Ex-NVIDIA Engineer Patent Issue With Open-Source

[phoronix]

Phoronix: Ex-NVIDIA Engineer Patent Issue With Open-Source

An ex-NVIDIA engineer that had a patent concerning high compression rate texture mapping attempted to attack an open-source project for supposedly violating this patent related to software graphics texture compression. The open-source software in question is Crunch and it's written by a Valve Software developer...

http://www.phoronix.com/vr.php?view=MTE0Njg

[plonoma]

Two questions come to my mind:

- if it's patented, how did it become part of the OpenGL standard?
(How is it possible that the whole industry uses those things without thinking and getting themselves in problems?)

- when will the patents expire?

I'm not going to google because I want this kind of information added in future articles. That's why I read news sites: to find information well presented with good information in one place.

[DeepDayze]

To me it is a good sign that that ex-nVidia developer has extended an olive branch to the OSS community after this brouhaha has come to light. What has happened to that Valve developer's work? Did he ever have a chance to defend his work and show it hasn't violated that patent or ever worked around the patent in question?

What about if someone has improved upon the original patented idea and is that deemed a derivative work?

[log0]

Don't forget floating point textures. How someone could grant a patent for this is beyond my comprehension.

[F i L]

Patents are so retarded.

[schmidtbag]

Patents are so retarded.

Today, yes, but otherwise they're one of the best things that ever happened to inventors. The problem is companies like apple want to patent every stupid little thing just so they can push out competition, when really the whole purpose of patenting was to encourage creations, while protecting them from being stolen or manufactured by others.


I don't blame this nvidia get for getting upset but this is a pretty arbitrary thing - unless explicitly told, how is anyone supposed to know such a patent exists, and how did this nvidia guy happen to know this code was being worked on? I personally feel that patents for software need to be based on a complete library or program, not a piece/feature of a library or program.

[Compholio]

It took a little digging, but it looks portions of the S3TC patents were considered invalid for obviousness (page 276):
http://www.scribd.com/doc/61046442/1...lusions-of-Law

Maybe Valve could get some clarification for us on whether the patents are still applicable with those parts thrown out, or better yet just get HTC to give open source a free license.

[Dukenukemx]

Two questions come to my mind:

- if it's patented, how did it become part of the OpenGL standard?
(How is it possible that the whole industry uses those things without thinking and getting themselves in problems?)

Don't know how but S3 created the technology, and I think HTC owns them now? HTC did sue Apple over the patent, but Apple got away with it.

- when will the patents expire?

I think it's 14 years but could also last for 20. The S3 Savage3D that first came with S3TC was introduced in 1998. So I'd imagine that it's already been 14 years. Either that or we have to wait another 6 years.

Though it's strange that a company like S3 is now in the hands of HTC. This now involves cell phones like Android and iOS, which is now a huge patent war zone. Like I said, if Apple got away with it then maybe the patent is over by now?

[allquixotic]

I'm for abolishing software patents entirely, but I'd be almost as happy if they'd impose more strict regulations around them:

1. Vetting the patented topic to make sure that it isn't obvious, in a method I'll call "expert jury" -- where Professors and top craftsmen in every field are required to register for a patent jury, and are randomly selected based on their industry of expertise to sit on a panel evaluating one patent every couple of years
2. Making the patented information immediately public at the beginning of the patent submission process, with a user friendly website where individuals or companies can go to submit evidence of prior art or to send notice that they have in their physical possession such evidence (i.e. if the proof can't be sent in a file attachment)... if prior art evidence is evaluated and confirmed to be legitimate, the patent application is rejected
3. Requiring that all patents related to mathematical algorithms, software, or computer information technology must have an official licensing channel for individual members of the public; any citizen or corporation must be allowed to buy a license to use the patented subject within a product that they either create themselves, or have purchased or obtained from someone else. They must be permitted to purchase this license without buying any other products from the patent licensor, and the maximum fee for the patent license must be no more than _____ (fill in the blank; something like $10 USD maybe) for a non-redistributable license and ______ (fill in the blank; something like $5000 USD per "product" maybe) for a license that can be conveyed onto the licensee's customers. Each patent license must not restrict the number of different applications in which an individual licensee can use the patent; for example, a licensee of a codec can use that codec on an unlimited number of devices and operating sytsems. For redistributable product licenses, it is acceptable to require that downstream licensees (i.e. those licensees who receive an indirect license as part of a product purchase) may only use the patent in the particular product that they buy from the direct licensee entity.
4. Reduce the patent validity term for software/algorithm/math/information technology patents to 2 years.

If these rules were in place:

1. Floating point textures would never have become a patent because it's too obvious and would never pass the expert jury or the prior art vetting.
2. There'd be a way for an individual to buy an S3TC license for all of your devices/OSes on an unlimited basis for a $10er. Ditto for MP3, AAC, ad nauseum.
3. Legitimate inventors who come up with something totally inspiring, transformational, and nothing even remotely similar to what's been done before would be able to make a constant stream of cash flow from licensees.
4. People who want to comply with the law would be easily able to by knowing the full extent of what patents are out there, and how to affordably license them.
5. There would be a sea of vendors able to cheaply release products containing previously-patented technologies just two years after the inventor has made their gravy. This means that vendors would be less likely to start patent wars and litigation (saving taxpayer dollars in the court system) and more likely to just wait out the 2 year term, or else license the patent. Either way, vendors wishing to use the latest state of the art technologies would be able to do so without having to give a ridiculous amount of money like 20% of their profits to the patent licensor.

[DeepDayze]

It took a little digging, but it looks portions of the S3TC patents were considered invalid for obviousness (page 276):
http://www.scribd.com/doc/61046442/1...lusions-of-Law

Maybe Valve could get some clarification for us on whether the patents are still applicable with those parts thrown out, or better yet just get HTC to give open source a free license.

That would be a great question for someone in the audience to pose to the Valve engineer...maybe if he can get the answer to that question that would be great and even better if HTC could be convinced to grant a patent license to the OSS community. In return HTC could have a fully working and supported driver and library or their own devices