Ok, maybe not all people, and not all of the time. But in ranking the kinds of prior art I'd like to be able to assert against a tech patent, off of the top of my head, I'd rank system references pretty low:
- A U.S. Patent: Simple and easy.
- A foreign patent: Proving authenticity and publication is usually easy (but sometimes not).
- A journal publication: You may have to jump through some hoops, but no big deal.
- A Wayback Machine reference: Now one of those hoops is waiting (and waiting...) for a declaration through the Internet Archive's procedures. But it's not hard to get.
- A book. Now you may be dealing with librarian declarations.
- . . . [various others] . . .
- System references, e.g. prior art products.
With a system reference, the burdens of proof start to multiply. You may have a user manual that shows some claim elements, but does it accurately describe the system? Which version? Do you have the source code? Compiled code? Schematics? Do you have a physical device? Is it something you got off of eBay? Are the parts original?
Can you prove that the product was sold in the U.S. at the relevant time, and that the manuals, parts, schematics, and/or code match the product that was sold? Are you asserting it all as one reference, or is it really an obviousness combination? And so on.
It can get tricky.
Amazon Just Found This Out the Hard Way
Months ago, after a discovery dispute where the plaintiff alleged that Amazon asserted too many prior art references and obviousness combinations, Judge Andrews ordered them to reduce the number to
four specific references or combinations of references [per claim], but I do mean specific, not A plus B plus one of 4,000 other references.
Plaintiff then accused Amazon of failing to do so. Instead, it said, Amazon folded various publications, demonstrations, and source code into system references, and then asserted various combinations of those systems.
According to plaintiff, the contentions were set up like nesting dolls. The surface level invalidity grounds looked like this:
Not so bad, right? But the footnotes list some more references:
Ok, so this system reference involves some 24 exhibits. But that's not all. Apparently the exhibits themselves listed still more references. According to plaintiff, for example, a single one of the charts involved more than 20 publications:
It's not hard to imagine why plaintiff might argue that this consisted of more than one "reference or combination[.]"
Judge Andrews Was Initially OK With Amazon's Approach
In his initial order, which we discussed in more detail last month, Judge Andrews recognized how tough proving up system art can be:
I am not going to limit Amazon as requested by Plaintiff. I understand that any of the asserted prior art, such as [system reference] OAA, refers to OAA as it existed at some particular point in time. I do not know whether Amazon’s twelve sources for the OAA are more like twelve pieces of a jigsaw puzzle where one needs each piece in order to see the picture, twelve pictures taken by multiple family members of a new college graduate receiving a diploma where the pictures mostly show the same thing, but the details differ slightly, or twelve pictures of the same person taken once a year for twelve years at school picture day. My impression is the situation is more like the middle one . . . But if I did what IPA has asked, I would essentially be guessing as to how much Amazon actually needs its dozen sources. If it needs them, then it should have them.
But he noted plaintiff's argument that Amazon failed to provide a single chart for each combination that maps it to the claim elements. Amazon said that it had. Judge Andrews asked to "verify" Amazon's representation by reviewing the claim chart for a single combination that he chose at random.
Then Came the Charts
In his order after reviewing Amazon's example claim chart, Judge Andrews described what he received:
I asked for a claim chart showing how “combination #3 . . . ” read on claim 1 . . . . I received a three-page letter with 170 pages of attachments in response. . . . Even after spending more time than I should have, it is hard for me to explain what I received, but I will try.
He describes that two of the references, a patent and a publication, were detailed using "traditional-looking" claim charts.
But the third reference, a system, was disclosed by “9 papers, 2 specifications, 1 presentation, and source code" spread over seven charts.
And the last reference, a thesis used as a "secondary reference," was tucked into a chart alongside "maybe thirty-four" other secondary references.
According to Judge Andrews:
I think through a lot of tedious effort, [plaintiff] can see for each asserted reference what it is that Amazon says it discloses. . . . I do now better understand [plaintiff’s] frustration. If I am properly understanding the claim charts, essentially, for the one combination, Amazon says, here are nine references (not counting [the secondary reference]), each of which discloses every limitation, and, in combination, they also make the claim obvious. I doubt that such an approach actually gives any reasonable notice as to what Amazon’s theories are.
The Court Forced Amazon to Cut Back
Judge Andrews resolved this by ordering Amazon to cut back to just a single reference to describe each system:
I am going to order Amazon to reduce its asserted prior art as follows: (1) for systems . . . , absent a declaration under oath as to why more than one reference (not including source code) is needed to describe the system, Amazon needs to pick its best reference describing the system, and the other references will no longer be asserted in support of the system . . . .
He also ordered Amazon to eliminate all duplicative references from its obviousness combinations:
where there are multiple primary references . . . , absent a declaration under oath as to why more than two primary references are needed (i.e., what is there might be missing from one that is not also missing from the others), Amazon needs to pick a combination that does not involve multiple references that meet the same claim limitations.
And that's why people hate having to assert system references.