JPEG Committee On The Ball, Seeks Prior Art

Sangui5 writes: "It seems as if the JPEG Committee has noticed the recent patent fuss, and is working on the prior art angle. Good to know that even though there's a new standard, the committee is standing by their previous work."

Isn't JPEG just a FFT?

[FelixCat]

I thought that JPEG was just a Fast Fourier Transform of the encoded data. How can patent this with so much history in the Mathematics literature?

Crazy

Sorta like Rambus

[PingXao]

Didn't Rambus get slapped for this sort of trick? If I remember correctly, they held certain IP which they did not disclose during the standards meetings. Then they waited until lots of other companies were using those standards, which incorporated their IP. When the momentum was already strong, they attempted to collect absurd royalties, and let loose their legal dogs to pounce on anyone who didn't ante up. Sounds verrrrrryyyyyy familiar.

Re:This is necessary

[Anonymous Coward]

I'm not sure I'd even praise the JPEG group for taking swift action - I'd say they're doing what's necessary to combat Forgent's crime. Doing their job as a standards body like an officer does his job as a member of the police.

The thing to understand here is that the way things are lately, we consider it an unusual and joyous occation when a computer industry body does its job properly or acts in the public interest..

OpenGL/Microsloth patent dispute is the same deal

[Whammy666]

This is the second major IP bombshell to hit the computing community in the last few weeks. First, the implied threat of M$ playing the patent card with their acquisition of an openGL patent, and now this. This seems to be new tactic among greedy corporations which involves seeking out widely used patents whose owners have allowed free use by non-commercial entities, purchasing the patent, and then announcing new restrictions in a effort to cash in on it's popularity.

It seems to me that a patent that has been released into the public domain (at least for non-commercial use) should remain so if and when the patent is sold. I don't believe that there is any law requiring this, but anyone selling an 'open patent' should include a requirement that it remain open as terms of the sale to avoid this very situation.

It would be interesting to see what would happen if someone should decide to challenge a patent that was open only to be closed at a later date. Think about the series of events: Group A invents an image compression algorithm and grants me license to use the patent free of charge. I develop a group of products based on this agreement. Everything is cool until Group B buys the patent and says I can't use the patent anymore (or worse, demands back royalties). But wait, my products were based on a agreement I had with Group A, not Group B. Group B came in after the original agreement and is trying to change the terms of my agreement with Group A after I've executed the agreement. I would argue that Group B would be compelled to honor any agreements that Group A had in force at the time of the purchase as part of the package of buying the patent.

===
All your patents are belong to us.

Re:Jpeg

[mikecarrmikecarr]

So how will this impact the whole GIF vs. PNG thing?

GIF's continue to be bad; PNG's continue to be good

Ignoring the philosophical reasons, PNG's are better:

1. (In my experience) PNG's are smaller
2. They support a variety of compression standards (see pngcrush [sourceforge.net])
3. They support a larger number of colors. GIF's used a 8 bit palette; PNG's can do truecolor, greyscale or 8 bit palette
4. They support animation (through the related MNG standard)
5. They support transparency through alpha channels. Alpha channels are a very good thing that I could rant and rave about (but I won't ;)
6. They support gamma correction
7. They support more intelligent interlacing than GIF's

See http://www.libpng.org/pub/png/pngintro.html [libpng.org] for more information, if you're so inclined.

Re:Something I don't get.

[Erasmus Darwin]

"Now you could say "Trademarks and Patents are two different things" but they are really aren't."

I think the reason why patent holders don't have to immediately prosecute is that patents are considered less readily visible than trademarks. For example, if another company opens a burger chain named "McDonalds", a lot of people are going to notice. If, on the other hand, a company infringes on McDonalds' (made up) patent for cooking a hamburger for 98.742 seconds, it might take awhile for it to become known.

In short, it takes 30 seconds to find out what someone is publically calling themself, but it can take considerably longer to reverse engineer one of their products. IP protection/enforcement laws seem to reflect this disparity.

That being said, I do think that something has to be done about people pulling the submarine patent non-sense. But I still think you'd be doing a disservice to treat this IP identically, as there are differences. Heck, even the length that the IP exists is different in both cases (as patents need to have a fixed life while trademarks should continue as long as the manufacturer makes the product; there's no compelling reason to suddenly declare that anyone can make a car called a Ford simply because it's been XX years since Ford began using the trademark).

Re:Computers are not magical beings

[ukryule]

It should be *harder*, not easier, to prove lack of prior art.

You've got it the wrong way round. The Patent Office does not prove lack of prior art, it proves prior art. What I would say is "It should be *easier*, not harder, to prove prior art."

The failure of a database query hardly constitutes "lack of prior art"

That is exactly what happens in practise. A patent agent has a few hours to understand the patent in front of him, search for prior art, compare all the prior art he finds to the proposal, and then justify rejecting/accepting it.
There is no way they are going to go trawling through archived usenet postings/search the web for detailled date-stamped documentation for every case. They *need* an easy way to search for something to compare the proposal against.

Your suggested database would result in *more*, not *fewer* bogus patents being granted, because it would accelerate the application process without adding any protection above and beyond what's already there.

I'm not suggesting that the time/patent is decreased - simply that the patent agent has more powerful tools to prove the (in)validity of a patent in that time. That has got to be a good thing.

I work for JPEG2K, posted this story Friday..

[warpedrive]

.. And was rejected. The exact story, and I posted it because we were in committee discussing it. Someone else posts it, and it's news? I really thought that whomever arbitrates this, was interested in content, not in particular authors. changes my idea of how this forum works. Bit more elitist than I thought.

Submarines

[shmlco]

Which may not help if someone comes up with another submarine patent. It's difficult to insure everything is royalty-free when you don't know the patent is there, or that one you don't think applies in fact does.

As the Forgent mess shows...

Re:Prior Art?

[Wanker]

The first mention of JPEG in the Google archive [google.com] dates from Jan 1990, but references papers presented in 1988 and 1989.

Re:How will a database fix things?

[markmoss]

The fundamental problem is that patent examiners look primarily at the database of existing patents. There is no way that someone who is not actually working in an industry could be expected to keep up with all the relevant publications and products, so they have to depend on what has been indexed into a database - and for most technology, the patent database is fairly inclusive. This doesn't work for software; many basic algorithms were developed before software could be patented, also many programmers are rather anti-IP and would rather place their new algorithms in the public domain by simply making them public. (It is of course quite possible to get a patent and make it public domain - the USPTO even has a special patent form for this - but few people want to go throught the paperwork.)

This works legally, but not practically; because the patent office is unaware of what is not in their database, they are quite likely to grant a patent on ideas which an expert _working_ in the field would recognize as not new. Most notoriously, the Australian patent office granted a patent on the wheel; yep, there have been no prior patents on the wheel, even though there's 5,000 years of prior art. I am not sure if that patent examiner was remarkably stupid or went along with the joke, but fields where the existing technology is less well known (image compression software, for instance), it will certainly always be possible to slide public domain ideas by the examiners as long as they do not have a database of public ideas that is as well indexed as their patent database.

Of course, if there was prior art, you can always go to court and invalidate a patent. The problem is that once the PO signed off on it, the courts consider the patent valid until proved otherwise. If you have unquestionable evidence of prior art (e.g., the patent description is copied right out of Knuth), it's still very expensive and takes years to get to present it in court. If the equivalence between the prior art and the patent claims is murky - and it usually is, because people filing questionable patents never use the normal industry terms to describe their "invention" - it's going to be a long, expensive court case, with the outcome depending on whether the judge and jury manage to comprehend the issues. Or it might be quite difficult to prove that the shareware source code you are presenting as prior art actually dates from 1980. And after you go through all this and win, in the US usually you can't get your legal costs back from the company asserting the bogus patent.

You can recover your costs and more if you can prove it was truly fraudulently filed - but that's one reason the filers use odd jargon, so at worst they can claim they invented the algorithm independently and never saw the writeup of it in "Proceedings of the ACM".

The "odd jargon" issue will limit the usefulness of database searches, but still there is a much better chance of finding non-patented prior art if the examiner has a database of non-patented art to search than if he only searches the patents... If a patent has to be taken to court, a database of public domain source code and algorithms would make it easier to find the prior art, provide proof of the original date, and make it more difficult to file and assert bogus patents without being found liable for fraud.

Given the patent office's recent record of errors exceeding even the norm for government agencies, I would recommend a different approach. Reduce the role of the patent office from approving patents to merely recording patent forms in a public database; putting the forms in the database does not imply that it's a good patent. This database will include both patents and public-domain ideas. Patents require a filing fee sufficient to cover the PO's expenses, but there is no fee to post an idea to be free to the public (if it doesn't infringe on prior patents).

Along with the forms and filing fee, the inventor has to send a $20,000 bond to pay off anyone who successfully challenges the patent within the first three years. As soon as the forms are posted to the database, the inventor or agents can start asking anyone else using the idea to stop or pay royalties. But anyone can also challenge the patent, whether or not they are in infringement.

And we need a special, technologically sophisticated, court to rule on patent issues. That is, you need judges with degrees in engineering or science as well as in law. The initial challenge requires a brief summary hearing before a judge, with an informal presentation of evidence. (Brief and informal so that $20K bond will be sufficient.) If the patent is less than 3 years old and has not been previously upheld by a court, there is a presumption _against_ the patent - that is, the inventor must present a preponderance of evidence to uphold the patent. If the inventor withdraws the patent or the judge rules against the patent at this point, the challenger gets reasonable and necessary expenses plus a $5K profit, or $10K if the prior art was in the database before the patent was filed. The inventor does have a motive to withdraw if he's likely to lose, because the longer the proceedings go on, the more he'll pay. If the challenger loses, he does not have to pay the inventor's expenses; defending the patent once at a summary judgement is just a normal expense of getting the patent.

All prior art presented to the court goes into the patent and public ideas database. If the patent is invalidated, it stays in the database - marked as invalid, with the court ruling given, and so anything in it that wasn't in a prior patent becomes public-domain.

The loser in the summary judgement can request another hearing before a 3-judge panel, or request a full jury trial - but in either case he has to pay the court costs and the other side's expenses until the final judgement. In jury trials, the jury pool will be working scientists and engineers, and be paid appropriately, so this gets rather expensive... The court will have to power to assess costs and the winner's legal expenses against the loser, and to fine either party if egregious behavior such as knowingly filing falsely is revealed during the trial.

Re:I work for JPEG2K, posted this story Friday..

[Anonymous Coward]

I tried to post it when it was breaking news -- IN EARLY JUNE. Slashdot is as political and commercial as any other website. They will not publish something they do not agree with, and quite frankly, I expect this post to disappear.