Hello, I'm interested in using the Fluid Dynamics for Games paper in a small commercial application, but despite a lot of Googling I'm still unclear on exactly what is patented, for example is it only in combination with the FFT that the patent applies? So far I've been unable to contact Stam about it so does anybody know exactly what the situation is?

Quote:

Original post by Novelty I'm still unclear on exactly what is patented

Welcome to the legal minefield!

You just uncovered one of the many troubling aspects of software development. You can hire lawyers, legal aides, and other researchers to find patents that might cover the things you are doing. Ultimately there is nothing preventing some obscure patent holder from sending their attack lawyers on you for a real or perceived violation.

{Cue people the globe describing how their own regulators handle this kind of patents better or worse than other places.}

Depending on your situation (you never really made it clear why you were interested) you should talk to a lawyer, your company's legal team, or your own boss for CYA documentation. Or you can just try to avoid the details in the patents you find and otherwise pretend that no restrictions exist.

Business is all about risk management. There is a constant risk of legal threats, including patent claims. The most you can do is try to minimize the risk, you cannot eliminate it.

You can read the patent filing here.

> So far I've been unable to contact Stam about it

It's unlikely the author is going to help you go around a patent his employer paid for and benefits from. Just think about it. You need to find that out on your own, or better, get advice from a patent attorney.

-cb

Welcome to the wonderful world of intellectual property rights - you will need to make sure you have indemnity insurance (see http://www.gamesure.com/ and others for info on what that is) to cover you against IP related claims.

Quote:

Original post by cbenoi1 You can read the patent filing here. > So far I've been unable to contact Stam about it It's unlikely the author is going to help you go around a patent his employer paid for and benefits from. Just think about it. You need to find that out on your own, or better, get advice from a patent attorney. -cb

Is this a common assumption? I'd like to pay the owner to use it where applicable, which isn't turning out to be easy. Because that could explain something.

However, yes, I've read that and that's where the complications emerge, the patent description is mostly (say, 95% or more) exactly the same as the paper except a fast Fourier transform (FFT) is specified instead of a less accurate Gauss-Sidel relaxation, the patent text goes on to say relaxation is not precise enough. The FFT is another patented algorithm.

So it appears to my untrained eye that Stam could be offering a way around the FFT patent in the original paper which uses relaxation, and in doing so allows public domain use of his own fluid solver incurring neither patent at cost of speed and accuracy. This does not seem to be specified anywhere I've so far seen it just seems to make sense from reading the patent and paper.

Quote:

Original post by Obscure Welcome to the wonderful world of intellectual property rights - you will need to make sure you have indemnity insurance (see http://www.gamesure.com/ and others for info on what that is) to cover you against IP related claims.

Great, I'll look into this.

Quote:

Original post by Novelty FFT is another patented algorithm.

Nope.

The original FFT algorithms were from Gauss in the very early 1800s. Those certainly aren't covered by patents.

There are hundreds of research papers describing variants of the algorithm, and those aren't covered by patent, either. (Maybe one or two managed to get a patent by an overzealous company and a naive patent cleark, but they would be the exception and not the rule.)

You can download optimized libraries for FFT for most modern processors. Intel's Performance libraries and AMD's Core Math libraries both include great implementations.



There are, however, some hardware-based FFT accelerators that are covered by patent. Those are patents on the hardware accelerators, not on the algorithms themselves.

> I'd like to pay the owner to use it where applicable,
> which isn't turning out to be easy.

The assignee is SGI Inc, not Stam. Did you try and contact SGI directly?

> Is this a common assumption?

Assumption? Hell no. I worked at a few large corporations and we've always been given indications to stay clear of any patent-related question and instead relay them to the Legal Dept. Even for published patents. Any admission by an employee may be used to undermine the patent's validity in court. One company I worked for even had a document distributed to developers and we had to sign a waiver stating we have read it all. Cover to cover.

> {...} which isn't turning out to be easy. Because that
> could explain something.

If that request was indeed forwarded to the Legal Dept internally, it's there that you will find the answers to your questions.

-cb

Quote:

Original post by frob Quote: Original post by Novelty FFT is another patented algorithm. Nope. The original FFT algorithms were from Gauss in the very early 1800s. Those certainly aren't covered by patents.

Ah, I was reading the text to mean Stam was using a type of FFT specific to the fluids somehow which was patented, but I just realized I read this page wrong:

http://www.nada.kth.se/~gustavt/fluids/

I'm still looking at the current status between Alias and SGI before I dig to find legal contact for either.