[throwawaykf]

somewhat active on Ask Patents. In fact, I've submitted an answer that is pretty sure to kill at least one Google patent application, and possibly another from Uniloc.

A few comments on this article:

1) This is a very unusual case; most answers (and almost all questions) from "lay engineers" completely misunderstand the scope of the patent, since they don't even know what claims are. And even if they do, they are very lax at interpreting claims. And even then, most posters frequently misunderstand the terms used (case in point, Spolsky's very post! https://news.ycombinator.com/item?id=6084884). All this leads them to post irrelevant prior art.

That does not mean there are no useful answers at all; there are, but they mostly come from people who are somewhat versed in patent law (such as agents, lawyers and examiners). Some re-wording of claims, such as what Micah Seigel does in his posts, helps, but for the proportion of useful answers to go up, we need more education about how patents work for this to be useful. It's really not that hard; heck I did it!

2) It's wayyy too soon for a victory lap because that was only the first non-final rejection, for which a response has already been filed. Statistically, this application will undergo 2.5 more rejections [1] and (based on my guestimate) at least one Request for Continued Examination (RCE) before being abandoned or (more likely considering the applicant) issued with much narrower claims.

3) Patents are worded so not (primarily) to be obfuscating, but rather because of legal, technical and some silly historical reasons. For instance, pronouns are very rarely used because any indefiniteness can be cause for invalidation. Obfuscation will not help much, because you are not trying to get it past lay engineers, but patent examiners, who have a technical background and are (usually) adept at reading patentese. Complaining about how hard it is to read patents is like a Blub programmer complaining about Lisp. You simply need to learn the language to appreciate what you are reading.

4) Most "software" patents (which can't even be cleanly categorized as such) are not crappy, at least with respect to all other patents. There are studies presenting this view [2, 3], but it's also based on my experience having read hundreds of patents. Almost none are revolutionary, but just as few are really as bad as the media portrays. The PTO has gotten pretty good at finding prior art (interestingly around the same time Google came around), and the really broad patents are dying out.

The "crappy software patents" view is common mostly because tech media routinely publishes uninformed (or disinformed? [4]) rhetoric, mostly because they garner some easy rageviews, and audiences accept it without critical thought. I do think the bar for non-obviousness should be different, but solving that is a difficult, almost-philosophical problem.

5) In response to various comments on this thread regarding pay-for-prior art schemes, initiatives such as Article One Partners already exist.

I am not a patent lawyer or an agent, but I believe in the patent system, as I have actually worked for the mythical small-guy firm that was ripped off by the big guys and almost died, but eventually prevailed with patents. You don't hear these stories much because typically the small guys don't have the PR budget for it [4]. (And also because many of those with patents turn to trolls, who like to keep a low profile.)

I have only recently become personally invested in the patent system, but I want all inventions, including mine, to be truly novel and worthwhile. And I want people to get off their butts and do something rather than complain about patents on HN. This is why I support Ask Patents.

[1] http://www.uspto.gov/dashboards/patents/main.dashxml [2] http://papers.ssrn.com/sol3/papers.cfm?abstract_id=650921 [3] http://papers.ssrn.com/sol3/papers.cfm?abstract_id=970083 [4] http://www.paulgraham.com/submarine.html

[dctoedt]

Re (2): In 1924, legendary federal judge Learned Hand [0] remarked that "the antlike persistency [sic] of [patent] solicitors has overcome, and I suppose will continue to overcome, the patience of examiners, and there is apparently always but one outcome."

[0] http://en.wikipedia.org/wiki/Learned_Hand

[1] Lyon v. Boh, 1 F.2d 48, 50 (S.D.N.Y.1924), copy available at http://scholar.google.com/scholar_case?case=9696597566965196....

[throwawaykf]

Agreed. It is an unfair match. It is, statistically, almost a tautology: The set of examiners is necessarily limited by government funding constraints, but the number of patent attorneys and agents they contend with on a daily basis is only limited by the market, which is huge. The examiners will eventually be overwhelmed.

I think this is one reason that most examiners (at least IME) have their default mindset to "Reject! Reject! Reject!" Also, this is why something like Ask Patents is invaluable to even the odds.

I don't agree with the quantitative approach, but I can't help think that technology can help. Google has already (in my opinion) helped the PTO greatly narrow claims the past decade; similar technology can help even more.

I have some background in NLP. And I know it's surprisingly effective when it comes to domains with specific jargon (cf. Watson and medical language). I've lurked long enough to know some here (such as VanL) have already experimented in this area. Personally, I have toyed with the idea of constructing parse trees out of multiple technical texts and claims, "normalizing" them using ontologies, and trying to find matches (i.e. prior art) using various tree-matching algorithms. I have a feeling it would be very effective. (Maybe Google already does this!)

But that does not address the problem of identifying patents that are quantitatively invalid but qualitatively valuable. To me, that is the more important long-term problem.

[kazagistar]

Increase the filing fee, to pay for better examiners directly? Patents that cost a few thousand dollars?

[pbhjpbhj]

This weighs against individuals too heavily. The UK for a time had a zero filing fee but they've gone back to a small fee again (to avoid getting so much chaff).

It's the renewal fees where you should be charging highly IMO; hyperbolic year-on-year increases would be an interesting option to model.

[xmodem]

I know this sort of thing is unpopular in the United States, but how about a pay-according-to-your-means model? So individual inventors working for themselves could file quite cheaply, but a publicly held corporation with a >$1b market cap pays a much bigger fee for the same thing.

[bstpierre]

Or perhaps even a pay-based-on-number-of-filings model. First application is $500, with a doubling for each filing, up to a max of $10k. (Or whatever multiplier and ceiling you want to have.) For small filers, the legal fees will dwarf the USPTO fees; for large filers with dedicated legal assets, the USPTO fees are still fairly minor but might start to be large enough to deter some of the frivolous filings.

[JohnsonB]

>4) Most "software" patents (which can't even be cleanly categorized as such) are not crappy, at least with respect to all other patents. There are studies presenting this view [2, 3], but it's also based on my experience having read hundreds of patents. Almost none are revolutionary, but just as few are really as bad as the media portrays. The PTO has gotten pretty good at finding prior art (interestingly around the same time Google came around), and the really broad patents are dying out.

That's only a person assessment, not a fact, as you are making out it is. The consensus in the tech industry among almost everyone (from software developers to tech legends) is that the patent system is broken and continues to routinely pass invalid and valueless patents.

>The "crappy software patents" view is common mostly because tech media routinely publishes uninformed (or disinformed? [4]) rhetoric, mostly because they garner some easy rageviews, and audiences accept it without critical thought.

No, it's because it's actually the truth and the software community is not nearly as susceptible to junk journalism as you make out.

>I am not a patent lawyer or an agent, but I believe in the patent system, as I have actually worked for the mythical small-guy firm that was ripped off by the big guys and almost died, but eventually prevailed with patents.

For every example of a situation as you described, there are 10 of a patent troll shutting down another small-guy firm or making them work for slave wages.

[VladRussian2]

>You simply need to learn the language to appreciate what you are reading.

when i read my ones (basically the same crap that other 99% software patents are) in patent legalese they sound more pompous, yet still an utter crap.

[throwawaykf]

The pompous aspect comes (IMO) from primarily historical reasons. However, in my experience, the judgement of them being "utter crap" is simultaneously (1) mostly uninformed and (2) yet mostly correct.

Let me clarify: people think they are crap for the wrong reasons, such as vastly overestimating the scope of the claims. But they are (IMO) still mostly crap because they are ridiculously narrow and not necessarily clever... Yet they are still valid because that's what the patent system is tuned to.

I have said this before in other comments -- the way patents are examined is mostly quantitative: find enough pieces of prior art where enough snippets of texts cover the text of the claims, and you have a rejection. No thought is given to the quality of the claims. For instance, I'd wager the RSA patent (which even the most die-hard anti-patent folks typically admit is novel) would not be granted today because each aspect of the claims is mentioned somewhere in the prior art. It is the combination of steps which is novel. Yet todays examination process would shoehorn the claims into a ridiculously narrow set of steps that (the attorneys hope) would still typically be infringed.

Examination these days is a mechanical process whereby examiners just try to find enough references with language in them to cover all steps in a claim to issue an easy rejection. (Yes, the "rubber stamp" is more likely used to reject rather than allow.)

So how do you try to qualitatively judge a patent when you have only a few hours appreciate all the prior art out there and appreciate what is being claimed and to do so?

The answer is, you can't; the only way to have a scalable process is to issue actions on a quantitative basis rather than a qualitative one.

As I said, it's a difficult, almost philosophical question.

[jacques_chester]

> The "crappy software patents" view is common mostly because tech media routinely publishes uninformed (or disinformed? [4]) rhetoric, mostly because they garner some easy rageviews, and audiences accept it without critical thought.

"Inventive patent filed covering obscure corner of programming" is not going to garner pageviews, for the same reason that "Grandmother walks home safe from 1000th time without being assaulted by ne'er-do-wells" doesn't appear on any newspaper outside The Onion.