Marc is a patent attorney and shareholder of the intellectual-property law firm of Poms, Smith, Lande & Rose in Los Angeles, CA. Marc specializes in computer law and can be contacted at 73414.1226@compuserve.com.
Like it or not, software patents are here to stay. Courts are enforcing them, the Patent Office is liberalizing rules for granting them, and copying usually does not have to be proven.
Unfortunately, few nonlawyers have any idea how to determine whether a software patent has been infringed. The patent owner often cries infringement when the software in question contains what he regards as his invention. Those accused of infringement, on the other hand, often feel protected if their system lacks some of the features described in the patent.
Both of these approaches are wrong. This month, I'll explore how to determine the question of infringement.
The first thing you'll need is a copy of the patent. Patents can be ordered from the Commissioner of Patents and Trademarks (Washington, DC 20231) for $3.00. A faster source is MicroPatent at http://www.micropat.com, where you can download an entire patent for 25 cents per page or order just its text by return e-mail for $3.00.
For all but the most superficial analysis, you will also need a copy of the "file wrapper" and "references cited." The file wrapper is the name the Patent Office gives the file it maintains on each patent application. After someone files for a patent, all of the communications between the Patent Office and the inventor are stored in the file wrapper. A copy of this file can also be ordered from the Commissioner of Patents and Trademarks for $150.00.
On the face of every patent is a list of references cited. After an application for patent is filed, the case is assigned to an examiner who is an expert in the relevant technology. The examiner conducts a search for the closest technology previously known. The patent applicant may also submit prior art. Collectively, these documents are known as the "references cited" and are itemized on the first page of every patent. If the document is itself a patent, a copy of it can be obtained by following the procedures discussed earlier. If it is a published article or other item, a copy is usually contained in the file wrapper.
A patent usually has at least one illustration of the embodiments of the invention, followed by text. The first section, "Background of the Invention," usually describes preexisting related technology and its problems. Next come the "Summary of the Invention" and "Brief Description of the Drawings." The last (and usually longest) section is typically captioned "Detailed Description of the Preferred Embodiments."
None of these sections delineate the scope of the technology embraced by the patent. That scope (and what you need to study carefully to determine the question of infringement) is exclusively set forth at the end of the patent in one or more separately numbered paragraphs called "claims."
Each claim describes the boundaries of what the patent covers. A good analogy is a real-estate deed. A deed contains phrases such as "40 feet north of the oak tree at the northeast most portion of the property, 30 feet due east," and so on. A patent claim is just like the deed on real property.
A "direct" infringement of the patent occurs when each element recited in a single patent claim is present in the accused party's product or process. (Some claims recite a process; others a product.) Here are some guidelines:
If only the terms of one claim must be satisfied, you might be asking yourself why the patent contains multiple claims. The answer is simple: The validity of a patent claim can be attacked during litigation by proof that it embraces a system which was known in the prior art. Patent applications therefore usually contain several claims, each of a different scope. Broad claims embrace the greatest number of systems, but are the most vulnerable to a validity attack. Having a spread of claims provides backup.
It is usually best to analyze infringement of the broadest claim, that is, the claim with the fewest number of recited elements.
In many patents, no single claim is broadest. For example, claim 1 might include the language "a register containing the address of a local subroutine," while claim 5 might require "a 16-bit register containing the address of a subroutine." Claim 1 is broader than claim 5 because it embraces a register with any number of bits. Claim 5 is broader than claim 1 because it embraces a subroutine which is not local. In this common situation, several claims must be analyzed before it can safely be concluded that there is no infringement.
When determining infringement, "dependent" claims can safely be ignored. A dependent claim refers to another claim; for example: "The system of Claim 1 further including a compression algorithm."
Dependent claims add additional elements and thus are always narrower in scope than the claim to which they refer. "Independent" claims, which do not refer to any other claims, are broadest, and should be studied.
The precise meaning of language in a patent claim is often unclear when it is first read. It will usually become clearer after the entire patent is studied. The documents in the file wrapper, particularly the remarks of the patent applicant, also often clarify the patent claim. During the efforts to procure a patent, the applicant often distinguishes the invention from the prior art and, in the process, usually provides clarification of his patent claims.
Even after this study, the exact meaning of certain claim phrases may still be uncertain. In this case, ask yourself what this phrase would have meant to a person of ordinary skill in the art at the time the patent application was filed. If the claim recites a "software module," would this have embraced firmware embedded in ROM? (Don't ask me for the answer; I'm just a patent lawyer!)
Consider also whether other language in the patent or its prosecution history explicitly or implicitly defines the claim phrase in a manner contrary to its ordinary meaning. Patent-claim language is given its ordinary meaning in the art, unless a contrary meaning was clearly expressed.
Although language in the claim is interpreted in light of the entire patent and its prosecution history, I repeat that a patent claim is not restricted to the illustrative embodiments of the invention discussed in the patent, unless the patent claim itself recites the features of those illustrative embodiments. Thus, when the patent claim says "register," it is not restricted to a 16-bit register, even though a 16-bit register is the only type of register described.
One type of patent-claim element does not follow this rule: an element written in "means-plus-function" format. These elements recite a "means" for performing a specified function. An example is "memory means for storing a pointer." Under the general rule just outlined, this element would be interpreted to encompass any type of memory device that stored a pointer. But the rule for claim elements written in means-plus-function format is different. These claim elements embrace only the specific means described in the patent for performing the specified function and its equivalents.
If the only type of memory described in the patent were RAM, the claim language "memory means for storing a pointer" would be restricted to RAM and its equivalents.
Determining equivalency for language in means-plus-function format is a mushy business. Generally, the courts look to see if the element in the accused party's software is regarded by persons of ordinary skill in the art as interchangeable with the elements described in the patent. Although some latitude is afforded, it is usually less then all of the means that could perform the recited function.
There is often an element in the software of the accused party that is similar to the element in the patent claim, but not accurately described by it. If the difference is "insubstantial," an infringement will nevertheless be found under the "doctrine of equivalents."
Suppose the patent claim requires "RAM to store a pointer," but the software in question utilizes a shift register to store a pointer. Plainly, a shift register is not RAM. But in many applications, the difference might be insubstantial. The "doctrine of equivalents" was created to ensure that such an insubstantial difference does not avoid infringement. If the accused party's software contains an equivalent of the missing claim element, it can nevertheless be an infringement.
The classic test for determining equivalency is whether the element in the software in question performs "substantially the same function, in substantially the same way, to achieve substantially the same result" as the element recited in the patent claim.
During an inventor's efforts to obtain a patent, he often makes clear that his invention does not embrace certain types of systems. When he does, the doctrine of "file-wrapper estoppel" bars him from later attempting to interpret his patent claim to embrace these systems under the doctrine of equivalents. He is said to be "estopped" (Latin for "stopped") from interpreting his claim to embrace these systems.
Consider the following example. The inventor applied for a patent and recited "memory for storing a pointer" as one of his claim elements. This claim was rejected by the examiner because of a prior patent, which disclosed a shift register that stored a pointer. In response to this rejection, the inventor amended his patent claim, replacing "memory for storing a pointer" with "RAM for storing a pointer." Along with this amendment, he argued that his invention is RAM for storing the pointer and that RAM is very different from the shift register shown in the prior art. His claim was then allowed.
Now, the inventor makes an accusation of infringement against software that has every element of the patent claim, but uses a shift register instead of the claimed RAM. Had the foregoing amendment and remarks not been made, the inventor might credibly argue that the accused party's shift register constitutes an insubstantial variation of the claimed RAM and, accordingly, that the accused party's software infringes under the doctrine of equivalents. But because the inventor told the Patent Office that his invention did not embrace a system using a shift register, he is now barred under the doctrine of file-wrapper estoppel from claiming otherwise.
Most software-patent claims recite elements in addition to software, such as memory, a display, or a microprocessor. In the computer industry, however, it is common for a company to sell only one of these components.
Can a person be held liable for infringement if he has manufactured or sold only one component of the infringing system, such as a piece of software? The answer is yes--as a "contributory infringer." But three legal requirements must first be met.
One who induces another to infringe a patent can also be held liable as an "inducer" of infringement.
Like contributory infringement, inducement requires knowledge of the patent. It does not, however, require the person to have manufactured or sold any component, let alone one with no substantial use other than in the combination of elements recited in the patent claim.
An allegation of inducement of infringement is often brought against officers or directors of companies alleged to have manufactured or sold a system which directly infringes a patent claim. The individuals did not manufacture or sell the product themselves, but they induced their company to do so, sometimes with knowledge of the patent. Along with the company, these individuals are often sued to ensure that there will be an adequate source to pay any judgment, as well as to increase the pressure to settle.
Table 1 summarizes the requirements for the three types of patent infringement.
Whether or not the accused party's software meets the requirements of a patent claim is not the only consideration. Although the Patent Office has decided that the invention warrants a patent, that decision can be reversed in a lawsuit for patent infringement. The validity of a patent is most easily assailed when the infringer locates prior art which is closer to the patented invention than the prior art of which the Patent Office was aware. (This art is listed on the face of the patent under References Cited.)
Another popular ground for attack is when the inventor failed to disclose in the patent the "best mode" of which he was aware for implementing his invention. Substantial delay in asserting a claim for patent infringement can also cause a loss of rights. For a more detailed discussion of these other considerations, refer to my column "Patents: Best Protection for Software Today?" (Dr. Dobb's Sourcebook, September/October 1995).
The steps for determining infringement have now been set forth. The real difficulty often lies in predicting how a court will interpret the meaning of certain patent-claim phrases, whether a court will find equivalency in missing but similar elements, and whether a court will limit equivalency by the doctrine of file-wrapper estoppel.
Patent owners should move quickly to enforce their patent, as rights can be lost by delay. But they should also move carefully. Companies accused of infringement can sue the patent owner to have the patent declared invalid. A reckless charge of infringement, therefore, can lead to loss of the patent.
Companies charged with infringement should almost always seek a written opinion from a patent attorney before proceeding to manufacture or sell the product after being notified of the patent. Otherwise, they are likely to be found to be a "willful" infringer if they are sued for infringement and lose. Once found to be a "willful" infringer, the court can treble the damages and assess attorneys' fees.
Direct Contributory Inducement
Infringement Infringement Infringement
All claim Yes No No
elements needed?
Knowledge of No Yes Yes
patent required?
Non-staple claim No Yes No
component needed?
Inducement No No Yes
required?