several decades later, the result is simple. The patent office grants the patent because the invention is not in the archive.
This is crazy. We should not use 19th century administrative process in the 21st. We should not assume that government knows everything, or that their records can stretch over inventions when such inventions was unpatentable. Doing 20 years governmental intervention into the market should be a serious undertaking where the government are responsible if such intervention cause serious harm or is done faulty.
So for a rejected application, the USPTO gets $1,600. For an accepted application that results in a 20-year patent, the USPTO gets $16,280.
Plus, it's more work to reject a patent than it is to issue it. To reject a patent, the Office has to explain why the invention should not get a patent. To issue a patent, the Office merely has to accept the Applicant's claim that a patent should issue. Issuing a patent gives ten times more money for less work.
Fee schedule from http://www.uspto.gov/web/offices/ac/qs/ope/fee031913.htm:
Basic filing fee - Utility 280.00; Utility Search Fee 600.00; Utility Examination Fee 720.00; Utility issue fee 1,780.00; Publication fee for early, voluntary, or normal publication 300.00; Patent Maintenance Fees Due at 3.5 years 1,600.00, Due at 7.5 years 3,600.00, Due at 11.5 years 7,400.00.
See http://www.uspto.gov/about/stratplan/ar/2011/mda_06_01_03.ht...
A large portion of PTO revenue is actually diverted to other agencies/budgets.
Google "PTO Revenue diversion". This has been an issue for various reasons for years.
So issuing more patents does not actually help the PTO in that sense, because they don't get the money they earn anyway.
They still have no real incentive, and don't really see themselves as the gatekeepers they really should. So you are right, but for the wrong reasons.
At least one good start would be for them to stop calling people filing patents "customers", as if it was their job to serve these folks.
Also, the patent system should be modified to allow some way to handle customer discovery. Something like a single provisional filing fee allows you to file many iterative documents repetitively (with MVP source code perhaps) and those later provide a basis for priority date if proof is supplied they were customer-tested.
If you need to know that it's possible in order to make it, then it is non-obvious.
Maybe the algorithm is obvious, ie, to do X you need to do steps 1, 2, and 3, but X itself, the basic functional creation that employs the algorithm, is not obvious.
Swipe-to-unlock, the basic algorithm, how to implement it is pretty clear. But matching and researching which gesture works with which functionality, to the degrees customers find it natural took effort and time, and it produced a creation that was not obvious before that research.
Besides, if this is already part of the current rules, why do we need to amend the rules for it? Just enforce the rules better.
It's a great example of another big problem with software patents - the idea that adding "... on a computer" makes it somehow brand new.
"The purpose of the inventive step, or non-obviousness, requirement is to avoid granting patents for inventions which only follow from "normal product design and development", ..."
What you described for swipe-to-unlock is very much the normal product design and development process, so doesn't (well, _shouldn't_) qualify it for patent protection.
And yes, I agree: the rules DO need to be enforced better.
[1] http://en.wikipedia.org/wiki/Inventive_step_and_non-obviousn...
In fact, the law as described by Barton in the earlier section of his paper shows that the non-obviousness criterion was deliberately weakened over time, specifically when the Patent Act was passed in 1952 that eliminated a requirement for a "flash of creative genius."
Anyway, I am not a law expert. I am an entrepreneur. My interest is to be able to invent without fear that after I invented and researched and showed that some form of product was viable, some large company (Google, Microsoft, Apple, etc) or a competitor or any other group would then replicate it, claiming to have arrived at the product independently. The bottom line is that it wouldn't foster innovation, it would hinder it. I would have no incentive under the patent system to go to investors and secure funds to do research if in the small chance that the research is fruitful, I will not be able to protect it.
[1] Draft version of Barton's article: http://emlab.berkeley.edu/users/bhhall/ipconf/Barton901.pdf
That's actually a great idea--it's basically a petition with a donation required to sign (thus making it into something that actually has autonomous power, rather than some names on papper.)
Documented code would be a great step forward if we decide we have to keep patents around because at least your money is buying source code in some form.
I think common sense applied to the process would be enough, rather than mandating languages that may or may not be relevant in the future. Though Patent System and Common Sense might be bit of an oxymoron.
Or C# (ECMA), Common Lisp (ANSI), SQL (ISO), Fuzzy Control Language (IEC) ...
I read a patent application the other day that basic said that someone with a knowledge of the field would be able to reproduce the invention using the patent as a guide. The application was 50+ pages long and I couldn't get past about 10 of then without dying a little inside.
All patentable subject matters work this way. In some engineering areas I've worked in, you license the patents from one company and license an implementation design from another. The notion that computer algorithm patents are somehow special in this regard betrays a lack of familiarity with other engineering disciplines.
(I agree that patents are nigh unreadable but then I come across many cases of people correctly reconstructing e.g. computer algorithms with nothing but a patent to work from so it appears to be eminently possible.)
The code should be included to uphold the spirit of the patent. A patent is a temporary monopoly granted in return for telling everyone how your invention works. If all you do is enumerate what your invention does in your claims, you haven't described how it works, and if the patent is granted, you've effectively only described any implementation of your idea, which is a huge problem.
Providing source code narrows the patent to AN implementation of an idea, and to experts in the field, actually explains in a common language (not legalese) exactly how it works.
Have a read of the brief the EFF put together for the PTO. It's somewhat readable, even for an engineer.
The problem is that algorithms and math are not considered patentable. The guidelines for membership in the patent bar specifically exclude mathematics coursework as a valid educational background. The reasoning, I believe, is that abstract laws should not be patented.
However, with the rise of software patents, we clearly are allowing algorithms and mathematics to be patented. However, the patent bar actually specifically excludes the background that would be most useful in evaluating these patents.
It makes sense if you think of the patent system as a cartel. By allowing math to be patented, you expand your domain. By denying people with a mathematics background to sit for the patent bar, you restrict your competition.
Sometimes, when you're confused by a bad policy, just ask yourself: what would a person with no morals, plenty of power, and a single minded desire to enrich himself do? Usually, that will explain the policy.
The PTO simply has no ability to change patent law. We apply it as it's written by Congress and interpreted by the courts. A few points on the law without getting into criticism of the agency (you can get easily find that elsewhere):
1. The burden on patent applicants is too low. There's a presumption that a patent will issue as soon as it's filed. The patent examiner has to prove that whatever is claimed was obvious, didn't work, wouldn't work, etc. If the patent examiner can't meet that burden, the patent issues. The patent applicant doesn't have to tell us about the state of the art prior to the "invention", what the improvement is, or why it deserves a patent.
2. As discussed upthread, the obviousness standards are too low. If you can't find evidence that each and every aspect of a claim was in the prior art, you have to allow the patent. Even then, a patent attorney can attack the rejection if the different prior art documents wouldn't fit well together. As an examiner, you can't simply say that something would have been obvious, with no supporting evidence. If you try to make such a rejection, the patent attorney will appeal your rejection, you'll get reversed on appeal and the patent will issue anyway.
3. The process naturally skews towards patent applicants. Most (nearly all) patent applicants have an attorney who is fighting 100% for as broad a patent as possible. The examiner is supposed to be a neutral arbiter who applies the law in an evenhanded fashion. There's no party fighting for the public, pushing back hard against the patent attorney. It's like a courtroom where only one side of the story is presented.
You can find a lot commentary/criticsm of the patent examination process online BTW, but not generally on the tech sites - look more towards legal/academic literature.
(Speaking on my own behalf, not the agency's)
How many patent agents are there employed by PTO? Are they all full time?
I wonder to what degree examiners use this before granting a patent.
I agree that the law needs to be changed. Instead of the burden of proof being on the examiner to prove obviousness -- which is basically impossible, as it's a subjective judgment -- I believe we should place the burden of proof on the applicant to show objective evidence of nonobviousness. 35 U.S.C. § 103 (http://www.law.cornell.edu/uscode/text/35/103) needs to be clarified to establish that the burden of proof falls on the applicant (during the application process) or plaintiff (in an infringement proceeding).
We already have a pretty good list of some things that could constitute objective evidence of nonobviousness; see http://en.wikipedia.org/wiki/Graham_v._John_Deere_Co. [the trailing period is part of the URL]. To these we could add publication in a peer-reviewed journal or conference proceeding.
I'm not necessarily suggesting that a patent couldn't be granted absent such evidence -- an applicant might reasonably argue that there has not yet been time for her invention to have enjoyed the commercial success that would constitute one kind of evidence. In any case, a granted patent should be annotated with the evidence provided; if the patent is later litigated, the court should review that evidence and allow additional evidence to be presented. If the evidence at the time of litigation were still insufficient, however, an infringement suit could not proceed.
I think this requirement would not much affect most fields other than software. For a drug patent, for example, the applicant could simply list all the other molecular structures they tried that didn't have the desired properties. Most non-software patents have substantial R&D behind them; it would be sufficient to provide evidence of the work that was done.
It would make a huge difference in software, though. Mere novelty would no longer be sufficient. There would have to be some objective reason to believe that another person, faced with the same problem, would have difficulty coming up with the same solution.
I think this proposal addresses all three of your points. It would obviously increase the burden on applicants and patent holders. It would reverse the burden of proof on obviousness. And it would give the examiner -- and then the defendant, if the patent is litigated -- a powerful way to push back against the applicant/plaintiff.
A pretty good argument has been made that we should just legislate that developing, distributing, or running a program on generally used computing hardware does not constitute patent infringement.[1] Why not just do that?
[1] http://www.wired.com/opinion/2012/11/richard-stallman-softwa...