Comment by ohazi | Hacker News Reader

Comment by ohazi | Hacker News Reader

> > The cost is further reduced by employing only static or dynamic ram as a means for holding the state of the system. [...]

> Dynamic ram? Can I poke fun at them for not knowing what RAM is? Yes I can.

> The above paragraph succinctly describes malloc(), which is perhaps one of the most fundamental advancements in modern computing. The problem is that Al didn’t invent it. Memory management (the ability to take RAM and then give it back when your function is done) goes back to the dawn of computing.

No, this is wrong.

Nathan is referring to dynamic memory allocation, which is what malloc does, but that isn't what this line in the lawsuit is talking about.

Dynamic RAM, or DRAM is RAM that stores state in a capacitor. This is what your laptop uses -- DDR4 SDRAM is "Double data rate (version 4) synchronous dynamic RAM". In DRAM, the charge on the capacitors slowly dissipates, so DRAM needs to be "refreshed" periodically, which is why it's referred to a "dynamic" RAM.

Static RAM, or SRAM, is RAM that stores state in a register-like structure (usually cross-coupled inverters). It's "static" in that it doesn't need to be refreshed -- as long as the supply voltage is maintained, the bits stored in an SRAM will remain intact.

I agree that this patent lawsuit is bullshit, and I really hate to derail the conversation, but I'm surprised by Nathan's misunderstanding here, given that SparkFun makes embedded boards that use all kinds of memory. SRAM/DRAM isn't unusual terminology. The distinction doesn't even matter in the context of this line in the lawsuit, but if you're going to make a point of nitpicking about something, you should really make sure that you're actually right about it first...

You are mistaken. The lawsuit itself is clearly referring to dynamic memory allocation. It refers to "static or dynamic ram" to mean RAM in general, but the paragraph has nothing to do with the characteristics of SRAM vs DRAM.

From page 6 of the lawsuit [0]:

> in fixed function implementations all possible functional requirements of the necessary data processing must be implemented on the silicon regardless of the final application of the device or the nature of the data to be processed.

> Therefore, if a fixed function device must adaptively process data, then it must commit silicon resources to process all possible flavors of the data.

> The inventors therefore sought to provide a new apparatus for implementing systems on a chip that will enable the user to achieve performance of fixed-function implementation at a lower cost.

> The lower cost is achieved by removing redundancy from the system. The redundancy is removed by re-using groups of computational and storage elements in different configurations.

> The cost is further reduced by employing only static or dynamic ram as a means for holding the state of the system.

> All of this reconfiguration can take place dynamically in run-time without any degradation of performance over fixed-function implementations.

> The present invention is therefore an apparatus for adaptively dynamically reconfiguring groups of computations and storage elements in run-time to process multiple separate streams of data and control at varying rates.

As you can see, it's saying that rather than pinning RAM (as previous fixed-function approaches did in the 80s), this system allows you to request memory dynamically.

I mean, it's a garbage patent anyway and not really worth so much time trying to analyze what they are claiming, but Nathan isn't wrong - the patent really isn't referring to SRAM and DRAM.

[0] https://cdn.sparkfun.com/assets/home_page_posts/3/9/1/7/Alta...

The lawsuit is decidedly not referring to dynamic memory allocation. There's no way to "request" a runtime-specified amount of memory dynamically here.

The memory in question is a tiny, fixed-size block of RAM inside the chip that's directly connected to the various logic blocks. There is no way to allocate or request anything -- that memory is all you've got, and all of the bits have been spoken for by the time the chip is taped out.

What they seem to be describing is something resembling a C union mapped directly onto that RAM block. That's not dynamic allocation, that's "these bits are interpreted differently depending on which logic blocks are active, so that we don't have to have a separate RAM block for each functional unit, because that would be a waste of space."

Every legal combination and order of configuration structures in this RAM block is specified and tested, that's why it's not dynamic allocation. If anything, it's a really big list of possible static allocations.

I agree with you that the SRAM/DRAM distinction is immaterial, but that's just how patents are written.

But of course, all this is a complete tangent. Nothing that SparkFun makes has anything to do with any of this.

What chance does our legal system have of gauging the validity of these patents when we can't even get a forum of technical people, with no horse in the race, to agree on what these things mean?

In a reasonable world, benefit of the doubt shouldn't be given to the patent holder when there is genuine confusion. It should be incumbent upon the filer to describe their invention in a way that is clear to experts in that relevant field.

Sadly I suspect the world we live in isn't so reasonable.

The patent holder didn't receive their patent by accident. It was reviewed by an examiner in the patent office. Patent examiners are skilled technical researchers who are trained to comprehend the matter they are examining. Once a patent has been granted, indeed, there is benefit of the doubt that the patent is valid. However - when it comes to whether the patent applies to a given invention, that's a matter of debate in a courtroom.

That’s certainly a good description of how the patent registration process is designed to work. Unfortunately reality, as represented by the broad proliferation of exceedingly moronic patents, begs to differ.

In think you are grossly overestimating the effort and expertise being applied in patent approval.

Or more specifically, I think you’re overestimating the expertise of the examiner in the relevant field.

Patent examiners seem to grant all sorts of crap that is prior art. I have very little faith in them.

Edit: removed example, it was wrong.

This is also completely ignoring economics of quality for government - the folks working there typically are either bottom rung (willing to accept non-competetive lower pay), or as a 'civic duty' of some sort. There is some exception perhaps for the intelligence community (NSA et all) because they have had somewhat a monopoly in that legally speaking, although again, there are private entities out there you never hear of which may be of higher quality.

But even there, talk to pretty much any vet - you can quickly surmise that as far as classifications go, most stuff is over classified without valid reason, because the safest thing to do for the low level peon is slap a "top secret" on anything and everything...

And there's plenty of evidence otherwise as others have mentioned - most patents are garbage and should never have been given out.

My understanding is that this is basically the endgame of all patents related to computing. You want to patent something so vague and poorly defined that it can be used to make a claim against anything.

Yes, that is very common for patents, but is not the case here. The patent and even lawsuit are very explicit that the "invention" works with both static and dynamic RAM. There is no point where they claim that dynamic RAM is better or some novel part of their invention. They are instead excluding the need for things like ROM.

> As you can see, it's saying that rather than pinning RAM (as previous fixed-function approaches did in the 80s), this system allows you to request memory dynamically

I must have missed the "as you can see" part since I see nothing in the quoted passages that would contradict the notion of a device configured by the contents of some SRAM or DRAM cells. At the very best one could say that the text might be ambiguous. After all, "cost is further reduced by employing only static or dynamic ram as a means for holding the state of the system" is a statement that you could easily make about a bog-standard von Neumann computer as well.

No, you are obviously wrong here. Even the wording of the lawsuit is extremely clear: "The cost is further reduced by employing only static or dynamic ram as a means for holding the state of the system." - that is, they only use S/D RAM for holding the state of the system, no other kind of memory.

I have no idea how you can misread this to think it refers to requesting memory dynamically - they explicitly say "static or dynamic ram" - they claim the system can work with static RAM just as well as dynamic RAM. How do you or SparkFun come up thinking they are saying dynamic RAM is better?

You are cherry picking:

"The ‘434 patent refers to the aggregate of the dynamically reconfigurable computational and storage elements as a “media processing unit.”

They are broadly talking about loading things in RAM. If there is a more specific purpose, its not mentioned here.

I.e. there is no special purpose chip here which happens to be embedded in some other special purpose device, it honestly sounds like some engineer wrote a driver that could live in RAM and they tried to patent that.

It then specifically goes on talk about it being a thing at runtime, not some pre-soldered state, i.e. software, so yes, malloc.

Utterly ridiculous.

Requesting memory “dynamically” has nothing to do with “dynamic’ in ‘dynamic memory”.

Requesting dynamically means that RAM is allocated while a program is running, not at compile or simply launch time. That is a software consideration.

Dynamic RAM means memory that requires refreshes to maintain the information it stores. A completely unrelated hardware concern.

The word dynamic simply means something that happens in time so is used in a myriad of otherwise quite different computing concepts.

I’d just like to add that “Dynamic Programming” is one example of word fluff. “Dynamic” is somewhat of an ambiguous marketing buzzword, unfortunately.

> Requesting dynamically means that RAM is allocated while a program is running, not at compile or simply launch time. That is a software consideration.

Yes, that's part of my point. The lawsuit and patent clearly say dynamic/static RAM, they never talk about software and never mention requesting/allocating memory, dynamically or otherwise.

I remember when I toured a research nuclear reactor and I was talking with the nuclear phd student and I said something about an alpha particle and the protons and neutrons. And they were trying to say alpha particles aren't made up of protons and neutrons. Made me realize than that even an expert in the field might not know something that I would have thought was common knowledge.

There are some pretty obvious followup questions:

- What are alpha particles made of?

- When an element undergoes alpha decay, its atomic number is reduced by two. Where do the protons go?

An alpha particle is a particle produced during alpha decay. It's always just a helium-4 which is made of two protons and two neutrons but with no electrons. This is why we have to mine for helium because it is produced in the ground by alpha decay.

It's different from normal helium because it doesn't have any electrons and it has an empty orbital which means it really really would like to have some electrons and also it's going crazy fast when it is created so that's why it's dangerous. But then once it takes those electrons from somewhere it becomes normal helium. You just don't want that happening in your body.

It seems like the other commenters in this thread are alluding to the electron thing maybe? Not exactly sure what point they're making.

I believe that is clear, those questions were intended to be read as ones to put to the nuclear engineer who claimed an alpha particle was not made of two nuetrons and two protons.

As far as I understand, the claim is actually that an Alpha particle is made up of quarks and gluons, the same number you would find in 2 protons and 2 neutrons, but that the proton and neutron are different things.

It sounds like there's a very technical distinction about exactly how you define a particle. Essentially, I believe that the claim is something like saying that modeling an alpha particle as 2 neutrons and 2 protons loses some information that you would have if you model it as 12 quarks, because the 12 quarks interact in (slightly) different ways than a proton and a neutron would.

But that sounds like a difference between two models, not a difference between two physical objects. To me it seems that if you were to follow this line of thinking religiously, you could then never say things like "an atom of carbon has six protons", since, well, they're not quite protons according to this logic.

Right. I was pretty surprised and amused when I was learning nuclear chemistry and realized: an alpha particle is also a helium-4 nuclei, a beta particle is an electron/positron, and a gamma ray is just high energy electromagnetic radiation.

This is sort of because the discoveries were made before people were really aware of helium nuclei, electrons/positrons, or high energy EM, they just ran physical experiments in a lab and saw there were different kinds of radiation that had different physical properties, and under close inspection of those properties, the nature of the radiation was deduced.

It's not uncommon for someone to be so focused on their specific area of a field that they loose sight of the rest of the areas.

They may have been so into their own bellybutton lint that they were very concerned about the difference between bound quarks on a QCD lattice compared to just elmer's gluing 2 free neutrons and 2 free protons together or something like that.

Yes, I have heard a lot of times how atom nuclei and alpha particles aren't made of protons and neutrons. It's always on this sense.

Some times the difference is relevant (and people dealing with radioactivity see those times more often than others), but some people insist on the pedantic distinction even when it isn't.

The binding energy of protons & neutrons into nuclei is less than 1% of their mass, so the energy required to separate a nucleus, like an alpha particle, into all its constituent protons and neutrons, is negligible in comparison with the energies required for interactions so intense that they would generate any other hadrons except protons and neutrons, for example pi mesons, which would expose the inner structure of protons and neutrons.

This is an important feature of nature, whose consequence is that the most convenient way of understanding the world is that the first approximation, which is useful for almost all practical purposes, is to consider everything as composed of protons, neutrons and electrons, which are bound by strong, electromagnetic and gravitational forces and that there are also positrons (antielectrons), which can be generated in sufficiently intense interactions and which will annihilate with electrons.

Only for few purposes it is necessary to go beyond this and take into account that there are many other hadrons composed of quarks, besides protons and neutrons, and that there are also other leptons.

The distinction between these 2 level of approximation is important and whoever likes to think that alpha particles as not being made of protons of neutrons fails to get it.

Of course, for many purposes you can ignore the strong nuclear forces and you can consider the surrounding world as being made of electrons and of a little less than 300 kinds of long lived nuclei, which are bound by electromagnetic and gravitational forces.

At this other level of approximation, you should ignore the proton-neutron composition of any nucleus, but when you talk about alpha particles, you normally mean helium nuclei having a kinetic energy high enough to not be negligible in comparison with the binding energies of nuclei, so you are not in a situation when you can ignore the nuclear compositions.

In any physical model, it is important to specify clearly which is the level of approximation at which it works. Mixing randomly various levels of approximation of the structure of things is usually a bad strategy.

Yes, but very smart people are often hyperfocused on their own particular bellybutton lint like I said.

For someone whose whole job revolves around the quark structure of nuclei they may see it important enough to "correct" someone who only thinks of nuclei as protons and neutrons flying around each other in relative isolation like planets.

Add a dash or ten of Asperger's (we're talking about nuclear physicists here) and that's how you wind up with an expert "correcting" a non-expert about a casual fact which is 99.9% correct.

Thinking that the expert was too dumb to know that fact is probably the wrong interpretation.

Thinking that the expert was making a social/communication mistake is probably the right one.

I would think that not thinking alpha particle as "made up" from protons and neutrons helpful abstraction in thinking about fission. Yes, they are, but we can't really form them or deform them at will. Specially when you already deal with free neutrons.

pretty sure the author is describing how the "runtime reconfiguration" described in the next two sentences of the lawsuit after the one you quoted is a memory allocator, not confused and thinking "dynamic RAM" means "dynamically allocated memory"

This line appears to suggest otherwise.

> Dynamic ram? Can I poke fun at them for not knowing what RAM is? Yes I can.

He's trying to nitpick about the term "dynamic" in this context, which he has misunderstood to mean "dynamic allocation" (i.e. the alternative to "static allocation"). The phrase "static or dynamic RAM" is basically never used to refer to memory allocation, it refers to the physical construction of the memory. This is especially true if the paragraph is next to a block diagram that includes an ALU... They are five levels removed from any sort of allocator.

The rest of the paragraph doesn't describe a memory allocator, it's describing runtime reconfiguration of a circuit, which you can do (and often do do) using a statically allocated configuration block or even a hard-wired dedicated memory. Again, nothing to do with malloc.

Whether a configuration memory is SRAM or DRAM based is immaterial, but patent language likes to make a point of enumerating the possibilities.

I read that line as poking fun of the fact that RAM is not capitalized because the lawyers that wrote this think "ram" is a word, not an acronym

Eh, shrug. I don't think so.

In any case, the fact that SparkFun doesn't even do semiconductor design makes this whole lawsuit even more ridiculous. They put chips onto boards, they don't design ALUs with special configurable control bits!

It actually doesn't matter if SparkFun designs chips. They do buy them, and re-sell them in devices. Patent law continues to apply to all downstream implementations of a product, until the claim is "exhausted" by a license. So, if the maker of the SoC doesn't have a license from this troll, then the troll is free to sue all of their customers, and in turn their customers, all the way down until someone pays up.

It's a bit absurd to imagine, but if someone has a patent on a CPU design, which is then used in a radio, which is installed in a car, which is driven by a taxi driver, then the patent holder can in principle sue the taxi driver for using their radio without licensing the patent.

In practice, patent trolls will figure out who they can sue who is big enough to extract $$, but not so big as to run the risk of them fighting back and invalidating the patent.

Yeah I know that this is how it works, I just think it happens to be the worst possible way for it to work. If anything, pointing to the upstream manufacturer should be the end of it.

The fact that someone can sue me over some proprietary implementation of something buried inside a chip I bought where the only way to defend myself is for me to go chase down NXP or ARM is indeed absurd.

I know nothing about patent law but this strikes me as false. I cannot be sued for turning on a radio in a vehicle that I bought or rented. I've never heard of a court thats accepted such frivolous claims either, nor of any law that asserts such.

The manufacturer of that radio can possibly be downstream sued, sure. That sounds at least plausible if they're assembling said radio without proper license for the parts. But simply using a radio in my car doesn't strike me as something enforceable nor legitimate in any way.

How do you figure a user downstream could be sued for not licensing a radio in a car they bought/rented? You're claiming we're all legitimate targets at this point in the eyes of the law (anyone using a radio in a car) so I'd like to understand this.

Anyone who makes use of patented technology that has not been licensed by someone upstream can, indeed be sued, at least in principle. Consider the case of Innovatio, who tried to go after businesses who were providing WiFi for their customers by using routers that contained chips that contained potentially infringing circuitry. https://www.eff.org/deeplinks/2014/02/infamous-wi-fi-patent-... . The reason we don’t see more suits of end customers is that it’s not economically feasible - nobody would sue millions of consumers to collect $.10 each, when they can sue the upstream manufacturer just once and collect hundreds of thousands or millions of dollars in royalties.

That's exactly what I understood it to mean when I read it.

I read it the other way, but now that you mention it, this interpretation makes much more sense.

> The rest of the paragraph doesn't really describe a memory allocator, it's describing runtime reconfiguration of a circuit, which you can do (and often do do) using a statically allocated configuration block or even a hard-wired dedicated memory.

IMO you are being much too generous in your interpretation of this trolling.

The claim is that pcDuino is infringing because it has a Cortex A9 with addressable memory. This chip does not have any of the fancy reconfiguration you have in mind, but the trolls believe it implements their patent, so I don't see how they can mean what you think they do.

No, I completely agree with you.

I'm nitpicking about a nitpick, and I feel kind of bad for derailing the conversation, because the patent assertion is completely ridiculous.

I just think that Nathan shouldn't have made the jab about malloc, because making that kind of mistake detracts from his otherwise very valid criticism.

But that would also be a wrong interpretation: the runtime reconfiguration refers to something more similar to (but differen from) the programmable gate array of an FPGA, where the interconnect between all the components gets re-configured to achieve different logic while other computations are running.

Reply on news.ycombinator.com