Mathematical Technical Concepts Testimony

Winsorizing: How to Complicate a Simple Idea

Last year I came across the use of this term and had to look it up. I ultimately understood the idea as soon as I read the description, but it made me think of the propensity of technologists to invent terminology. But, of course, one benefit of creating specialized languages is excluding those who do not understand the language.

For example, the legal profession excels at taking terms and evolving them to have novel or bewitching meanings.

I learned this term when reading the information in a financial trading case. Then, I figured out why it was called “winsorizing.” It was nothing mysterious: the person who first described it in literature germane to that field was surnamed Winsor.

I was recently reminded of this as I read the trial transcript in preparation for my testimony. The attorney called attention to this term and explained its meaning. I realized that an essential part of what I do when demystifying technology is to try and come up with intuitive explanations for what we do and why.

So, winsorizing is the process of removing “outliers” from the data. For example, the attorney spoke to a jury of a dozen people. The example I would have used would have been to explain that if I had a group composed of the jurors plus Elon Musk, the average net worth of the group would be around the net worth of Elon Musk divided by 13. One data point dominates that group. Thus, the idea of removing “outliers” is to remove points that have a disproportionate effect on the group. If we don’t do that, we might conclude that jurors are rich, which is not likely the case. Removing such samples usually leads to better analysis, but it may not be evident until you explain why. Tying it to something the reader (for a report) or audience (jury) can understand is powerful and helpful in making a random term “make sense.”

The Elon of Pineapples

Another example of this I like to use also appeared to me while reading the trial transcript. Covid is still very much an active threat, and they are testing people to ensure they do not have Covid. There is a presumption that a positive test result means you have Covid, but that is often not the case. This is a general problem with tests.

My usual way of explaining this is to start with a highly accurate test: 99.5%. So, for 1000 people, that test will find five positive results even if none of those people have the condition.

In this case, the “positive predictive value” of the test is 0%. Therefore, the probability of someone with a positive test result having the condition (whether it doesn’t exist or doesn’t apply to the population) is always a false positive.

That’s not the real world in which we live, though. The condition does exist but is rare. So, if the need for which we are testing is 0.5%, then 5 people in 1000 have an actual positive test. Our 99.5% accurate test will find almost 5 people (in the 995, that shouldn’t be positive), and thus we expect to see 10 people with a positive test result. Not the “positive predictive value” is approximately 50% (half-true positives, half false positives.)

Why does this matter? The “rapid antigen” test for Covid is between 45% and 97% accurate. It becomes more challenging to evaluate some of this because as the test becomes less accurate, we have to start worrying about false-negative results. For the moment, let’s ignore that because my point is more focused than that. At 97%, we get 30 false-positive results in a group of 1000 people. So, how do we protect against that?

Easy. We omit people that are more likely to be negative. This helps increase the prevalence of the condition. In my earlier example, the prevalence was 0.5%. If the prevalence had been 5%, so 50 people in 1000 were positive, and we had 5 false positives (4.75 since a false positive can only happen in the group of 950 negative people), then our positive predictive value is much better: 90%.

Thus, usually, we’re told to only test “if you have symptoms.” This helps you eliminate the people that are unlikely to have Covid, and thus for the group that does test, the prevalence is higher, and the “positive predictive value” of the test is better. In Court, the prevalence will be relatively low, and thus a positive result without symptoms is likely to be a false positive. With 97% accuracy, there are 30 false positives per 1000. If the prevalence is 1%, there are 10 true positives. The “positive predictive value” is 25% (10 true positives out of 40 total positives.)

Explaining things systematically takes time and patience, but it is advantageous because it helps people better understand what we’re doing and why we’re doing it.

Now, I hope my Covid test is not a false positive so I can provide my testimony to the court successfully.


Blockchain: Fad for Innovative Technology?

Recently, I wrote about how blockchains work. However, I did not really delve into what it is good for? Maybe the answer is (like war) absolutely nothing!

As is so often the case in the real world, the truth is somewhere in-between. Blockchain solves a specific, real problem. However, it has also become a shiny new buzzword and is used to promote some very dubious offerings. There is so much fraud and crime surrounding blockchain-based work that stories about hacks, such as cryptojacking – where a website installs malicious software into your web browser to “earn money” while solving blockchain problems.

The technology has expanded rapidly, with numerous uses of blockchain, both benign and malign in appearance. Some actually make a lot of sense to me. For example:

For example, an old friend reached out to talk about his latest venture, This is an intriguing example of a class of uses that I have seen for blockchains that make sense. In this case, it provides an interesting technology for improving the timeshare business. Buying and selling timeshares is difficult for many reasons, several of which can be addressed using a blockchain technology solution. For example, three reasons listed in this online article, 9 Reasons Why Timeshares Are a Bad Investment, can be solved by a solution like my friend’s: it makes ownership more transparent, which can be used to ensure the person selling the timeshare truly owns it; simplifies temporary rental of it, and; encouraging a better resales process. Of course, this does not fix all of the issues, but it is an excellent example of good use.

Another case is one I suggested to add value to a friend’s work on performing “livestock facial recognition.” Such a system could be combined with a blockchain representing ownership of the given animal, providing provenance (the chain of ownership,) ease of transferability, and better tools for preventing theft. Again, this is not something we can do yet, but the technology is far enough along to make sense, and it solves a real problem.

Other uses of blockchain technology are more challenging to evaluate. For example, the Ethereum blockchain technology model is widely used because it provides abilities beyond the basic blockchain idea. A crucial part of that is the idea it can contain a contract. While businesses routinely use contracts today, such agreements are written in natural language formats that can have ambiguity. An Ethereum “smart contract” is written in a language that has a specific definition of its behavior. It enables someone writing a contract to formally validate that the agreement does what is expected. That is surprisingly hard – after all, we have lots of programmers writing many programs, yet we routinely find they struggle to “get them right.”

One specific example of a smart contract that I keep running into is the “non-fungible token.” The term “fungible” might be familiar to you, or perhaps it is slightly vague. Essentially, it captures that something can be replaced with an “equivalent” object. In cooking, many things are fungible: you can substitute margarine for butter, for example. The results aren’t necessarily identical. Some things are not substitutable. For example, a unique cultural or architectural element, such as the Mona Lisa, does not have any substitute. Thus, a “non-fungible token” is a “token” (entry on the blockchain) that represents verifiable ownership of something. This is the opposite of cryptocurrency. Indeed, few of us worry about the specific currency we have – if it is a £20 note, it is likely just as good as any other combination of currency adding up to the same amount. Of course, sometimes specific currency units become valuable for some reason, such as when they are misprinted in a way that makes them unique and interesting.

Personally, I have mixed feelings about NFTs. The scheme I suggested earlier with cows and timeshares makes sense. There are blockchain-based land title registries, which I think are a great use of distributed ledger technology. The challenge with the generic term “NFT” is that you need to understand what the NFT represents to determine if it has value. For example, I suggested NFTs that could represent ownership of a real thing, but many of the NFTs being marketed represent a reference to a virtual object. If the object is itself part of the NFT, say a digital image, and the ownership of the image is transferred, it might have value. Then again, that signed first edition of The Shining has value as well, but it does not give you the rights to do anything beyond owning that one copy. In other words, the right to create copies or derivatives need not be what was sold as part of the NFT. Thus, if you buy an NFT, you might find yourself asking if you bought a usable template for making bags of poo, or just a bag of poo itself, or a URL that points to a picture that someone made of a bag of poo that anyone else can use or access. The value of this is something you can leave on your own.

From my perspective, the interesting aspect of all this is trying to break things down and explain the process: what is a blockchain, what is a smart contract, what is a Turing Complete language like Solidity, how do these get used, etc. While potentially complicated, I have found most people can understand the basics. From that basic model, it’s then possible to explore some specific issues, whether it is for crypto-currency, smart contracts, NFTs, or any other uses that people keep finding for blockchain.

I expect that as the interest in NFTs continues to expand, I’ll have more opportunities to put my skills to good use, explaining how these technologies work and applying that to the legal cases that continue to arise around them.

Claims Litigation Patents


Generic framework for embedded software development

Each morning I receive an e-mail detailing the patent legal cases that have been filed from the fine folks at RPX Corporate, part of their RPX Insight service. I generally don’t have time to review all the cases, but I often pick several complaints to read. The morning as I write this (June 28, 2019) I noticed a lawsuit against Microsoft. Since I am involved in the technology space, cases involving technology companies pique my interest.

As I began to read the basic description of the case – the use of modular software development practices in embedded systems – the more it piqued my curiosity. After all, I did quite a lot of work with Windows NT starting before it was released in 1993. One of its key features was the use of it in embedded systems. Since I had intended on describing how I do patent analysis, I thought this would be a good one with which to start, as I have some expertise in this area, having co-authored Windows NT Device Driver Development, a book that is still commonly available and still relevant today.

The place to start is at the actual original patent: this is the definitive document on what this patent covers. It includes the specification, which is text and art that explain the problem as well as the general intent of the techniques of the patent itself. In my experience, having been involved in about a dozen patent applications and patent prosecutions, the specification is generally drafted with the input of both the inventor(s) and the patent attorney. This isn’t a requirement – someone familiar with the structure and drafting of patent applications as well as the technology could certainly do both. In my experience those skill sets seldom show up in a single person.

The actual patent is, however, the claims of the patent. These describe what is claimed by the inventors. The specification can explain existing technologies or alternative solutions, but those are background, not the invention. Thus, the claims set forth the specifics of the invention claimed by the patent.

Patents consist of independent claims and dependent claims. The independent claims stand on their own, while dependent claims rely upon an independent claim or another dependent claim. For example, dependent claims might specify particular details of a specific embodiment (implementation or realization) of the patent relative to some prior claim of the patent.

So I start by looking at the claims. If you have not read a patent before, you might find the format of it peculiar; that’s because it is shaped by many years of legal precedent, custom, and so forth. As the law regarding patents evolves, so does the language used in patents, so over time you will notice a shift in how patents are drafted.

This patent has two independent claims and 43 dependent claims. Claim 1 is an independent claim:

A method for producing embedded software, comprising:
providing one or more generic application handler programs, each such program comprising computer program code for performing generic application functions common to multiple types of hardware modules used in a communication system; generating specific application handler code to associate the generic application functions with specific functions of a device driver for at least one of the types of the hardware modules, wherein generating the specific application handler code comprises defining a specific element in the specific application handler code to be handled by one of the generic application functions for the at least one of the types of the hardware modules, and registering one of the specific functions of the device driver for use in handing the defined specific element; and compiling the generic application handler programs together with the specific application handler code to produce machine readable code to be executed by an embedded processor in the at least one of the types of the hardware modules.

Ideally, it is good if someone familiar with the field of the patent can read the claims and understand the invention. In reality, it is common to find that the patent uses language which is not immediately obvious. The general rule then is that if the specification defines the terms, then that is the meaning we should give that term. If the specification does not define the term, then we should give it the ordinary meaning that someone who knows the field would ascribe to it (“one of ordinary skill in the art”).

One of the challenges of patent litigation can actually be figuring out what the patent means. This often involves trying to understand the context of terminology within their historical context. So, if you’re trying to prove you understand what some term that “everybody knew” actually meant you will find yourself looking through old reference materials: books, dictionaries, devices, you name it. An easy mistake to make is to assume the current meaning is, in fact, the meaning at the time the patent was filed. Many terms do retain their meaning, but some terms shift over time as the field of the invention evolves.

So what does this mean? The term “embedded software” to me means “software that operates on a device and is inherently involved in the functionality of that device”. We use embedded software all the time – it is in numerous things, including televisions, refrigerators, cars, wearable devices, etc. Thus, this is not a particularly limiting term to use at this point. Operating systems software is typically one of the essential parts of embedded systems.

So to my first read, “A method for producing embedded software” seems like a broad sweeping statement. If you have an embedded device, you need to be able to produce the embedded software, assuming there is a computational device of some sort involved (e.g., a central processing unit).

Let’s move on then: “providing one or more generic application handler programs”. Here is where the terminology starts to narrow down what this invention is about. I’m not quite sure yet what a “generic application handler program” is (remember, I haven’t read the specification yet – I’m trying to understand the patent from the claims). My mental model at this point is that we have some programs that provide a range of functions. The term generic suggests to me that there is some mechanism for providing specialization – the idea of using a “plug-in” model. This is surprisingly common in systems software; in fact it is a common technique in software in general. The first time you sit down to write the software to talk to the very first printer, you really don’t know what that software should look like. By the time you are writing the same software for your 5th or 10th printer, you’ll have noticed that there is quite a lot of commonality about how they operate, though there are typically some sections that are specific to the device. Thus, you can extract “common functionality” and separate it from “device specific functionality”. To me, this would mesh with the term generic.

I must admit, I’m still a bit vague on the significance of “generic application handler programs” because it seems broad. But let’s continue further, perhaps things will become clearer from context: “each such program comprising computer program code for performing generic application functions common to multiple types of hardware modules used in a communication system”. Well, programs in my area do correspond to computer program code for performing… something. This is one of those points where I start thinking about what one means by “computer program code”. I suspect it won’t matter, but this is one of the areas that can quickly become fuzzy. I’ll save that for a more theoretical blog post at some point.

Anything that consists of a series of instructions that can be carried out by a computer likely falls into the category of “computer program code”; having said that, I suspect in this case they mean “binary code that corresponds to the instruction set of the computer”. I will run with that for now. That leads to “… for performing generic application functions…” This use of generic still leaves me feeling this is quite broad. The next bit helps narrow it a bit: “… to multiple types of hardware modules…” This is why I was interested. Hardware! Now we are talking about an area in which I’m steeped in experience, having written my first Ethernet driver back in 1987. If you are not familiar with Ethernet (IEEE 802.1), that is one of the earliest standardized computer communications mechanisms. And then the last bit “… used in a communication system;” I’m feeling warm and fuzzy now, because “communication system” to me normally means “network” and, as I observed earlier, that’s an area which which I’m familiar well before 2001. In fact, one of the earliest examples of noticing commonality of hardware devices was networking. Rather than ask every network device vendor to write a separate device driver for their device, we could construct a common code module that implemented the things that were common across devices (hence making it generic) and leave the much smaller bits of interacting directly with the hardware, which often varies according to the specific hardware, to a developer familiar with that hardware. For programs using the network, this is also beneficial because it means we don’t need to rewrite those programs to use different network hardware. One early example of this was the Network Device Interface Specification (NDIS) developed by Microsoft and 3COM and included in various Windows versions, including Windows NT.

At least the picture is getting clearer at this point. Let’s see where we go from here: “… generating specific application handler code to associate the generic application functions with specific functions of a device driver for at least one of the types of the hardware modules”. I think this sounds like the invention wants to generate the code so that the developer can “fill in the blanks”. This would ease the burden on the developer building the hardware specific code – much like giving them a template with comments that say “add code here to get your device to do X”. This sounds like meta-programming to me. The idea of doing this is certainly not new, but perhaps it was novel in 2001. I’d need to do more research to find out.

Fortunately, a model of what this patent means is forming for me.

Next we have “wherein generating the specific application handler code comprises defining a specific element in the specific application handler code to be handled by one of the generic application functions for the at least one of the types of the hardware modules”. Remember when I was talking about the language of patents? This is a great example of it. The goal here, as I understand it, is to try and be inclusive with the language. If we start with a communications device, we usually start simply: we have a send operation and a receive operation. But if I have a communications device that only wants to receive, perhaps because it is only monitoring network traffic, I wouldn’t need a send operation at all. Rather than list all the possible operations and the various permutations, the intent here is to capture the idea that “we have a list of ‘generic functions’ that might be implemented, but don’t all need to be implemented”.

Another example? You don’t need any support for writing to a CD-ROM. It’s a waste of time. If you look at the Windows CDFS file systems code for example, you will note that it has routines for handling reading from a CD-ROM, but not writing to a CD-ROM. Makes sense. So the intent of this language is to capture the broad range of possibilities that occur in the device space. While that CDFS code is fairly recent, it is not remarkably different than the version that Microsoft publicly distributed in 1994 (the first time I saw it).

One benefit of breaking this wall of text up into smaller bite sized chunks is that we have an opportunity to breath because the sentence continues: “wherein generating the specific application handler code comprises defining a specific element in the specific application handler code to be handled by one of the generic application functions for the at least one of the types of the hardware modules”. I admit, I had to read (and re-read) this several times to try and make sense of it. So I’ll try to translate it back: we have some code that we need to talk to the hardware. We have to have some way to map the relevant parts of that hardware specific code to the generic operation being performed. In other words, to return to my hypothetical network device, I have to have some way to map from the generic send operation to the hardware specific send operation bits – after all, I can’t call the hardware specific receive code, since it can’t send a message. That makes sense to me.

We’re making progress! Next fragment: “and registering one of the specific functions of the device driver for use in handing the defined specific element”. I’m quite familiar with the concept of registering specific functions – this is a common model for layering abstraction in systems. For example, I’ve done a fair bit of file systems work over the years and it is quite common to find that file systems “insert” themselves into the system by registering the functions they provide to perform particular operations. This sort of decomposition of functionality into “common” (or generic) and “specialized” has been around for quite a long time; certainly longer than I’ve been building systems software.

Now the last fragment: “and compiling the generic application handler programs together with the specific application handler code to produce machine readable code to be executed by an embedded processor in the at least one of the types of the hardware modules.” We have some very specific bits at this point. Note that it says compiling. That would immediately eliminate (at least for this claim) the possibility that this code is not compiled – so this wouldn’t apply if the code were interpreted. I could nit-pick the point about “together”. Would this exclude the case in which you just linked the code together (e.g., you used a pre-existing library), but I’ll defer considering that further for now. I don’t really have much problem with “machine readable code”. The limitation that it is “to be executed by an embedded processor…” is similarly not really very limiting. Essentially most, if not all, the CPUs we used in 2001 (or today) can be used as embedded processors. In fact, we normally use the same device drivers for desktop computers, which are not embedded systems, and stand-alone devices, which are embedded systems. Thus, again, not really very limiting. The last bit just says it has to have something to do with at least one piece of hardware. That makes sense to me – why bother compiling code for hardware, if that hardware isn’t present in the system?

I’ll likely re-read this again, but breaking it down in this fashion helps me better understand the general scope – of the first claim.

Actually, the second claim is much easier, since it just builds upon the first claim:

A method according to claim 1, wherein providing the generic application handler programs comprises providing an application program interface (API) to enable a system management program in the communication system to invoke the generic application functions.

So, this claim just specifies one of the generic interfaces is some sort of management operation. To allow a program to access this, we need an API. APIs are certainly not new, nor are system management programs. This claim is much easier to understand. In my experience, dependent claims typically provide these sorts of small, focused instances of the broader claim upon which they are based.

At this point I’ll break. There’s more analysis that can be done here, but I’ve made a good start. I am a bit surprised at the breadth of this patent. Such breadth is a two-edged sword for the patent holder when it comes to enforcement: the broader the claims, the more likely they are to find many people potentially infringing upon the patent, but also the more difficult it is to defend the patent against work that might anticipate this. Narrow patents, in my experience, are more difficult to enforce but when you find an instance where someone is practicing the patent, the narrow patent is more difficult for them to challenge successfully.

Lucio Development LLC v Microsoft Corp, Case 6:19-cv-283.


The Journey Begins

Thanks for joining me!

Good company in a journey makes the way seem shorter. — Izaak Walton

Jump through the portal

This blog, unlike those I have done before, is focused on my consulting work in the litigation support domain. Since I am working with technology, I thought I would start looking at interesting patents, which I find through the patent dispute process, and discuss them in the context of how I would approach them as an expert.

I am the primary inventor on 11 US patents in the technology space. I was personally involved in their prosecution. I have been involved in several patent disputes in the past and while I have yet to testify at trial, I have been through the other stages of the process.

In addition to inventing those patents, I also owned them for a while, as they were assigned to me after leaving my last company. I ultimately went through the patent disposition process as well, working with a broker to sell them. Each aspect of my involvement in the patent process has taught me quite a bit about how it works.

In the coming weeks and months I’ll be sharing different aspects of the patent process from my own unique perspective. I expect to discuss:

  • An Expert’s perspective on patent litigation. I get a daily report of new patent cases filed in the United States from the folks at Rational Patent (RPX). In all fairness, I don’t have time to go through all the complaints filed on a given day normally, so I pick those that look of interest to me.
  • My perspective on patent prosecution. What distinguishes a good patent from a bad patent from my perspective as an expert as well as someone who has gone through more than a dozen patent prosecutions.
  • My experiences in monetizing my patents. For small inventors, this can be one of the most challenging aspects of the patent process. Indeed, it is only by going through the process that I’ve learned quite a bit about the process and how it works.

As an expert, one of my goals is to help demystify technology as much as possible. Arthur C. Clarke said: Any sufficiently advanced technology is indistinguishable from magic. My goal is to demystify the technology, so it is no longer magic. Hence, my tag line, a portmanteau to honor Clarke’s memory: Any sufficiently advanced magic is indistinguishable from technology.