Hacker News new | past | comments | ask | show | jobs | submit login
Google is trying to patent video compression use of Asymmetric Numeral Systems (encode.ru)
116 points by dvfjsdhgfv on July 12, 2017 | hide | past | web | favorite | 53 comments

Is it too late to consider killing software patents? I guess the battle is over and we just have to bite the bullet and be happy with these half-assed "countermeasures".

I'm probably biased because I mostly read about software patent stories on HN and other similar outlets, but it seems like 99.9% of the time they're only used for patent trolling or killing the competition.

Patents are supposed to protect the inventors by giving them the time to license and market them before the competition can swoop in and create a similar product without having to pay for the R&D.

So do people around here have "success stories" regarding the use of software patents to actually reward creativity instead of just playing the "patent cold war" game?

I think the CMU hard drive patents are a good example: https://www.google.com/patents/US6201839. But ideally, a patent is never litigated. It simply allows an R&D house to exist as an independent entity without forcing it into becoming a product/manufacturing company. I don't know about SV, but outside SV, investors do care if your tiny little company has defendable IP.

I worked at a startup doing cognitive radios. We were an R&D shop only--our competitors were companies like Raytheon and Lockheed-Martin that had manufacturing muscle. We patented everything, and most of it was what you'd call a "software patent" (even though in the long run it would be partially implemented in an FPGA or running on a microcontroller). It was expensive to pay PhDs to spend a year doing Matlab simulations to figure out what algorithms would work.[1] Even more expensive was then spending years testing those in the field (because lots of things that work in simulation fall flat in uncontrolled real-world environments). Copyright was pointless. Writing the code was the easy part.

I'm a bit on the fence--I think the problems with patents are overstated, but the benefits often are too. But carving out "software patents" from other kinds of patents strikes me as odd. There is no clean line--it's often easy to take something that might be done by an ASIC and move it into software and vice versa.

[1] E.g. one project involved getting disjoint listen-before-talk networks to synchronize their "listening" periods without being able to communicate with each other directly.

Patents are an offensive weapon by design. Getting a patent does not allow you to make a gadget; it allows you to sue someone else so that they cannot make a gadget.

It shouldn't be surprising or unexpected or frowned upon that they are used to kill competition - that's basically their raison d'être.

Summary from Reddit (https://www.reddit.com/r/programming/comments/6h08z5/google_...):

Asymmetric Numeral Systems ([1]) is entropy coding family currently replacing Huffman and arithmetic coding in data compressors, among others, of Apple, Facebook and Google, thanks to being up to 30x faster [2]. Its author has made it public to prevent pathology of arithmetic coding, which wide use was blocked by patents for many decades ([3], [4]). However, currently others are trying to patent basic applications of ANS – including Google for AV1 video compressor (initially suggested by ANS author, who has helped them for the last 3 years: [5]) in very general patent application, to prevent others from using it in image and video compression – claims and sources: [6] [1] https://en.wikipedia.org/wiki/Asymmetric_Numeral_Systems [2] benchmarks: https://sites.google.com/site/powturbo/entropy-coder [3] https://en.wikipedia.org/wiki/Arithmetic_coding#US_patents [4] Charles Bloom comment: http://cbloomrants.blogspot.com/2015/05/05-21-15-software-pa... [5] https://groups.google.com/a/webmproject.org/forum/#!topic/co... [6] Google claims and sources: https://encode.ru/threads/2648-Published-rANS-patent-by-Stor... Clarification: the "30x speedup" is for decoding of the fastest arithmetic coding (~50 MB/s) vs of ANS (~1500 MB/s) in benchmark [2] for similar compression ratio (better than Huffman, which standard implementation: zlibh from gzip has ~300MB/s). The inventor of ANS is Jarek Duda.

The title sounds like Google is doing everyone a favor. From reading the thread on encode.ru it appears to me that Google is patenting it so that they have the patent, not for purely altruistic reasons.

If it was for the public good, the patent would be filed via a non-profit that Google can't control.

I don't think the title is positive.

I assumed it to mean that Jarek Duda had joined Google, but it is Jarek speaking in the thread.

I follow the data compression sphere a bit. It really seems like Google is making a mistake here. From a patent perspective, since ANS is already "prior art", it seems like it doesn't deserve a patent.

Presumably Google wants to avoid receiving a misinformed lawsuit about ANS that they would then have to fight to prove that an ANS patent shouldn't be valid. Still, not putting Jarek in the loop (and like he mentions, not engaging with his university more) is wrong.

The best takeaway:

if anybody is also afraid of Google having a very general patent for using ANS (both tANS and rANS) in image and video compression and can explain why the above claims are just a natural mix of known methods:

go to: https://efs.uspto.gov/efile/portal/efs-unregistered

select "Existing application/patent" and then "Third-Party Preissuance Submission under 37 CFR 1.290"

as "Application number" enter: 15/370840

"Confirmation number": 5111

"I don't think the title is positive."

The title of the post was changed. It used to read that they were patenting for the public good.

"it appears to me that Google is patenting it so that they have the patent, not for purely altruistic reasons"

This is not correct. In fact, as far as i know, the code was already open sourced, and the patent would end up released as part of AV1/etc anyway.

"If it was for the public good, the patent would be filed via a non-profit that Google can't control."

It'll be released that way, so who cares who files and pays the fees?

You really don't want non-profits filing patents. Really. This is pretty much the most horrible thing i can think of. It's infinitely easier to keep them from going bad (Example: MPEG-LA) if you can make their corporate bylaws, etc, prevent them from filing patents.

>You really don't want non-profits filing patents. Really.

>This is pretty much the most horrible thing i can think of.

>It's infinitely easier to keep them from going bad (Example:

>MPEG-LA) if you can make their corporate bylaws, etc,

>prevent them from filing patents.

? Non-profits are corporations. They have by-laws. They also have several advantages over for-profits in the case of dissolution; eg, you can set up a non-profit so that the IP they hold is not simply sold to the highest bidder.

FTR, Xiph.Org is a registered 501(c)3 non-profit, and we hold patents.

Holding is different than filing. As I said, i'm fine with holding, just not filing. For precisely the reason you say.

Non-profits can in fact, "go bad" the same as any other corporation.

If they can only hold patents, and not file them, you at least cut off a source of that happening.

Is it possible to work around? So is anything in the legal world.

But it makes for a nice barrier.

> If it was for the public good, the patent would be filed via a non-profit that Google can't control.

Couldn't they just publish the techniques by some other means? E.g. some scientific journal. I mean, a publication is strong evidence of prior art.

I've been wondering this, and I'm not a lawyer so I have no experience in this regard. But!

If you patent something, and don't have a licensing fee or anything like that, I would imagine that it would deter some patent trolls from getting a false patent and suing people using that technique. Fighting in court costs a lot of money for people like me (and probably you) so anything that keeps me out of court is a good thing.

Hmm, so actually we need a different construct. Perhaps USPTO could add a checkbox to the patent form, saying that the patent will not be used aggressively in a court of law. Still not ideal, since I suppose the patent application procedure will cost the same.

That would mean that the USPTO tacitly admits the fact that the patent system is broken and they need hacks to work around it. I think that would be strange coming from them.

If the patent system was working correctly there would be no need such a thing, either something is patented or it isn't. If it isn't it shouldn't be possible for a third party to patent it post-facto since there's already prior art. At least, that's the spirit of it as far as I'm aware.

In theory yes, but it's less effective than filing a patent app, letting it publish, and then abandoning. Patent examiners tend to spend more time searching patent prior art because the prior art disclosures tend to be kitchen-sink exercises that make it easier to hit all elements of a claim relative to more targeted technical publications.

Are summary judgements typically issued on prior art alone? These things are done as a means of avoiding litigation. If you have to take the claim to trail, even when you win, you've already lost.

Or at least, that Google would pledge all of their patents if their patents had altruistic goals.

There is a tiny number of patents they've pledged not to sue offensively for: https://www.google.com/patents/opnpledge/patents/

The fact that they have not pledged most of the rest indicates that someone there is maintaining their ability to offensively sue in the future.

Have they ever used their parents offensively ever? The only case I know of is the Waymo one, but that was a special case where they had proof of their data being stolen. The other case would be against Microsoft but IIRC Microsoft was the one who started that case so it was technically "defensive" use.

Companies change. You can't assume that a company not offensively using patents today will not offensively use them tomorrow. If based on some misguided belief of corporate altruism (something that very simply does not exist), you grant a company unwarranted rights, they will come back to abuse them later. Even if you have unwavering trust in Larry Page and Sergey Brin's unquestionable morals (I couldn't say this with a straight face, but anyways), the reality is, they will not run the company forever.

If you've noticed something about most patent trolls, they're has-beens. When you can't compete, you sue on patents. Right now, as one of the most valuable companies on the planet with multiple products that are monopolies, obviously that does not describe Google today.

But it does describe Google at some date in the future. And when Google's profits are crashing, and they're trying to figure out how to stay in business, some business exec is going to realize that treasure trove of patents can solve all their problems...

But we don't live in a perfect world. Someone has to hold this patent. If it isn't Google, it'll be someone else. If you can give me any other company which you think is better situated to hold this copyright, go for it. The system is broken, and this is the best we can do right now.

Why not just not patent it if you want it available for everyone? The current version of Google would not enforce the patent. In the future if Google's leadership, organization or position in the market changes someone would be able to obtain the patent and then no one would have the favor.

Because someone else less scrupulous might. At our current juncture it's impossible to trust that the patent review system will effectively capture prior art events.

Capturing prior art products is hard. But Google could just publish articles describing the technology in a journal. That is solid, easily obtainable prior art.

You can have all the prior art you want and still end up in court, having to make your case, because you're not the one who got the patent. For what it's worth, and slightly related, there are design mocks on the chromium site that didn't get implemented and always looked like deliberate prior art to me, so I suspect the idea is not new to the lawyers in Mountain View.

We don't know what's the rationale here. For all we know, Google might even be OK if the patent is not granted to them, as long as it isn't to anyone else, either. The radio silence is to be expected when lawyercats are involved, unfortunately.

[Ex-Google employee, not a lawyer, etc.]

Getting a patent doesn't mean the patent office doesn't give someone else a patent that more or less covers the same thing.

Hah. Sadly true, but you'll probably agree that it's less likely.

Maybe, all the examiner records I've seen are definitely biased towards patent prior art rather than published works. But I'm not sure it's worth the expense of patenting all that work.

But every single major IEEE, ETSI, IETF, etc. standard ends up with 10 different companies getting patents on exactly the same shit.

Just wondering: Can a patent holder give up his rights and invalidate the patent? Or can a copyright holder put his work into the public domain?

> Can a patent holder give up his rights

What do you mean by "give up his rights"? Whether he can allow other people to use the patented idea?

> and invalidate the patent?

What do you mean by "invalidate"? Whether they can show the patent office prior art to their own patent and ask them to for that reason invalidate the patent?

> Or can a copyright holder put his work into the public domain?

That depends on the jurisdiction.

"What do you mean by "give up his rights"? Whether he can allow other people to use the patented idea?"

My understanding is that someone needs to hold a patent, be it a non-profit or commercial entity. What I wonder is if someone who could file a patent or already holds a patent for an idea could put this idea into the public domain without anybody holding a patent? Does this make sense?

> My understanding is that someone needs to hold a patent

Well, yeah? I mean, the whole definition of a patent is that it gives someone a monopoly over an idea. A patent without someone holding it would be as sensible a concept as a bank account without a holder. A bank account documents that money is owed to someone. If there is noone who it's owed to, then there isn't a bank account. And if there is noone who holds the monopoly ... then there is no monopoly?

> What I wonder is if someone who could file a patent or already holds a patent for an idea could put this idea into the public domain without anybody holding a patent?

Depends on what you mean by "public domain". If you own the patent, you get to decide who gets to use the patented idea. Obviously, nothing prevents you from deciding that everyone is allowed to use the idea without any further conditions. That you get to decide that is the whole point of the system.

I will not repost the entire thing here, as there's a few naughty words born of frustration, and this might well be a family forum for all I know.

My take: https://xiphmont.dreamwidth.org/84214.html

Thanks, very interesting take on the matter. Would you mind explaining "...it has to code everything in reverse. The latter especially is a technical showstopper. " ?

A previous post about this on HN:


Link provides steps for people to submit prior art to USPTO for this patent application (patent is not granted yet)

The patent does not cover rANS. rANS has been independently discovered a few times at this point. The patent filing covers Google's use of it.

To me what would make the most sense would to have a consortium of companies do this. Imagine Google, Facebook, Amazon, etc. all chipping in in a coordinated effort to prevent trolls from lapping up bullshit patents.

So much for publishing your work in hopes of creating prior art and preventing patent takeovers.

I blogged about this elsewhere. From reading the filing, Google is not patenting his work. They're patenting their specific application of it, and the claims seem to me fairly tightly tailored to AV1.

so 'email .. on a mobile phone' patent? not to mention He was the one helping them with this particular application, and even mentioned it almost 10 years ago in his original paper.

Can someone please summarize the patent?

Intellectual property is one of the worst things ever created, used extensively by that certain group of humans to hold back progress for the entire species. People in the future will view these laws as barbaric and evil. We would probably have advanced space ships, and certainly world peace and material abundance by now if not for people treating abstract ideas as property.

I broadly agree that intellecutal property isn't really property at all, and that copyright laws are unjust indeed, though I think it is hyperbolic to say that we'd have world peace and abundance just by their abandonment. More broadly the concept of property as a whole ought to come into question if we wish to go that far. Proudhon's essay What is Property? and Kropotkin's The Conquest of Bread shed light on this debate, making the case not for the sharing of private property by all but rather the very abolition of the concept itself.

>The inventor of ANS is Jarek Duda.

Google has thrown few grants for ANS research to Jarek and folks of encode.ru community, then patented their research behind their backs

Update: they did not, but they did approach Jarek's university with specific interest in that research work, and vaguely expressed interest in funding them.

Jarek's stance - "this is my work, not yours, this ends this discussion."

Google - "we have grounds to do so, and nothing you do will move us"

Update: I put things in a more neutral tone.

Can you provide some context or links?

This seems to contradict directly with what user SquareWheel writes in another comment and who now seems to have deleted that comment.

Who is Jarek - his page http://th.if.uj.edu.pl/~dudaj/

His original publication from - https://arxiv.org/abs/1311.2540 . That work was a result of two years of research that was well documented, and was making noise in encoding community. His first notes on use ans for quick entropy coding date to 2008.

First production grade code was published under GPL at around late 2013 with explicit intention to enforce copyleftness of derived works, and the patent poison pill clause.

Jarek claims that Google has abruptly ended communications with him few month after the time of first code publication

Update: it seems Google never made any grants to the group, but were only vaguely feeding them with "may bees" for two years, while requiring more and more detailed "research scope declarations." It is better now to let Jarek give his own first hand account of what was going on in between his university and Google's research people

It sounds like all these copyleft-warriors (like Stallman) could litigate. Isn't it what they are for?

No, that's not what they are there for. Why are you so snide?

I'm not. I quite literally mean it. It's their purpose, it's why all these organizations (I cannot remember the names now) are created, isn't it? They don't design anything, they don't write code, they do talking and litigating, which might, for once, be actually useful.

> They don't design anything, they don't write code

I don't like the tone you've taken. I'll assume you're merely uninformed.

Stallman's organization, the FSF, was created to protect the open-source nature of things that he was a primary creator of or significant contributor to such as gcc and emacs.

The Python foundation was created by the creator of Python, and so on.

To claim that there is no creative endeavor in these entities is to ignore the fundamental reason they exist.

see https://www.reddit.com/r/programming/comments/6h08z5/google_...

Google promising free license to anyone who implements AV1 would have been ok, if they patented something that is only useful for AV1. But they patented "using ANS for video" so now you can't use ANS in any competing video codec. In my opinion, people absolutely should fight to invalidate this patent using any work that mentions using ANS for video from a year (or more) before the filing.

I think you can, software patents don't work in EU, you'll have just trouble doing it in the USA.

This is incorrect, you are conflating algorithm patents and so-called "business method" patents (e.g. "$BUSINESS_ACTIVITY using the computer/internet").

Proper algorithm patents are supported in Europe because they are equivalent to a novel electronic circuit, which is generally viewed as patentable subject matter.

Applications are open for YC Summer 2019

Guidelines | FAQ | Support | API | Security | Lists | Bookmarklet | Legal | Apply to YC | Contact