Simplifying Patent Venue

Introduction

From the 1990s to 2017, life was good for plaintiffs in patent infringement lawsuits. In 1990, the Federal Circuit1 interpreted the patent venue statute—28 USC § 1400(b)—to allow patent venue in any district with personal jurisdiction over a corporate defendant.2 This interpretation let plaintiffs, especially patent trolls, forum shop nearly as much as they wanted.3 And forum shop they did, into the welcoming arms of the Eastern District of Texas—a federal district known, rightly or wrongly, for being friendly toward patent plaintiffs.4 But the Supreme Court’s 2017 decision in TC Heartland LLC v Kraft Foods Group Brands LLC5 changed things. That decision de facto reinstated § 1400(b)’s long-ignored “regular and established place of business” test, which provides that patent venue is proper if a defendant has a regular and established place of business in the relevant judicial district.6 Suddenly, venue was limited again.

In policy terms, TC Heartland has since proven to have both virtues and vices. On one hand, the decision curtailed patent-troll forum shopping. On the other, it effectively replaced the old patent venue rule (anything goes) with a standard (regular and established place of business). Despite guidance from the Federal Circuit,7 implementing this test at the district court level has resulted in thorny factual disputes and several difficult, unresolved interpretive questions.8

One such question is whether § 1400(b) requires that a defendant have a human agent9 conducting business at the “place of business” before the defendant can be sued in the judicial district. I will call this the “human-agent requirement.” In a world without a human-agent requirement, a defendant can have a regular and established place of business in a district even if the defendant uses automation to conduct that business remotely. Lest this all sound too abstract, consider the implications: Does a telecommunications device, like a router or server, on rented shelf space count as a regular and established place of business?10 What about a cell tower?11 Or an Amazon Locker, which lets people access packages without interacting with any of Amazon’s agents?12 Two of the three district courts13 to confront the human-agent question imposed that requirement as part of § 1400(b).14 Only one—in the Eastern District of Texas—disagreed.15 After an initial refusal to weigh in,16 the Federal Circuit recently sided with the majority, endorsing the human-agent requirement.17

At first glance, this lopsided case law gives one the impression that the Eastern District of Texas lost out simply because it was wrong. After all, the pro-requirement cases purport to derive the human-agent requirement by reading two intimately related statutes together. And on policy grounds, the cases hint that the requirement is an important tool for avoiding drastically overbroad patent venue. Because limiting patent venue was a core function of § 1400(b) (and its predecessor) in the first place, this is a weighty consideration. The Eastern District of Texas’s now-defeated opposition only adds to the appearance that the requirement is a much-needed limit on patent venue: the Eastern District is known for its forum-shopping-fueled popularity among patent plaintiffs (and patent trolls), and its stated reasons for rejecting the requirement weren’t entirely compelling.

But district courts were mistaken to adopt the human-agent requirement, and the Federal Circuit was mistaken to agree with them. The requirement has significant downsides: it is based on a statutory misinterpretation, and it will make patent venue litigation more confusing and costly than it already was. The requirement’s apparent upside—that it’s the only possible finger in the dam of patent venue—is illusory. Courts can limit venue, along slightly different (and, I argue, better) contours, simply by paying close attention to § 1400(b)’s text.

This Comment has four parts. Part I lays out the background law at play, with special emphasis on forum shopping and the widely acknowledged need to limit patent venue. Part II discusses existing case law and explains that the arguments currently in play lean heavily in favor of the human-agent requirement. This likely explains why the Federal Circuit adopted this requirement. Part III provides two reasons, discussed by neither the district courts nor the Federal Circuit, for rejecting the human-agent requirement at either the Federal Circuit or the Supreme Court level. In light of those reasons, Part IV provides an alternative path. In a human-agent-requirement-free world, courts could still put a robust limit on venue, without any of the requirement’s downsides. Specifically, courts could use the “place” requirement to define these limits instead. This route would enforce a venue limitation that really is in § 1400(b)’s text: venue is proper only if (among other things) the defendant has a substantial enough physical presence in the district to constitute a “place” of business.

I. Background Law

This Part outlines the legal history behind current patent venue requirements, paying special attention to venue’s scope. For present purposes, the biggest takeaway is that there is now a broad and well-substantiated judicial consensus that the patent venue statute should not be read to allow for unlimited (or nearly unlimited) venue. Doing so would both distort the statute’s plain meaning and undo its intended effect. Indeed, as mentioned above, this was the interpretive error the Federal Circuit fell into in 1990—an error the Supreme Court only recently corrected.

A. The Act of March 3, 1897

The current patent venue statute is a modified codification of Chapter 395 of the Act of March 3, 1897,18 (“the Act”), which had two primary aims: to clarify and to narrow patent venue.19 The relevant provision read as follows:

[I]n suits brought for the infringement of letters patent the circuit courts of the United States shall have jurisdiction,20 in law or in equity, in the district of which the defendant is an inhabitant, or in any district in which the defendant, whether a person, partnership, or corporation, shall have committed acts of infringement and have a regular and established place of business. If such suit is brought in a district of which the defendant is not an inhabitant, but in which such defendant has a regular and established place of business, service of process, summons, or subpoena upon the defendant may be made by service upon the agent or agents engaged in conducting such business in the district in which suit is brought.21

This statute was recodified in 1948—but because the recodification left the wording almost unchanged, the original remains important for interpreting the recodification.22

28 USC § 1400(b), which codifies the first half of the original statute, now provides that venue in patent infringement lawsuits (throughout, “patent venue”) is proper in either of two circumstances: “Any civil action for patent infringement may be brought in the judicial district where the defendant resides, or where the defendant has committed acts of infringement and has a regular and established place of business.”23 The second half of the original is now codified in 28 USC § 1694, which reads as follows: “In a patent infringement action commenced in a district where the defendant is not a resident but has a regular and established place of business, service of process, summons or subpoena upon such defendant may be made upon his agent or agents conducting such business.”24

The Act reflected, among other things, Congress’s worry that unlimited patent venue would lead to abuse. As the Supreme Court explained, “the Act . . . was a restrictive measure, limiting a prior, broader venue.”25 The Court elsewhere said the Act aimed to “eliminate the ‘abuses engendered’ by previous venue provisions allowing such suits to be brought in any district in which the defendant could be served.”26 In the words of one member of the House of Representatives, “[t]he bill is intended to remove [the previous] uncertainty and to define the exact jurisdiction of the circuit courts in these matters. . . . [I]t would tend not only to define the jurisdiction of the circuit courts not now defined, but also limit that jurisdiction.”27 In keeping with these aims, the Court has admonished lower courts that, like venue in other federal cases, patent venue “is specific and unambiguous; it is not one of those vague principles which, in the interest of some overriding policy, is to be given a ‘liberal’ construction.”28 And indeed, for nearly one hundred years, § 1400(b) and its predecessor did what they were supposed to do: clarify and limit patent venue.

B. VE Holding and the Rise of Forum Shopping

A 1990 Federal Circuit decision largely undid the statute’s limiting function. VE Holding Corp v Johnson Gas Appliance Co29 interpreted § 1400(b)’s use of the word “resides” to mean that a corporate defendant now resided not just in its state of incorporation (the previous rule),30 but in any judicial district in which it was subject to personal jurisdiction.31 The case’s reasoning was nuanced, but the upshot was relatively simple: when it came to corporate defendants, the Federal Circuit now interpreted § 1400(b)’s residence prong so broadly that it no longer limited venue at all.32 Plaintiffs suing corporate defendants could now use § 1400(b)’s residence prong to establish venue with ease. There was thus little reason to resort to the more stringent “regular and established place of business” prong, and that test became largely irrelevant.33

A few years after VE Holding, the Eastern District of Texas started to become a hotspot for patent cases.34 Whether the court was actively seeking this result is somewhat controversial. But if it was an intentional effort, it was an astoundingly successful one. By 2015, near the peak of the forum-shopping spree, over 40 percent of American patent infringement suits were being filed there.35 Theories abound as to why, exactly, patent plaintiffs found this district so appealing. The happy story is that the judges in the district are exceptionally knowledgeable about patent cases and manage their dockets efficiently; the sinister story is that the district’s procedural rules and juries are too plaintiff friendly.36 In any case, it is clear that patent plaintiffs love the Eastern District of Texas, and that VE Holding helped them get there: its broadening of patent venue made forum shopping drastically easier.37

Many, but not all, practitioners and scholars believe forum shopping (especially in the patent context) is a net harm that diminishes courts’ legitimacy and generates inconsistency and unfairness.38 More pointedly, a forum free-for-all is obviously inconsistent with the venue-narrowing aim of the Act of March 3, 1897.39 Even further, fighting forum shopping seems to have been one of Congress’s motives for creating the Federal Circuit in the first place.40 Yet despite some incomplete attempts by Congress41 and the Federal Circuit42 to curtail the practice, VE Holding remained in place for seventeen years. This allowed the Eastern District of Texas to gain and maintain dominance in the world of patent litigation.

C. TC Heartland and Its Aftermath

The Eastern District of Texas’s ascendancy, and the forum shopping that enabled it, took a punch in 2017. In TC Heartland, a unanimous Supreme Court abrogated VE Holding and held that “a domestic corporation ‘resides’ only in its State of incorporation for purposes of [§ 1400(b)].”43 This holding drastically restricted the scope of § 1400(b)’s residence prong, and unsurprisingly, the “regular and established place of business” prong immediately became relevant again. The impact of this change has been astonishing: in the year following TC Heartland, the Eastern District’s share of patent infringement filings dropped from 44 percent to 15 percent.44 Many filings shifted to the District of Delaware.45 This would seem to be a good thing.46 As patent law scholar Professor J. Jonas Anderson aptly puts the point, “Delaware, unlike East Texas, has some incentive to maintain an even-handed approach to patent law,” because failure to do so would “risk innovative companies choosing to incorporate elsewhere.”47

In a 2017 decision, In re Cray, Inc,48 the Federal Circuit gave district courts guidance on applying the “regular and established place of business” test. The court explained the test in terms of three requirements: “(1) there must be a physical place in the district; (2) it must be a regular and established place of business; and (3) it must be the place of the defendant.”49 Despite this guidance, district courts ran into several difficult questions when applying the test.50 As previously noted, one of these questions was whether, in addition to the three prongs explicated by In re Cray, § 1400(b) imposes a human-agent requirement—in other words, whether venue is proper under § 1400(b) only if the defendant has a human agent doing business in the relevant judicial district. The Federal Circuit recently answered “yes,” holding that § 1400(b) does indeed impose a human-agent requirement. This Comment argues that was the wrong answer.

Notably, In re Cray repeated and emphasized the Supreme Court’s view, discussed in Part I.A, that § 1400(b) should be read as a venue-limiting statute.51 The In re Cray court also discussed the history and policy concerns behind the Act and warned district courts against conflating personal jurisdiction with venue—in other words, it emphasized that VE Holding is no longer the law.52 Because all three authorities—the history behind the Act, the Supreme Court, and the Federal Circuit (by now, at least)—support reading § 1400(b) to limit patent venue, this Comment takes that as a given.53 Two questions, therefore, remain: First, is the human-agent requirement, all things considered, a good way to achieve this goal? And if not, what other avenues do courts have to stop patent venue from becoming a VE Holding–like free-for-all?54

II. Justifications for the Human-Agent Requirement

This Part has three purposes. First, it explains why courts would impose the human-agent requirement in the first place. The requirement isn’t in the text of § 1400(b) or any other statute, after all. A review of pro-requirement rulings suggests that the reason is a mix of statutory (mis)interpretation and an understandable attempt to deploy the requirement as a venue-limiting device. Second, this Part lays out the Eastern District of Texas’s reasons for not imposing the human-agent requirement and explains why those reasons are not compelling. Finally, this Part discusses the Federal Circuit’s adoption of the human-agent requirement, which relied on largely the same pro-requirement rationales advanced at the district court level. This sets the stage for the rest of the Comment, which will provide new and more compelling reasons for rejecting the human-agent requirement. The rest of the Comment will also show how, if the Federal Circuit or Supreme Court were to rule against the requirement, courts would still be able to limit patent venue.

A. Reasons for Imposing the Human-Agent Requirement

The courts in the Southern and Northern Districts of New York were the first to embrace the human-agent requirement. In Peerless Network, Inc v Blitz Telecom Consulting, LLC,55 the Southern District asked whether a breadbox-sized shelf of internet routers counted as a regular and established place of business.56 Roughly speaking, a router is a computer whose primary purpose is to direct data from a sending computer to a receiving computer.57 Many routers, like those in this case, can and do operate without a human being at their location. The court held that, though the defendants’ agents “direct[ed] telecommunications traffic through” the router shelf, the shelf could not be the basis for venue because the defendants did “not engage in business from the shelf itself.”58

In Rensselaer Polytechnic Institute v Amazon.com, Inc,59 the Northern District of New York asked whether Amazon Lockers were regular and established places of business. Amazon Lockers are “secure, self-service kiosks where [customers] can pick up Amazon.com packages at a time and place that is convenient for [them].”60 The court imposed the human-agent requirement—but found the requirement satisfied in this case. Amazon Lockers could serve as the basis of venue, the court said, because the independent contractors who serviced and repaired the Lockers on Amazon’s behalf counted as Amazon’s “agents” regularly doing business at the Lockers.61

The main rationale62 for imposing the human-agent requirement is an interpretive one. This interpretation reads § 1400(b) (the patent venue statute) together with § 1694 (the patent service of process statute). Peerless Network was the first to lay this reading out.63 But because Rensselaer fleshed it out in greater detail, I will focus on the latter case. The court emphasized that because the predecessors of § 1400(b) and § 1694 were—in the original Act—right next to each other, they “must be read together given their common statutory history and structural connection.”64 The Rensselaer court noted that the Act allowed process to be served on “the agent or agents engaged in conducting such business in the district.”65 And § 1694, similarly, refers to “his agent or agents conducting such business.”66 The court concluded: “[t]he use of the definite article and possessive determiner suggests that Congress assumed that a defendant with a place of business in a district will also have agents conducting such business in the district.”67 Thus, § 1400(b), read in light of § 1694, implicitly contains a human-agent requirement.68 Notice the court’s hop from Congress’s assumption to the conclusion that § 1400(b) contains a human-agent requirement. The court was, I suspect, relying on something like the following syllogism,69 to justify the hop:

1. Congress assumed70 a defendant with a regular and established place of business in a district would also have an agent doing business in the district.
2. Any assumption Congress71 makes about the state of the world at the moment of a statutory requirement’s satisfaction becomes a necessary condition for that satisfaction.
3. Therefore, Congress imposed a human-agent requirement.

Courts may also see the human-agent requirement as a way of keeping venue limited, thus vindicating the Act’s original purpose and avoiding relapse into a VE Holding–style venue bonanza. Perhaps this consideration was in the background of these two decisions—especially Peerless Network. Both opinions explicitly acknowledged, as has the Federal Circuit, that the Act aimed to limit a formerly broader patent venue.72 And in Peerless Network, the human-agent requirement was the only thing stopping the plaintiff from obtaining venue based on a router “the size of a breadbox” (because the rest of the statutory test had been satisfied).73 It is natural to ask: If plaintiffs can establish patent venue on the basis of a measly router, where does it stop? To borrow one judge’s phrase, a no-requirement world might “essentially reestablish nationwide venue, in conflict with TC Heartland, by standing for the proposition that owning and controlling computer hardware involved in some aspect of company business . . . is sufficient” for venue.74 Maybe the human-agent requirement, then, is necessary to hold the venue floodgates shut. Of course, this Comment contends that is the wrong conclusion. Not only is imposing the requirement interpretively mistaken and detrimental to the judicial administrability of patent cases (as Part III will argue)—it is also (as Part IV will argue) not the only way to keep venue from unduly expanding.

B. The Eastern District of Texas’s Responses

The Eastern District of Texas was the only district court to reject the human-agent requirement, and the Federal Circuit has since shot it down on this point.75 This Section explains that, even though the human-agent requirement really should be rejected, the Eastern District’s reasons for doing so were not compelling. The question in SEVEN Networks, LLC v Google LLC76 was extremely similar to that presented in Peerless Network: Google owned and operated a few servers77 on leased shelf space within the Eastern District of Texas.78 The court held that this setup sufficed for venue under § 1400(b).79 And, because Google had no agents on site operating the servers, reaching this result required rejecting the human-agent requirement.80 The court gave two main reasons for doing so.81

First, the SEVEN Networks court suggested that the human-agent requirement does not “accord with conceptions of places of business stretching back to at least the turn of the 20th century.”82 The court pointed to a litany of historical examples of warehouses being considered regular and established places of business.83 The court did not spell the argument out any further, but the view may be as follows: The original understanding of the phrase “regular and established place of business” included warehouses. Warehouses are visited by agents too infrequently to satisfy a human-agent requirement. Therefore, the original understanding of the phrase “regular and established place of business” could not have involved a human-agent requirement.

Second, the court appealed to the non-superfluity canon of statutory interpretation.84 The non-superfluity canon says statutes generally shouldn’t be read in a way that renders them (or another statute) superfluous.85 The court pointed to the Leahy-Smith America Invents Act86 (AIA), which Congress enacted in 2011. Section 18(c) of the AIA provides:

ATM EXEMPTION FOR VENUE PURPOSES.—In an action for infringement under section 281 of title 35, United States Code, of a covered business method patent, an automated teller machine shall not be deemed to be a regular and established place of business for purposes of section 1400(b) of title 28, United States Code.87

Embracing the human-agent requirement, the court said, would render AIA § 18(c) superfluous. And ATMs “are not operated, in person or remotely, by employees of the owning financial institution.”88 That means ATMs do not satisfy the human-agent requirement. But if that is the case, what does AIA § 18(c) do? It simply says ATMs, which already are not regular and established places of business, do not count as regular and established places of business. Thus, the court concluded, endorsing the human-agent requirement would “necessarily render[ ] this express exemption superfluous.”89 The requirement should therefore be rejected.

The Rensselaer opinion undercuts both these arguments. The Rensselaer court explicitly responded to the ATM non-superfluity argument by pointing out that ATMs, even though they are self-service facilities, require regular maintenance and servicing.90 According to the Rensselaer court, the person maintaining and servicing the ATM could count as the agent doing business at the ATM.91 And using this more generous conception of who counts as the requirement-satisfying agent, AIA § 18(c) makes sense again: it exempted ATMs because (even with a human-agent requirement) ATMs would otherwise be sufficient for venue.92

This same point also (implicitly) responds to the SEVEN Networks warehouse argument. On the Rensselaer court’s broader view of who can count as a business-conducting agent, there’s no reason to assume, as the SEVEN Networks court did, that warehouses lack business-conducting agents. Instead, it’s just as plausible that historical warehouses—much like modern-day ATMs—typically featured workers doing on-site service, maintenance, or transportation. Because warehouses were likely never agent-free locations in the first place, the fact that they were historically considered regular and established places of business says nothing about the human-agent requirement.

C. The Federal Circuit Endorses the Requirement

Eventually, the Federal Circuit resolved the district court split by siding with the Rensselaer and Peerless courts.93 The court, in In re Google LLC94 (Google III), was once more faced with Google servers on leased shelf space.95 The court held that the servers did not constitute a regular and established place of business, because no agent conducted business there.96 This decision, though obviously significant because it came from the Federal Circuit, doesn’t need much discussion here. Its stated reason for adopting the requirement was nearly identical to that in Rensselaer.97 Now that the Federal Circuit has endorsed it, the human-agent requirement is here to stay until either the Federal Circuit reverses course or the Supreme Court gets rid of it.98 As the rest of this Comment argues, the sooner this happens, the better.

III. Where Agency Goes Astray

The SEVEN Networks court didn’t provide compelling reasons to reject the human-agent requirement—which is likely why it ended up outgunned by other districts and the Federal Circuit itself. But the requirement should be rejected nonetheless. First, it’s based on a misreading of 28 USC § 1400(b) and 28 USC § 1694. Second, the requirement adds an entirely new (and expensive) layer of factual and legal uncertainty into the early stages of litigation. Either of these reasons alone, much less both put together, would be enough to justify rejecting the human-agent requirement—at either the Federal Circuit or Supreme Court level.

A. Statutory Interpretation

No court claims the human-agent requirement is explicit in either § 1400(b) or § 1694. Rather, pro-requirement courts interpret these two statutes together as implicitly imposing the human-agent requirement. That does not make these courts automatically wrong; written laws state things implicitly all the time. But anyone who would read an unwritten provision into a statute bears the burden of showing the statute means more than it actually says. This point is intuitive; it takes no appeal to a canon of interpretation to see it.99 Statutory interpretation, after all (whether textualist, purposivist, or otherwise), is the project of discerning what a statute requires. It is not the project of adding more to the statute than is actually there. Thus, unless there is a good justification for reading the extratextual human-agent requirement into § 1400(b), it is a mistake to interpret the requirement as part of the statute.

1. The service of process argument.

The service of process argument (that is, the argument based on reading § 1400(b) together with § 1694) might be either a legislative-intent or originalist argument.100 Roughly, the former makes the intention of the enacting legislature binding; the latter makes the original public meaning of the statute’s text binding. The Peerless Network and Google III courts did not specify how they conceived of the argument,101 whereas the Rensselaer court framed it in legislative-intent terms.102 Either way, I’ll here restate (in slightly more nuanced form) the same syllogism I presented in Part II.A:

1. The enacting Congress (or the public of 1897) assumed a defendant with a regular and established place of business in a district would also have an agent doing business in the district.
2. Any assumption Congress (or the original public) makes about the state of the world at the moment of a statutory requirement’s satisfaction becomes a necessary condition for that satisfaction.
3. Therefore, the statute imposes a human-agent requirement.

This syllogism, even if clunky, might seem persuasive at first glance. Why shouldn’t we take the assumptions of the enacting legislature or the public seriously? But as I am about to argue, the syllogism is unsound because premise (2) is false.

2. Why the service of process argument fails.

Premise (2) is false because there is a critical difference between assuming the world will be a certain way when a statutory test is satisfied and requiring that the world be that way before the test can be satisfied.103 More concretely, it’s one thing to assume that if the defendant has a regular and established place of business in a district, then the defendant will of course also have an agent doing business in the district. It’s another thing to require that, for a defendant to have a regular and established place of business in a district, the defendant must also have an agent doing business in the district. And the assumption does not entail the mandate. Thus, there’s little reason to believe premise (2).

A hypothetical should help bring out the point. Suppose that in 1940, the City of Chicago passed a statute that read, “(A) Parking a motor vehicle in a roadside Restricted Zone is illegal and punishable by a fine of $25, assessed by ticket; (B) if the police see a motor vehicle parked in a roadside Restricted Zone, they may immobilize it by putting a ‘boot’ on the vehicle’s wheel.”104 In 1940, everyone (city council and general public included) assumed that any motor vehicle that could be parked on the roadside would have a wheel—a wheel that a “boot” could be put on. But suppose that, in 2140, hovercars have been perfected and are now in widespread use. These hovercars function just like normal cars, but they, of course, need no wheels.

One day, an exceptionally stingy man, Hovercar Herbert, gets a ticket for parking on the roadside in a Restricted Zone. Herbert argues he’s not subject to the statute because the statute includes, implicitly, a “wheel requirement.” “The enacting city council and the people of 1940 Chicago,” Herbert complains, “assumed all vehicles satisfying provision (A) would have wheels. And of course they did—there were no such things as hovercars in 1940, and nobody had anticipated them yet (or if they did, they were nothing more than science fiction).” Herbert continues, “This assumption is encoded in the law—just look at provision (B)! But hovercars don’t have wheels. Therefore, my hovercar is beyond the scope of the statute.”

First, construe this as a straightforward legislative-intent argument. This version of Herbert’s argument fails with flying colors. The one thing Herbert has proved, by highlighting that the city council didn’t have hovercars on its mind in 1940, is that the city council could not have intended to impose a wheel requirement. A wheel requirement, after all, is only necessary if there’s such a thing as a wheelless motor vehicle that can park roadside. But as Herbert so astutely pointed out, the one thing we know the 1940 city council wasn’t thinking about, and could scarcely have conceived of, was wheelless motor vehicles parked on the side of the road. Not only is there no reason to infer a legislative intent to impose a wheel requirement, there’s every reason to think the legislature couldn’t possibly have intended such a requirement.

Construed as an originalist stance, Herbert’s argument fails again, but for a different reason. On an originalist view, the inquiry centers around the meaning of the statute’s text, and that meaning is defined by reference to what a reasonable member of the public, at the time of enactment, would have taken the text to mean.105 Here, unlike in the legislative-intent situation, Herbert’s argument doesn’t fail because the original public wouldn’t have expected the statute to apply a certain way to hovercars.106 Rather, Herbert’s argument fails because the original public meaning of the statute involves no mandate at all. There simply is no wheel requirement in the text’s original meaning. And the fact that the original public, like the enacting city council, would surely have assumed that any roadside-parked motor vehicle would have wheels is simply irrelevant: that’s an assumption, not a mandate. On an originalist reading, then, the text’s public meaning—and therefore the law—does not include a wheel requirement.107

Herbert’s failed arguments help show why premise (2) is false.108 On a standard legislative-intent theory, one of the few things we know with certainty is that the enacting Congress didn’t intend to impose a human-agent requirement. This is because there was absolutely no need for such a requirement in 1897. How could there be? The idea of conducting business without a human agent present would’ve been just as unrealistic to the 1897 Congress as the idea of hovercars was to our hypothetical 1940 city council. Telecommunications and full-on automation, obviously, weren’t on anyone’s mind at the time. Congress, then, didn’t intend to impose a human-agent requirement. Similarly, the text of the Act, as it would have been interpreted by the general public of 1897, simply contains no mandate—just like the text of the hypothetical parking statute. And the fact that the original public would have assumed no defendant could have a regular and established place of business without a human agent present does nothing to change that. Assumptions just do not equal mandates.

If the Act’s legislative history really did suggest Congress intended to impose a human-agent requirement, that would admittedly weaken the above arguments. Indeed, the Google III court appealed to legislative history, citing the congressional floor debates:

The Congress that enacted the venue statute stated that the “main purpose” of the statute was to “give original jurisdiction to the court where a permanent agency transacting the business is located.” Furthermore, that Congress explained that only a “permanent agency”—and not “[i]solated cases of infringement”—would be enough to establish venue. Congress’ characterization of a “regular and established place of business” for venue purposes as a “permanent agency” reinforces the applicability to venue of the agent requirement of the neighboring service provision.109

This is a puzzling line of argument. My response is just to reiterate my above point: Representative John Lacey, just like all congressmen at the time, would’ve had no conception of an agent-free place of business. Such a thing is not possible without some form of automation. So his use of the word “agency” to refer to a business merely shows that Lacey assumed regular and established places of business would feature agents. But that assumption provides no evidence at all for the idea that he, or the rest of Congress, intended to impose a human-agent requirement. And predictably, neither the rest of the House floor debate110 nor the House committee report111 nor the Senate floor debate112 suggests that a human-agent requirement crossed anyone’s mind at all. Thus, even when legislative history is invoked, it provides no support for the human-agent requirement.

The human-agent requirement is not in § 1400(b)’s text, and there is no compelling reason to read it into the text. The most textually faithful reading of § 1400(b), then, is one that takes all the statute’s requirements seriously—but without imposing an additional human-agent requirement. In other words, if a defendant has a regular and established place of business in a judicial district, then the “regular and established place of business” test is satisfied. And this is so even if no human agents conduct business at that location.

3. Three objections from a defender of the human-agent requirement.

This criticism of the assumptions-based syllogism, I believe, is ultimately correct. But it is not airtight. A defender of the human-agent requirement has several responses available, and I present them here as objections to my aforementioned argumentation. The first objection presses on service of process, arguing that if § 1694 is the only way to serve process on a patent defendant, then the human-agent requirement is de facto a part of the law. This objection is easily disposed of. The second objection is more substantial but likewise mistaken: it’s based on the idea that the expected applications of statutory provisions (that is, the outcomes their drafters or the public expected them to have) are binding. The third objection reframes the case for the human-agent requirement as an imaginative reconstruction argument. This objection is perhaps the trickiest, but it turns out that an imaginative reconstructive framework weakens, rather than strengthens, the case for the human-agent requirement.

The first objection is as follows: Even if historical assumptions are not binding, the combination of § 1400(b) and § 1694 only gives one way to serve process (on the defendant’s “agent or agents conducting [ ] business” in the district).113 So doesn’t proper service of process require an agent anyway? If so, the human-agent requirement, even if not actually in any statute, might as well be. The problem with this objection is that, as the Rensselaer court itself acknowledged, § 1694 simply “conveyed that a plaintiff has an alternative means of service beside the general methods of service now specified under Rule 4 of the Federal Rules of Civil Procedure.”114 Thus, § 1694 merely adds, presumably for convenience’s sake, another way to serve process in an infringement suit; it doesn’t provide the exclusive means of serving process in such a suit.

The second objection requires a bit of background. Some scholars have maintained that the expected applications of constitutional provisions are binding (as opposed to viewing only the linguistic meaning of the provisions’ text as binding).115 So if the drafters or original public behind a written law expected the law to apply—or not apply—in a certain way, later interpreters are forever bound by that original expected application.116 Perhaps the most famous advocate of this view was Justice Antonin Scalia. He maintained, for example, that the death penalty couldn’t possibly violate the Eighth Amendment because other parts of the Constitution explicitly contemplated executions.117 And, because this means the Framers clearly didn’t think the death penalty counted as “cruel and unusual,” Justice Scalia concluded that the death penalty isn’t terminated by the Eighth Amendment.118

Our doughty objector, then, using Justice Scalia’s expected-applications view as a springboard,119 might argue: “This is just like the death penalty situation. The text of § 1694 shows that, in 1897, everyone assumed § 1400(b) would only apply to businesses with agents. And, as Justice Scalia tells us, that expected application is binding. So there is a human-agent requirement after all.”

This objection fails for two reasons. First, Justice Scalia was arguably mistaken on this point. As Professor Ronald Dworkin famously pointed out, there’s nothing odd about believing the drafters of a law (a) legally codified a principle and (b) acted, in other areas of law and politics, in ways inconsistent with that principle.120 And if they did, it’s perfectly natural to treat the codified principle—and not the inconsistent alternative action—as binding.121 In fact, Dworkin points out, elevating expected applications above the meaning of a law’s text is arguably inconsistent with Justice Scalia’s own textualism.122 Dworkin’s point, notably, has been well-taken in the literature.123 If Justice Scalia was mistaken here, there’s no point in leaning on his view to support the human-agent requirement.

The second response is more important. Even if giving expected applications binding force isn’t a mistake, the objector has made a fatal error by equating the Act with something like the death penalty. The only reason Justice Scalia’s death penalty argument has appeal in the first place is because Justice Scalia has strong evidence that the Framers and original public (which one is relevant depends on one’s version of originalism) believed a proposition like, “the Eighth Amendment does not bar the death penalty.” Specifically, the fact that other parts of the Constitution explicitly contemplate the death penalty does suggest that the Framers and original public believed this. Justice Scalia may be right about that; I believe he goes wrong by giving binding force to that belief. But the human-agent requirement is in a different ballpark—there is no reason to think the Congress or public of 1897 believed the analogous proposition: “The Act requires that a defendant, in order to have a regular and established place of business, have an agent doing business at the place.” In fact, there is every reason to think nobody believed this proposition. As discussed previously, the very fact that agent-less places of business were inconceivable in 1897 means a human-agent requirement would have been totally superfluous at that time. Thus, even Justice Scalia’s expected-applications view can’t save the human-agent requirement.

The final objection is somewhat tricky, but it might be the most interesting of the trio. Our indefatigable objector, one last time: “Though the enacting Congress didn’t explicitly enact a human-agent requirement, that’s only because it didn’t know one would be needed. If that Congress were alive and convened today, it would have enacted a human-agent requirement.” The objector goes on: “How do I know this? I know it because the requirement is now the best way to achieve the enacting Congress’s legislative intent. And that matters. Thus, your fancy syllogism fails to represent what very well could be the real reasoning behind the Peerless Network and Rensselaer decisions.” The objector offers this syllogism instead:

1. Courts should interpret a statute in the way that—given the state of the world today—best achieves the legislative intent behind the statute.
2. Interpreting § 1400(b) to impose an implicit human-agent requirement is the best way to achieve the legislative intent behind the Act.
3. Therefore, courts should interpret § 1400(b) to impose an implicit human-agent requirement.

This objection is using what Judge Richard Posner once called “imaginative reconstruction.”124 Broadly, this approach asks the judge to “try to think his way as best he can into the minds of the enacting legislators and imagine how they would have wanted the statute applied to the case at bar.”125 Judges should consider the statutory text, background, and structure, as well as the legislative history and related statutes.126 They should also look at the “values and attitudes, so far as they are known today, of the period in which the legislation was enacted.”127 Finally (and as we’ll see, importantly), judges should look out for “any sign of legislative intent regarding the freedom with which [they] should exercise [their] interpretive function.”128 For example, a statute might stipulate that it should be read broadly.129 In contrast, a statute passed “against a background of dissatisfaction with judicial handling of the same subject under a previous statute or the common law” should be read narrowly, since it’s likely the statute aimed to constrain, rather than empower, judges.130

Assume imaginative reconstruction is a good way to read statutes. This objection still makes little headway. First, the Act aimed to resolve a troublesome circuit split.131 The Act was intended “to remove the uncertainty which [had arisen] as to [patent] jurisdiction by reason of the conflicting decisions of the various circuit courts.”132 This suggests, by Judge Posner’s own lights, the Act (and thus, § 1400(b)) shouldn’t be read broadly: the Act was passed “against a background of dissatisfaction with judicial handling of the same subject under a previous statute.”133

Second, even if we did apply full-bore imaginative reconstruction, we’d end up with even more reason to reject the human-agent requirement: rejecting the requirement fulfills the Act’s legislative intent better than imposing the requirement does. Granted, an important aim of the statute was to limit patent venue.134 But eschewing the human-agent requirement in favor of faithfully applying the rest of § 1400(b), as described in Part IV, achieves this aim in largely the same way as keeping the requirement (though not identically).135 And that alternative does a significantly better job of vindicating another of the statute’s key goals: reducing uncertainty about the scope of patent venue.136 Thus, even under a hardcore imaginative-reconstruction approach, the best route is to avoid reading a human-agent requirement into § 1400(b).

B. The Judicial Administrability Problem

The human-agent requirement introduces an entirely new layer of expense and uncertainty to the early stages of patent litigation—precisely the stages least in need of more expense and uncertainty. First recall, if TC Heartland’s policy virtue was its limitation of forum shopping, its policy vice was its switch from a venue rule (venue is proper wherever personal jurisdiction obtains) to a venue standard (venue is proper wherever the defendant has a regular and established place of business).137 True, standards may increase the odds of equitable venue (that is, having proper venue where it would be fair to litigants to have venue and not otherwise) in some cases.138 But in the patent venue context—where forum shopping, uncertainty,139 extortionate settlements, and slow decisions have consistently dogged courts and litigants—courts should avoid additional standards.140

After TC Heartland, lower courts have no choice but to apply the standard-like “regular and established place of business” test. The problem with the human-agent requirement, then, isn’t that the requirement turns a rule into a standard, but that it takes a standard-like test and needlessly adds a new, distinctly tricky standard to it. And critically, the inquiry does not turn on the same factual material that would have come out in discovery anyway.

The Rensselaer decision illustrates the point. To start, notice that “agency” is a common law standard rather than a rule. Agency is “the fiduciary relationship that arises when one person (a ‘principal’) manifests assent to another person (an ‘agent’) that the agent shall act on the principal’s behalf and subject to the principal’s control, and the agent manifests assent or otherwise consents so to act.”141 In Rensselaer, whether venue was proper ended up turning on whether the employees of a third-party contractor—employees tasked with configuring, maintaining, and repairing Amazon Lockers—were Amazon’s agents.142 To answer this question, the court considered the contract between Amazon and the contractor, the business procedures of both companies, and depositions of employees of both Amazon and the contractor.143 This inquiry did not turn on the same factual material that already would have come out in discovery. Indeed, the only reason the corporate relations between Amazon and its break/fix contractor were relevant was for purposes of the human-agent requirement.144 Thus, the requirement takes an already messy test and adds a unique set of costs.

After Google III, the tricky situation in cases like Rensselaer is just the start of the costs and complications. Consider the issue Judge Evan Wallach’s concurrence pointed out in Google III: Does a Google end user—that is, someone who uses Google’s services on a computer—count as an agent of Google?145 In Judge Wallach’s own words:

Given the absence from the record of information sufficient to understand Google’s business model, the question remains for the District Courts to determine whether Google’s end users become agents of Google in furtherance of its business by virtue of voluntarily or involuntarily sharing information generated on Google’s servers. If, for example, by entering searches and selecting results a Google consumer is continuously providing data which Google monetizes as the core aspect of its business model, it may be that under the analysis in which I today join, Google is indeed doing business at the computer of each of its users/customers. Because this is a question I believe should be entertained by District Courts, I concur.146

I take no stance on the correct answer to this question. I simply want to point out two things: First, the question is a consequential one that, if answered wrongly, could cause a return to overbroad venue and forum shopping. Second, and independently, answering it will involve legal and factual difficulties—difficulties that wouldn’t present themselves without the agency requirement.147

One possible outcome is that all district courts decide Google’s end users (that is, you and I) are Google’s agents and that (some of) our computers otherwise satisfy the “regular and established place of business” test.148 If this happens, Google is subject to patent infringement suit in every district in the nation. Given that one of the human-agent requirement’s stated benefits is to limit patent venue, this would be an exceedingly strange result.149 More likely, suppose most district courts decide end users aren’t Google’s agents. Even if this happens, any district that favors broad patent venue (like the Eastern District of Texas) is free to disagree, using Judge Wallach’s concurrence as a means of yet again haling Google into its courtrooms. The end result: forum shopping is back on the menu, at least until the Federal Circuit steps in with another writ of mandamus.

To see the kinds of practical difficulties the human-agent requirement will cause, consider how a district court might go about answering Judge Wallach’s question. Someone sues Google for patent infringement, arguing venue is proper because Google has a regular and established place of business at the computer of any given end user in the district. First, the parties will have to gather evidence about some particular user whose computer is the (alleged) place of business. This will take time and cost money. Next, the court must dive into the nitty-gritty of who counts as an agent. This isn’t so bad—agency is a well-developed area of the common law, after all. But then the court will have to do a fact-sensitive inquiry, asking whether this particular user counts as an agent—and if so, whether this particular user’s computer otherwise satisfies the “regular and established place of business” test.150 These are all thorny questions of first impression, bound to result in uncertainty and interdistrict disagreement. And if a given court finds no venue, that isn’t the end of the matter. It might just mean the plaintiff can come back to the same district with a different end user and try again.151 After all, all it takes is one regular and established place of business to satisfy the test.

That isn’t the worst of it. As Judge Wallach’s concurrence points out, this entire inquiry depends not only on the end user’s circumstances, but also upon the “core aspect[s]” of the defendant’s “business model.”152 Thus, courts will have to ask whether any given tech company’s business model (Zoom’s, Oracle’s, or perhaps Amazon’s, for example) involves end users in a way that renders those end users the company’s business-conducting agents. Every time a different tech company is sued under this end user theory, the district court will have to undertake another fact-intensive inquiry into that company’s business model. Not only will this take time and drive up discovery costs, it will invite still more disagreement and uncertainty at the district court level.

The Federal Circuit, like the district courts it sided with, may have seen the human-agent requirement as a useful way to limit patent venue. But it’s important to recognize that this route comes at a cost. The requirement will inject new fact-intensive, standards-heavy questions into the early stages of patent litigation. Discovery costs will increase, and district courts will have more and more opportunities to disagree and thus more chances to generate inefficient uncertainties. These downsides might be justified if the human-agent requirement were the only thing holding back the flood of unlimited venue. But as the next Part argues, that isn’t so. There is another, much easier way to vindicate Congress’s goal of keeping patent venue within reasonable bounds.

IV. The Alternative: “Place” as a Venue Limitation

The alternative I propose is simple. In fact, it’s boring—but in this context, that is a virtue and not a vice. The Federal Circuit (or the Supreme Court) should do away with the human-agent requirement and use § 1400(b)’s text to limit patent venue.153 An unadorned § 1400(b)—and especially § 1400(b)’s requirement of a “place”—can limit venue without any misinterpretation and without causing the aforementioned judicial administrability problems. A world where courts enforce a place requirement and no human-agent requirement, I argue, is preferable to a world where courts enforce both.

Before diving into the rest of this Part, I should clarify that I do not claim the place requirement should be the only limitation on patent venue. To the contrary, courts should faithfully and rigorously apply all of § 1400(b)’s requirements.154 Doing so will limit patent venue, and that’s a good thing. Still, this Part focuses on the place requirement for two reasons. First, courts—including the Federal Circuit in Google III—have given the requirement short shrift, so there is a need to point out its importance. Second, giving the place requirement more attention would produce the limiting effect the human-agent requirement is supposed to have. And while this limiting effect wouldn’t be identical to that of the human-agent requirement, I contend that the place requirement gets the answer right in the cases where the two diverge.

Part IV.A explains what a robust place requirement would look like, using Rensselaer as an example of “place” done well. Part IV.B shows that “place” can be a real limit on patent venue. Google III itself illustrates this point. Rightly decided, the Google III court would still have found no venue—and not because of agency, but because the place requirement wasn’t satisfied. Part IV.C explains why the “place” alternative is preferable to imposing a human-agent requirement.

A. A Robust Place Requirement

Section 1400(b), obviously, requires a “regular and established place of business” within the judicial district. The In re Cray court explicated the word “place” as a “physical place” requirement. The court framed this as the first of three prongs in the “regular and established place of business” test.155 The In re Cray court emphasized that the Eastern District of Texas, which heard the case below, “erred as a matter of law in holding that a fixed physical location in the district is not a prerequisite to proper venue.”156 Instead, the In re Cray court laid out a rule-plus-standard view of what counts as a place. The gatekeeping rule: “The statute [ ] cannot be read to refer merely to a virtual space or to electronic communications.”157 So courts are to consider physical space, and only physical space, when analyzing the place requirement. Next, the court explicated the place standard itself: the statute requires (vaguely enough) a “physical, geographical location in the district,” such as “a building or a part of a building set apart for any purpose or quarters of any kind from which business is conducted.”158 Thus, “place” involves a two-step analysis. First there is the gatekeeping rule that only physical, and not virtual, space, can be considered. Second, and more important, there is the standard: Is the physical presence substantial enough to be a place?159

I suggest the right way to handle § 1400(b) is to use the place requirement to limit venue. As discussed in the next Section, this would require the Federal Circuit to reverse course. Its recent Google III decision hamstrings the place requirement, making the place requirement too flimsy to limit venue effectively. If this precedent were out of the way, district courts could focus their inquiry on whether a defendant’s physical presence is substantial enough to count as a place. By paying attention to the statute’s text and giving place its teeth back, courts could gradually build a robust body of precedent about what counts as a place. Courts do this kind of thing all the time. I am intentionally not proposing anything too innovative, and I certainly don’t suggest courts should expand the notion of “place.” I don’t even propose turning the place standard into a rule (for example, by holding that a location is a place only if it has its own permanent address). That approach would be unintuitive, departing from the ordinary and imprecise meaning English speakers have always ascribed to the word. It would also be too rigid, inviting evasion of “place” status—and therefore venue—by defendants. Instead, with Google III out of the way, district courts could leave the place standard as a standard, building on their own cases to refine it over time. And though standards do, of course, have problems of their own (indeterminacy and a surfeit of judicial discretion come to mind), the place standard is a surprisingly workable one.

The court’s approach in Rensselaer is a prime example of a district court, pre–Google III, taking place seriously. In Rensselaer, the court determined that some Amazon Lockers counted as a place within § 1400(b)’s meaning.160 Though the court did find the place requirement satisfied here, it reached that result by treating this requirement seriously—and that’s the important point. First, the court laid out the standard as In re Cray articulated it.161 Then the court outlined the relevant facts. Roughly: the Lockers were large units; Amazon had secured licenses to put them up on another business’s property; Amazon didn’t allow that business to do anything with the Lockers without first notifying Amazon; and some of the Lockers were anchored to the ground with concrete bolts.162 Next, the court noted that the Lockers (obviously) were not “merely [ ] a virtual space.”163 The question therefore turned on whether the Lockers were physically substantial enough to count as an “established ‘physical, geographical location.’”164 So the court consulted the language of In re Cray and some decisions from fellow district courts, ultimately concluding that each Locker was big enough and permanent enough to count as a place.165 Finally, the court used differentiating facts to distinguish this fact pattern from that of cases with a contrary result.166 For example, the court pointed out that the Northern District of Ohio had suggested vending machines would not count as places.167 But, the court reasoned, Amazon Lockers differ in several ways from vending machines: they require more maintenance, they are larger, and they have their own addresses, to name a few.168

The Rensselaer court’s handling of § 1400(b)’s place requirement is important because of its process, which is a decidedly familiar one. In a world with a robust place requirement, courts could determine whether the place requirement is satisfied by doing a run-of-the-mill analysis of the facts, the statutory text, and the case law. And this is so even though “place” inevitably involves a standard and can’t readily be reduced to a pure rule. Granted, this standard involves some costs and some indeterminacies, as all standards do. But as I argue below, place can limit venue with decidedly fewer of both than the human-agent requirement involves.

B. The Place Requirement’s Ability to Limit Venue

The place requirement—if given serious attention and rigorous analysis—could be a robust limit on patent venue. Google III is a good example: in that case, place, rather than agency, could’ve done the venue-limiting work that needed doing. In other words, Google III reached the right conclusion (no venue) for the wrong reason. This Section sketches how, in my view, Google III should have been reasoned.

In Google III, the Federal Circuit analyzed the place requirement less thoroughly than did the district court in Rensselaer. The Google III court held that some Google servers on rented shelf space counted as a place under § 1400(b).169 The court spent most of its time establishing that property can count as the defendant’s place even if the defendant doesn’t have an ownership or leasehold interest in the property.170 But the court did not provide strong support for its holding that “leased shelf space or rack space can serve as a ‘place’ under the statute.”171 In support of this holding, it gave one hypothetical (“a defendant who operates a table at a flea market may have established a place of business . . .”).172 It also cited two district court decisions.173 The first decision, Tinnus Enterprises, LLC v Telebrands Corp,174 dealt with a much more substantial physical presence than a few computers on racks. The defendants had “holiday tables, As Seen on TV product sections, [and] product pallet displays” in retail stores within the district.175 So it’s hard to see how Tinnus Enterprises stands for a general proposition about shelves. The other decision was Peerless Network. The Peerless Network court gave the place requirement precisely one sentence’s worth of analysis.176

Perhaps the most striking thing about Google III is its rather laid-back analysis of the place requirement. The court did no Rensselaer-like inquiry into the shelves in question. There was no discussion of exactly how many servers Google had in the district; nor their size; nor of how many full racks (as opposed to shelves on racks), if any, were controlled by Google.177 Indeed, though Google did exercise control over the servers, the servers seem to have been rather transient after all.178 In a footnote, the court recognized that Google quietly removed them from the district after the start of the case.179 More important, Google III’s casually approving citation of Peerless Network means that even a computer “the size of a breadbox and the shelf on which it sits” is now enough to count as a “place” in any district in the nation.180 Simply put, that outcome is a strange one. A computer intuitively isn’t substantial enough to be a physical place, and it’s hard to see why putting that same computer on a shelf should make the difference.181

A better approach in Google III would have been to give more attention and rigor to the place requirement analysis. This would have yielded the same result (no venue), but without invoking the human-agent requirement. Perhaps surprisingly, the Eastern District of Texas once reached this very conclusion in a nearly identical Google-server-on-shelf case. The case was Personal Audio, LLC v Google, Inc,182 the holding of which SEVEN Networks later overruled.183 The Personal Audio court, much like the Rensselaer court, gave the place requirement teeth. First, the court properly ruled out electronic communications as constituting a “place.”184 The court then considered the servers at issue as purely physical entities. The court emphasized that the rooms hosting the servers were not the defendant’s property.185 Thus, the hardware itself was the only candidate for a “place.”186 The court then reached the intuitive conclusion: though “servers might be loosely referred to as a ‘place,’ or perhaps as the ‘location’ of ‘the cloud,’” they are simply not “a building or physical quarters of any kind.”187 That was that. As any layperson could tell you, a server is a thing, not a place.188

Here, however, a tricky issue arises. The technology industry, pesky as it is, has a tendency to constantly change. It doesn’t seem at all far-fetched to imagine that, in a few years, Amazon will have lockers or other structures that are physically substantial enough to satisfy the “place” standard in every judicial district in the nation. Or Google might build its own facilities full of servers all across the country, thus acquiring a physical “place” in any given district. What then? The human-agent requirement may very well find “no venue” in such situations.189 In contrast, the place standard would, in this hypothetical future, allow nationwide venue for any lawsuit against Amazon or Google. Forum shopping would be back.

While the place alternative certainly doesn’t limit venue along precisely the same lines as the human-agent requirement, this is a virtue rather than a vice. First, while nationwide venue could result for massive companies like Amazon and Google, this would happen only if these companies put down a substantial physical footprint in practically every judicial district. If companies want to expand in this way, nobody is stopping them. But by the same token, no tragedy has occurred if a tech behemoth or two—after voluntarily acquiring physical places of business across the nation—can be sued in any district across that same nation.190 This is simply the cost of “bestrid[ing] the narrow world [l]ike a Colossus.”191 In fact, the contrary result, which might follow under the human-agent requirement, would be puzzling. Why should the hypothetical Google of the future, which owns a building full of servers in a judicial district, be immune from suit there just because Google doesn’t have any agents at the facility? That building is a place—it is a tie between the company and the locale. It seems perfectly natural to let such a tie be the basis of venue.

Further, there are plenty of other companies, smaller than Google and Amazon, that have their own widespread servers or routers without their own “places” for them. These are precisely the kinds of companies that the “place” requirement would protect from nationwide suit, assuming these companies have no plans to acquire physical “places” across the nation. This might explain why the amici tech companies, in the course of urging the Federal Circuit to reject the SEVEN Networks decision, argued solely on place-requirement lines.192 Thus, though a rigorous place requirement doesn’t limit venue along exactly the same lines as the human-agent requirement, it allows for more venue in the right cases (cases where a company has set down a significant physical footprint tying it to a locality, even if it doesn’t have human beings working there) and disallows venue in the others (cases where a company has no such tie to a district).193

The place requirement, if treated seriously, can be a significant limit on patent venue. The right approach is twofold: First, the human-agent requirement has got to go. Second, the Federal Circuit should tighten up the place requirement. Doing so would not only improve the contours of patent venue, it would also, as I am about to argue, do so at a lower cost for courts and litigants alike.

C. The Place Requirement’s Advantages

The place requirement limits patent venue with none of the human-agent requirement’s downsides. The place requirement comports much more closely with § 1400(b)’s text.194 And the place requirement is more judicially administrable than the human-agent requirement.195

That second point is less obvious and therefore deserves some explanation. Recall the judicial administrability problem with the human-agent requirement: it is an unwieldy standard, and it adds too many factual and legal uncertainties to the early stages of litigation. But the place requirement is a standard, too (or at least, it involves a standard). So, isn’t it just as bad? Well, no. The critical problem with the human-agent requirement, as discussed in Part III.B, is not that it turns a rule-like test into a standard-like test. The problem, instead, is that it adds another standard to an already-too-mushy test.

The place requirement is, admittedly, standard-like. That’s not ideal for the early stages of litigation. But the human-agent requirement doesn’t replace the place requirement; it just adds to it.196 Even under Google III’s too-relaxed approach to place, district courts will still have to draw lines somewhere: Does a server, even without rented shelf space, count as a “place”? If so, what about a laptop? If not, what about five servers stacked on top of each other? And so on. Therefore, the relevant comparison isn’t between a world where courts use the human-agent requirement and a world where they use the place requirement. Rather, the comparison is between a world where courts use both the human-agent and place requirements, and a world where they use only the place requirement. Framed this way, the answer is clear: using only the place requirement is more judicially administrable than using both. This isn’t a close call, either: the human-agent requirement’s unique facts, discovery costs, and legal uncertainties are no small burden for courts and litigants to bear.

Conclusion

If the human-agent requirement truly were entailed by the text of § 1400(b) and § 1694, there would be no option but to keep it. But as we’ve seen, that simply isn’t the case. If this isn’t enough reason to reconsider the requirement (and in my view, it certainly is), it might be helpful to look to the requirement’s costs and benefits. First, will imposing the requirement cause any trouble? This Comment suggests it will cause plenty. Not only is it wrong as a matter of statutory interpretation, but it also makes patent venue litigation needlessly cumbersome.

The question, then, is does the requirement provide enough benefits to justify those hefty costs? Well, the requirement does help limit patent venue. This is a worthy goal, but § 1400(b)’s place requirement can achieve that goal in largely (though not precisely) the same way the human-agent requirement would. And the place requirement has none of the downsides. Instead of being based on a statutory misinterpretation, it comes from a straightforward reading of the statute. And instead of increasing the duration and expense of patent litigation’s early stages, “place” limits venue without incurring any additional cost. The bottom line: eschewing the human-agent requirement in favor of § 1400(b)’s unadorned text would be uninteresting, unoriginal, and downright uninspired. But it would also be interpretively correct and an improvement to the legal system.

• 1The Federal Circuit is the only federal appeals court with jurisdiction to hear patent appeals. 28 USC § 1295(a)(4).
• 2See VE Holding Corp v Johnson Gas Appliance Co, 917 F2d 1574, 1575 (Fed Cir 1990).
• 3Patent trolls are businesses that sit on patents and make their money by suing, or threatening to sue, for infringement rather than actually innovating. For an explanation of the patent-troll business model along with a real-life example, see Nick Statt, Apple Ordered to Pay Patent Troll More Than $500 Million in iMessage Case (The Verge, Apr 10, 2018), archived at https://perma.cc/S2ZW-UN2F.
• 4See Part I.B.
• 5137 S Ct 1514 (2017).
• 6Id at 1516–17.
• 7See In re Cray Inc, 871 F3d 1355, 1360, 1362–64 (Fed Cir 2017).
• 8See, for example, In re Google LLC, 2018 WL 5536478, *2–3 (Fed Cir) (Google I) (puzzling over whether the locations of Google’s servers constituted “regular and established place[s] of business”).
• 9As discussed in Part II.B, agency is “the fiduciary relationship that arises when one person (a ‘principal’) manifests assent to another person (an ‘agent’) that the agent shall act on the principal’s behalf and subject to the principal’s control, and the agent manifests assent or otherwise consents so to act.” Rensselaer Polytechnic Institute v Amazon.com, Inc, 2019 WL 3755446, *11–12 (NDNY), quoting Restatement (Third) of Agency § 1.01 (2006).
• 10See, for example, Peerless Network, Inc v Blitz Telecom Consulting, LLC, 2018 WL 1478047, *1 (SDNY) (router); SEVEN Networks, LLC v Google LLC, 315 F Supp 3d 933, 949 (ED Tex 2018) (servers).
• 11See, for example, In re Google LLC, 914 F3d 1377, 1380 (Fed Cir 2019) (Google II) (Reyna dissenting from denial of rehearing en banc), citing Personal Audio, LLC v Google, Inc, 280 F Supp 3d 922, 934 (ED Tex 2017).
• 12See, for example, Rensselaer, 2019 WL 3755446 at *2.
• 13I am using the term “courts” loosely here. Really, five district court opinions have addressed this issue: two in the Southern District of New York; two in the Eastern District of Texas; and one in the Northern District of New York. I recognize that these opinions are not binding for future decisions, even in the same district. See Part II.
• 14See Rensselaer, 2019 WL 3755446 at *11–12; CDx Diagnostic, Inc v United States Endoscopy Group, Inc, 2018 WL 2388534, *3 (SDNY); Peerless Network, 2018 WL 1478047 at *4.
• 15See Super Interconnect Technologies LLC v Google LLC, 2019 WL 3717683, *1 (ED Tex), revd, In re Google LLC, 949 F3d 1338 (Fed Cir 2020) (Google III); SEVEN Networks, 315 F Supp at 962.
• 16See Google I, 2018 WL 5536478 at *1–3.
• 17In re Google LLC, 949 F3d 1338, 1344–45 (Fed Cir 2020) (Google III). The court ended its opinion on a cryptic note, suggesting that somewhere down the line a machine might be regarded an “agent” in the relevant sense: “To be clear, we do not hold today that a ‘regular and established place of business’ will always require the regular presence of a human agent, that is, whether a machine could be an ‘agent.’ Such a theory would require recognition that service could be made on a machine.” Id at 1347. Aside from this qualification (upon which the case didn’t turn in any way), the court adopted the same human-agent requirement the lower courts had been considering. For that reason, and because this Comment’s arguments respond to either version of the requirement, I’m sticking with the “human-agent requirement” nomenclature.
• 18Ch 395, 29 Stat 695.
• 19Compare Act of Mar 3, 1875, ch 137, 18 Stat 470 (allowing suit wherever the defendant may be found and served process), with Act of Mar 3, 1897, ch 395, 29 Stat at 695–96 (stating that suit may be brought in the district the defendant inhabits, or the district where the defendant has a regular and established place of business and has committed acts of infringement). See also In re Cray, Inc, 871 F3d 1355, 1360–61 (Fed Cir 2017) (contrasting the two acts and discussing the purpose of the latter, which included resolving some interpretive questions courts had been disagreeing about). More precisely, only the patent venue sections of the Act had these two aims.
• 20Though the original statute used the word “jurisdiction,” the Court has long considered this to be a venue statute. See Stonite Products Co v Melvin Lloyd Co, 315 US 561, 562–63 (1942) (treating the original act as a venue statute); TC Heartland LLC, 137 S Ct at 1518 (discussing Stonite in those same terms). This Comment follows suit.
• 21Act of Mar 3, 1897, ch 395, 29 Stat at 695–96 (emphasis added).
• 22See Act of June 25, 1948, 62 Stat 869, 936, codified at 28 USC § 1400; 62 Stat 869, 945, codified at 28 USC § 1694.
• 2328 USC § 1400(b) (emphasis added).
• 2428 USC § 1694.
• 25Stonite, 315 US at 566. In support of this proposition, the Stonite Court approvingly cited several cases, including most notably Bowers v Atlantic, G&P Co, 104 F 887, 888–93 (CC SDNY 1900), which gives a detailed justification (on the bases of legislative history, legislative intent, and textual interpretation) of the view that the Act should be interpreted as a venue-limiting, rather than venue-expanding, provision.
• 26Schnell v Peter Eckrich & Sons, Inc, 365 US 260, 262 (1961), citing Stonite, 315 US 561.
• 2729 Cong Rec 1900, 1900 (1897) (statement of Rep Mitchell).
• 28Schnell, 365 US at 264, quoting Olberding v Illinois Central Railroad Co, 346 US 338, 340 (1953).
• 29917 F2d 1574 (Fed Cir 1990).
• 30See Fourco Glass Co v Transmirra Products Corp, 353 US 222, 226 (1957).
• 31VE Holding Corp, 917 F2d at 1583. Venue and personal jurisdiction are, of course, not the same thing. See, for example, In re Cray, 871 F3d at 1361 (emphasizing the difference in the patent venue context). For present purposes, it’s enough that, because personal jurisdiction is its own independent requirement, making venue as broad as personal jurisdiction is the same thing as eradicating venue’s limiting function.
• 32It is literally true that, under VE Holding, § 1400(b)’s residence prong provided no limit on patent venue whatsoever. In 1988, Congress broadened the definition of the word “reside” in 28 USC § 1391—the statute governing venue in run-of-the-mill federal suits—to mean that a corporate defendant resides in every judicial district in which it is subject to personal jurisdiction. Act of Nov 19, 1988, Pub L No 100-702, 102 Stat 4642, 4669, codified at 28 USC § 1391(c)(2). VE Holding held that this congressional alteration of “reside” in the general venue statute also modified the meaning of “resides” in the patent venue statute—§ 1400(b). VE Holding, 917 F2d at 1575. Critically, this meant that wherever there was personal jurisdiction (which is, of course, already necessary for a lawsuit), patent venue was automatically proper. Patent venue and personal jurisdiction, under VE Holding, were coterminous.
• 33This prong’s irrelevance under VE Holding contrasts sharply with its past prominence. For an example of the latter, see New Wrinkle, Inc v Fritz, 30 F Supp 89, 90 (WDNY 1939) (asking whether a corporate defendant had a regular and established place of business in the district).
• 34See J. Jonas Anderson, Reining in a “Renegade” Court: TC Heartland and the Eastern District of Texas, 39 Cardozo L Rev 1569, 1574 (2018) (emphasizing that the Eastern District has become the “leading court for patent litigation”); Hilda Galvan, Chad Everingham, Clyde Siebman, George Bramblett, and Xuan-Thao Nguyen, The America Invents Act: A Tribute to the Honorable John Ward, 15 SMU Science and Tech L Rev 459, 459 (2012) (describing the retired judge of the Eastern District of Texas as “one person who [ ] change[d] the world of patent law and also put Texas on the map”). But see Barrie McKenna, Venue Shopping? See You in Marshall (The Globe and Mail, May 4, 2006), archived at https://perma.cc/C7YW-LV6A (describing part of an interview with Judge T. John Ward, including his description of the patent-plaintiff rush as an “unforeseen consequence” rather than an intended aim).
• 35Owen Byrd, Announcing the Patent Litigation Year in Review 2015 (Lex Machina, Mar 16, 2016), archived at https://perma.cc/G75R-H966.
• 36See Anderson, 39 Cardozo L Rev at 1576 (cited in note 34) (collecting sources that propose a variety of explanations, including “East Texas’s notoriously friendly juries, or judges who are ‘[k]nowledgeable, [w]elcoming, and [o]rganized,’ or plaintiff-friendly procedural rules, unwillingness to transfer cases to a more convenient district court, differences in substantive law rulings, or a host of other reasons,” and concluding that “what is not up for debate is that the district benefitted [from]” VE Holding) (citations omitted).
• 37See Anderson, 39 Cardozo L Rev at 1578–79 (cited in note 34).
• 38See, for example, Loren Steffy, Patently Unfair (Texas Monthly, Sept 15, 2014), archived at https://perma.cc/8ZC5-37PM (stating that “the small East Texas town of Marshall may be the worst thing that ever happened to intellectual property law”); ATR Foundation, 2015/2016 Executive Summary: Judicial Hellholes, archived at https://perma.cc/8PYG-T6N7 (ranking the Eastern District of Texas the ninth worst judicial hellhole in the nation, and complaining that “[m]ost of the lawsuits are filed by patent trolls”); Jeanne C. Fromer, Patentography, 85 NYU L Rev 1444, 1445 (2010) (explaining that “[f]orum shopping can be harmful to the legal system by distorting the substantive law, by showcasing the inequities of granting plaintiffs an often outcome-determinative choice among many district courts, and by causing numerous economic inefficiencies, including inconveniences to the parties”); Rochelle Cooper Dreyfuss, The Federal Circuit: A Case Study in Specialized Courts, 64 NYU L Rev 1, 6–7 (1989) (discussing the problems with pre–Federal Circuit forum shopping in the patent context, and pointing out the inconsistency it caused, which in turn decreased the “incentive to invest in research and development”); Melissa Wee, TC Heartland v. Kraft and the Resurrection of the Place of Incorporation or “Regular and Established Place of Business” Test for Patent Venue, 33 Berkeley Tech L J 981, 1003–06 (2018) (arguing that, though forum shopping has its upsides, pre–TC Heartland forum shopping probably reached and exceeded the “point when the benefit of court expertise is outweighed by the inefficiencies of too high a concentration of patent cases”). But see, for example, Friedrich K. Juenger, What’s Wrong with Forum Shopping?, 16 Sydney L Rev 5, 5–6 (1994) (arguing that there is nothing wrong with forum shopping). But consider Timothy B. Dyk, Ten Prescriptions for What Ails Patent Law, 17 Stan Tech L Rev 345, 352 (2014) (praising the Eastern District of Texas (without praising forum shopping per se) for limiting discovery and the length of trials in patent lawsuits). At the time of writing, Timothy Dyk is a judge on the Federal Circuit.
• 39See Part I.A.
• 40See Commission on Revision of the Federal Court Appellate System, Structure and Internal Procedures: Recommendations for Change, 67 FRD 195, 220 (1975) (discussing and criticizing the practice of forum shopping in patent infringement suits). See also S Rep No 97-275, 97th Cong, 1st Sess 5 (1981), reprinted in 1982 USCCAN 11, 15 (“The creation of the Court of Appeals for the Federal Circuit will produce desirable uniformity. . . . Such uniformity will reduce the forum-shopping that is common to patent litigation.”).
• 41See Anderson, 39 Cardozo L Rev at 1581–86 (cited in note 34) (surveying failed congressional attempts to pass legislation curbing patent forum shopping).
• 42Id at 1586–90 (overviewing the Federal Circuit’s attempts to rein in the Eastern District, but noting that Federal Circuit judges have made pro–Eastern District comments and pointing out their general policy “to leave patent policy decisions to Congress”).
• 43TC Heartland, 137 S Ct at 1517. The Court focused on formal methods of statutory interpretation, emphasizing that “residence,” as used in § 1400(b), had long borne a “particular meaning as applied to domestic corporations.” This meaning, the Court explained, had not changed merely because Congress had amended a different statute (§ 1391). Id at 1520.
• 44Howard I. Shin and Christopher T. Stidvent, Patent Venue a Year After TC Heartland, 11 Landslide 42, 45 (2018).
• 45Id.
• 46For mention of the not-insignificant policy downsides of TC Heartland, and an argument that the human-agent requirement exacerbates these downsides with little benefit, see Part III.B.
• 47Anderson, 39 Cardozo L Rev at 1620 (cited in note 34).
• 48871 F3d 1355 (Fed Cir 2017).
• 49Id at 1360.
• 50For discussion, see In re Google LLC, 2018 WL 5536478, *5–6 (Fed Cir) (Google I) (Reyna dissenting) (noting the significant disagreement among district courts, and even within districts, regarding the application of the test).
• 51In re Cray, 871 F3d at 1360–61. The circuit followed suit on this point when it later endorsed the human-agent requirement. In re Google LLC, 949 F3d 1338 1346–47 (Fed Cir 2020) (Google III).
• 52In re Cray, 871 F3d at 1360–61.
• 53See Part I.A.
• 54A third, somewhat more specific, question is: Exactly how limited should patent venue be? Part IV.B addresses this question, contending that the human-agent requirement limits venue a bit too much.
• 552018 WL 1478047 (SDNY).
• 56See In re Google LLC, 914 F3d 1377, 1380 (Fed Cir 2019) (Google II) (Reyna dissenting from denial of rehearing en banc) (describing Peerless Network as a case about routers), citing Peerless Network, 2018 WL 178047 at *3–4.
• 57See Patrick McKeown, Information Technology and the Networked Economy 476 (Global Text 2009) (defining a “router” as “[a] computer that determines the path that a message will take from the sending computer to the receiving computer”).
• 58Peerless Network, 2018 WL 1478047 at *4 (emphases in original).
• 592019 WL 3755446 (NDNY).
• 60Amazon, Amazon Hub Locker, archived at https://perma.cc/PJD4-U5ZC. Customers can also return packages to Lockers. For a more detailed description of these kinds of Lockers, see text accompanying notes 161–68.
• 61Rensselaer, 2019 WL 3755446 at *11–13.
• 62The Peerless Network court also briefly mentioned two other sources of support. First, the court pointed to In re Cray’s statement that “the defendant must actually engage in business from [the] location.” Peerless Network, 2018 WL 1478047 at *4 (quotation marks omitted), quoting In re Cray, 871 F3d at 1364. But basing a human-agent requirement on this sentence alone—which was merely meant to clarify that an unused office space does not count as a regular and established place of business—would be a stretch. Second, the court pointed to a dictionary definition, which defines “‘place of business’ as ‘[a] location at which one carries on a business.’” Peerless Network, 2018 WL 1478047 at *4 (alteration in original), quoting Black’s Law Dictionary (West 10th ed 2014). This argument begs the question as to what it means to “carry on a business.” If someone does business remotely, is the “location at which she carries on her business” her location, the location at which the business is happening, or both? Only if we assume the human-agency reading is correct from the outset does the dictionary definition tell us anything about this question.
• 63Peerless Network, 2018 WL 1478047 at *4. In Personal Audio, LLC v Google, Inc, 280 F Supp 3d 922 (ED Tex 2017), notably, the Eastern District of Texas (of all places) hinted at a human-agent requirement. Id at 934. But for two reasons, I am not devoting much time to that hint. First, the court’s discussion of agency was mixed with its discussion of “place,” and the court never clearly adopted a human-agent requirement. Second, and more pointedly, the Eastern District of Texas has since rejected the human-agent requirement. See SEVEN Networks, LLC v Google LLC, 315 F Supp 3d 933, 961–62 (ED Tex 2018).
• 64Rensselaer, 2019 WL 3755446 at *11.
• 65Id (emphasis in original) (quotation marks omitted), quoting Act of Mar 3, 1897, 29 Stat at 696.
• 66Rensselaer, 2019 WL 3755446 at *11 (emphasis in original) (quotation marks omitted), quoting 28 USC § 1694.
• 67Rensselaer, 2019 WL 3755446 at *11 (quotation marks omitted).
• 68See id. The court did not state this conclusion explicitly, but it is of course implicit.
• 69Properly speaking, this isn’t a “syllogism” as formulated; it’s merely a two-premise semi-formal argument. But I am using the term “syllogism” loosely here because this is an easy way to distinguish this argument from other arguments throughout this Comment.
• 70The Peerless Network court said the Act “presume[d] that a defendant with a ‘place of business’ in a district will also have ‘agents conducting such business’ in that district”; the court did not specify whether its argument was based on legislative intent or original public meaning. Peerless Network, 2018 WL 1478047 at *4.
• 71Or, if this is interpreted as an originalist rather than a legislative-intent argument, the original public. See notes 105–07 and accompanying text.
• 72In re Cray, 871 F3d at 1361; Rensselaer, 2019 WL 3755446 at *8; Peerless Network, 2018 WL 1478047 at *2 (discussing the broad concern that § 1400(b) should not be read too broadly, along the lines of the discussion in In re Cray).
• 73See Peerless Network, 2018 WL 1478047 at *1, 3.
• 74Google II, 914 F3d at 1381 (Reyna dissenting from denial of rehearing en banc). See also id at 1380 (noting that a gaggle of technology companies had joined Google’s mandamus petition as amici). To be fair to Judge Jimmie Reyna, he was referring to the SEVEN Networks holding as a whole—not just to its rejection of the human-agent requirement. Still, the worry remains.
• 75In re Google LLC, 949 F3d 1338, 1343–44 (Fed Cir 2020) (Google III).
• 76315 F Supp 3d 933 (ED Tex 2018).
• 77A server, roughly speaking, is a computer whose primary purpose is to send information to other computers. See McKeown, Information Technology at 477 (cited in note 57). Servers, like routers, can and often do operate without the presence of a human being.
• 78SEVEN Networks, 315 F Supp at 949.
• 79Id at 964.
• 80Id at 961–62.
• 81Id. This is imprecise. The SEVEN Networks court touched on one other argument, but its force is limited by a significant shortcoming. The court pointed out In re Cray’s mandate that the § 1400(b) “analysis must be closely tied to the language of the statute.” Id at 962 (quotation marks omitted), quoting In re Cray, 871 F3d at 1362. Because the human-agency reading, according to the court, “finds no basis within the language of the statute,” it must be rejected. Id. But this argument, without more, begs the question: What does the language of the statute mean? So saying, as the SEVEN Networks court seems to have done, that “other courts’ interpretation is wrong because it’s not closely tied to the language of the statute” amounts to saying “other courts’ interpretation is wrong because that interpretation is out of line with the statute really means.” But that amounts to nothing more than the claim that “other courts’ interpretation is wrong because it’s wrong.” For this “it’s not in the text” argument to work, we need more—specifically, we need a reason for rejecting the Peerless Network court’s service of process argument. The SEVEN Networks court gives no such reason. This Comment does. See Part III.A.2.
• 82SEVEN Networks, 315 F Supp 3d at 962.
• 83Id at 958–59 nn 35–37. See also, for example, State v Hutton, 39 Mo App 410, 416 (1890) (“This act, as amended by the act of March 24, 1887, recites: ‘No such license shall authorize any merchant to sell vinous, fermented or spirituous liquors in any quantities, to be drank at his store, stand or warehouse, or other place of business.’”), quoting 1887 Mo Laws 217.
• 84SEVEN Networks, 315 F Supp 3d at 962–64.
• 85See, for example, Bruesewitz v Wyeth LLC, 562 US 223, 236–37 (2011) (describing the canon and explaining why it did not apply to the case at hand).
• 86Pub L No 112-29, 125 Stat 284 (2011), codified in various sections of Title 35.
• 87AIA § 18(c), 125 Stat at 331.
• 88SEVEN Networks, 315 F Supp 3d at 964.
• 89Id.
• 90Rensselaer, 2019 WL 3755446 at *12.
• 91Id.
• 92Id.
• 93The only slight (and for present purposes irrelevant) caveat here, as discussed in note 17, is the court’s unexplained suggestion that perhaps—someday in the future—a machine could count as an “agent” for purposes of the statute. See Google III, 949 F3d at 1344–45.
• 94949 F3d 1338 (Fed Cir 2020).
• 95Id at 1339–41.
• 96Id at 1343–44.
• 97The court reasoned as follows: “[W]e normally presume that the same language in related statutes carries a consistent meaning,” and § 1694 “plainly assumes that the defendant will have a regular and established place of business . . . only if the defendant also has an agent . . . engaged in conducting such business.” Id at 1344 (quotation marks omitted). Thus, “[i]n the absence of a contrary indication, these assumptions must govern the venue statute as well.” Id. The court also made a legislative-history point, which I address subsequently. See notes 109–12 and accompanying text.
• 98The latter possibility may seem improbable, but TC Heartland does suggest the Court has at least some appetite for overturning the Federal Circuit on interpretive grounds. See Part I.C.
• 99Maybe this is what the SEVEN Networks court was getting at when it said the agency reading should be rejected because it is not explicitly a part of § 1400(b)’s text. See SEVEN Networks, 315 F Supp 3d at 961–62. But, as note 81 explains, the SEVEN Networks court’s argument doesn’t get off the ground because the court offers no compelling reason for rejecting the service of process argument first advanced by the Peerless Network court. This Section aims to offer such a reason.
• 100It could also be an imaginative-reconstruction argument. Because the hypothetical I am about to give wouldn’t be very helpful in responding to that version of the argument, I deal with imaginative reconstruction later, in Part III.A.3.
• 101See Peerless Network, 2018 WL 1478047 at *4; Google III, 949 F3d at 1344. The Google III court did, however, reference legislative history as a buttressing point. See Google III, 949 F3d at 1345. See also notes 109–13 and accompanying text.
• 102Rensselaer, 2019 WL 3755446 at *11 (“The use of the definite article and possessive determiner suggests that Congress assumed that a defendant with a place of business in a district will also have agents conducting such business in the district.”) (quotation marks omitted).
• 103To put it another way, there is a difference between a lawmaker/member of the public assuming that “as far as I can see, legal requirement A will be satisfied only if state of affairs B obtains,” and a lawmaker/member of the public mandating that “legal requirement A implicitly contains sub-requirement A*, which is not satisfied unless state of affairs B obtains.”
• 104Suppose boots had been invented by this time.
• 105See Antonin Scalia, A Matter of Interpretation: Federal Courts and the Law 38 (Princeton 2018) (Amy Gutmann, ed.) (“What I look for in the Constitution is precisely what I look for in a statute: the original meaning of the text, not what the original draftsmen intended.”). Original-public-meaning-originalism, of course, is just one form of originalism. But not much turns on which kind of originalism we use here.
• 106To take that stance would be to veer into “expectations originalism,” which, as discussed below, would not be a good idea. See notes 115–24 and accompanying text.
• 107See Scalia, A Matter of Interpretation at 22 (cited in note 105) (declaring, in the context of statutory interpretation, that “[t]he text is the law, and it is the text that must be observed”).
• 108Premise (2), restated for convenience: “Any assumption Congress (or the original public) makes about the state of the world at the moment of a statutory requirement’s satisfaction becomes a necessary condition for that satisfaction.”
• 109Google III, 949 F3d at 1345 (emphasis and citations omitted) (alteration in original), quoting 29 Cong Rec 1900, 1900 (1897) (statement of Rep Lacey).
• 110See 29 Cong Rec 1900, 1900–02 (1897).
• 111See Infringement of Letters Patent, HR Rep No 2905, 54th Cong, 2d Sess 1 (1897).
• 112See 29 Cong Rec 2719, 2719 (1897).
• 11328 USC § 1694.
• 114Rensselaer, 2019 WL 3755446 at *11 (emphasis added). Specifically, FRCP 4(h) provides:

  Unless federal law provides otherwise . . . , a domestic or foreign corporation . . . must be served: (1) . . . (A) in the manner prescribed by Rule 4(e)(1) for serving an individual; or (B) by delivering a copy of the summons and of the complaint to an officer, a managing or general agent, or any other agent authorized by appointment or by law to receive service of process and—if the agent is one authorized by statute and the statute so requires—by also mailing a copy of each to the defendant.

  FRCP 4(h).

  Thus, because § 1694 is a federal law that “provides otherwise,” plaintiffs in patent infringement suits can serve process in one of three ways: first, they can follow FRCP 4(h)(1)(A); second, they can follow FRCP 4(h)(1)(B); third, they can follow § 1694 and serve process on an agent doing business at the regular and established place of business, if there is such an agent. In fact, § 1694 might be read as fitting into 4(h)(1)(B): § 1694 just authorizes a new type of agent “by law to receive service of process.” FRCP 4(h).

• 115See, for example, Raoul Berger, Death Penalties and Hugo Bedau: A Crusading Philosopher Goes Overboard, 45 Ohio St L J 863, 872–73 (1984) (arguing, along quite similar lines, that the meaning of a constitutional provision is determined, not by the semantic content of its words, but by the situations to which it originally did or did not apply); Jed Rubenfeld, Freedom and Time: A Theory of Constitutional Self-Government 179 (Yale 2001) (harshly criticizing any form of originalism that doesn’t treat expected applications as binding). Professor Jed Rubenfeld has since softened his views on this topic. See Jed Rubenfeld, Revolution by Judiciary: The Structure of American Constitutional Law 125–27 (Harvard 2005). For discussion and critique of these views, as well as an alternative view, see generally Christopher R. Green, Originalism and the Sense-Reference Distinction, 50 SLU L J 555 (2006). This is largely a debate within originalism. But that’s not the point here: the argument for the human-agent requirement focuses on the Act’s legislative intent and/or original public meaning—and the legislative (read: Constitutional Convention’s) intent and original public meaning of the Constitution are precisely what originalists focus on anyway.
• 116For a helpful description of both this view, commonly called “expectations originalism,” and the contrary view, commonly called “semantic originalism,” see Randy E. Barnett, An Originalism for Nonoriginalists, 45 Loyola L Rev 611, 620–29 (1999).
• 117Antonin Scalia, Common-Law Courts in a Civil-Law System: The Role of United States Federal Courts in Interpreting the Constitution and Laws, in Grethe B. Peterson, ed, 18 The Tanner Lectures on Human Values 79, 119 (Utah 1997) (originally given in 1995).
• 118See id; US Const Amend VIII.
• 119The difference between constitutions and statutes isn’t relevant for present purposes. That is what allows the objector to leverage Justice Scalia’s view in this way.
• 120See Ronald Dworkin, Comment, in Scalia, A Matter of Interpretation 115, 120–22 (cited in note 105).
• 121Id at 119.
• 122Id.
• 123See, for example, Keith E. Whittington, Originalism: A Critical Introduction, 82 Fordham L Rev 375, 383 n 49 (2013) (observing that Dworkin’s “basic caution has been widely accepted by originalists”). See also Barnett, 45 Loyola L Rev at 620–29 (cited in note 116) (characterizing the rejection of expectations originalism as an important part of the “New Originalism” and as a helpful way for this “New Originalism” to avoid some of the most prominent critiques of originalism).
• 124See Richard A. Posner, Statutory Interpretation—in the Classroom and in the Courtroom, 50 U Chi L Rev 800, 817–22 (1983).
• 125Id at 817. See also William N. Eskridge Jr, Dynamic Statutory Interpretation, 135 U Pa L Rev 1479, 1479–84 (1987) (laying out “dynamic statutory interpretation,” which focuses on “statutory text,” “original legislative expectations,” and “the subsequent evolution of the statute and its present context,” but emphasizing the third factor “when the statutory text is not clear and the original legislative expectations have been overtaken by subsequent changes in society and law”). But see Frank H. Easterbrook, What Does Legislative History Tell Us?, 66 Chi Kent L Rev 441, 445–47 (1990) (criticizing this view).
• 126Posner, 50 U Chi L Rev at 818 (cited in note 124).
• 127Id.
• 128Id.
• 129Id.
• 130Posner, 50 U Chi L Rev at 818 (cited in note 124). For a real-life example of Judge Posner applying this methodology, as well as some more nuances I haven’t described here, see Hively v Ivy Tech Community College of Indiana, 853 F3d 339, 353 (7th Cir 2017) (Posner concurring):

  Title VII of the Civil Rights Act of 1964, now more than half a century old, invites an interpretation that will update it to the present, a present that differs markedly from the era in which the Act was enacted. But I need to emphasize that this [ ] form of interpretation—call it judicial interpretive updating—presupposes a lengthy interval between enactment and (re)interpretation. A statute when passed has an understood meaning; it takes years, often many years, for a shift in the political and cultural environment to change the understanding of the statute.

• 13129 Cong Rec 1900, 1901 (1897) (statement of Rep Mitchell) (describing the circuit split in detail and explaining how the then-bill would resolve it).
• 132Id at 1900 (emphasis added).
• 133Posner, 50 U Chi L Rev at 818 (cited in note 124). See also text accompanying note 28.
• 134See Part I.A.
• 135See Part IV.B (arguing that though the physical “place” requirement doesn’t limit venue along precisely the same contours as the human-agent requirement, the unadorned “place” requirement is the test outputting the right result in cases where the two diverge). More specifically, see note 193.
• 136See Part III.B (explaining the uncertainty the human-agent requirement brings along with it); Part IV.C (explaining why the place requirement would generate less uncertainty than the human-agent requirement).
• 137See Paul R. Gugliuzza, Quick Decisions in Patent Cases, 106 Georgetown L J 619, 679–80 (2018) (criticizing TC Heartland on the grounds that “the decision will likely result in protracted pre-merits litigation and increase litigation costs”).
• 138See Peter Estall, Note, Venerunt, Viderunt, Vicerunt Venue: How TC Heartland and In re Cray Have Conquered Patent Venue for Corporate Defendants and How Congress Can Balance the Scales of Patent Venue Justice, 103 Minn L Rev 1523, 1559–63 (2019) (proposing that Congress adopt a six-part “multifactor equity test” for patent venue and arguing this would be fairer to litigants).
• 139Uncertainty results both from fact-intensive inquiries that could go either way in any given case and from the possibility that judges will disagree about how to implement standards.
• 140See Gugliuzza, 106 Georgetown L J at 679–80 (cited in note 137) (bemoaning the uncertainty and cost of the “regular and established place of business” test); Wee, 33 Berkeley Tech L J at 1006–08 (cited in note 38) (arguing that standards, because of the cost, uncertainty, and manipulability they introduce, are a poor fit for patent venue). For an early critique of the “regular and established place of business” test as too standard-like, see Richard C. Wydick, Venue in Actions for Patent Infringement, 25 Stan L Rev 551, 574 (1973) (“Since the shards of evidence can be arrayed in an endless variety of patterns, the law in this area tends to grow in bulk but not in substance.”). But see Estall, 103 Minn L Rev at 1559–63 (cited in note 138) (arguing for more patent venue standards in the name of equity).
• 141Rensselaer, 2019 WL 3755446 at *12, quoting Restatement (Third) of Agency § 1.01 (2006).
• 142Rensselaer, 2019 WL 3755446 at *12.
• 143Id at *4–5.
• 144See id. See also id at *12. For discussion of one problem with a multifactor venue test, see Wee, 33 Berkeley Tech L J at 1006–08 (cited in note 38) (“[H]aving no single factor dispositive of venue could lead to discovery on any one of the factors, as opposed to being able to stop the inquiry once it is clear that a single requirement cannot be met.”). Granted, after Google III, maintenance no longer counts as doing business, so this precise issue couldn’t arise in the future. Google III, 949 F3d at 1346 (“The better reading of the statute is that [ ] maintenance activities cannot, standing alone, be considered the conduct of Google’s business.”). But Rensselaer is still illustrative: not all contractors, of course, merely do maintenance. Some contractors really do engage in business on behalf of someone else. Therefore, even after Google III, courts will have to engage in a Rensselaer-type agency analysis for business-conducting contractors.
• 145Google III, 949 F3d at 1348 (Wallach concurring).
• 146Id (emphasis added).
• 147Further, without the human-agency requirement, it wouldn’t follow that every end user’s computer is Google’s regular and established place of business. Instead, because a computer (and the shelf on which it sits) should not be counted as a “place” under § 1400(b), the inquiry stops there if a place requirement is used. See Part IV (advancing a robust place requirement).
• 148As a knee-jerk reaction, it strikes me as implausible that an end user’s computer could count as Google’s regular and established business, unless Google in some way owns or controls the computer. Nonetheless, at least one judge on the Federal Circuit obviously thinks the question is close enough to bring up. And given the Eastern District of Texas’s tendency toward broad venue, it wouldn’t be at all surprising to see some decisions from that district that use this concurrence as a springboard for establishing patent venue, yet again, over Google. In other words, the saga continues.
• 149See, for example, Google III, 949 F3d at 1346–47 (finding no venue because there was no agent and characterizing this conclusion as concordant with the venue-limiting purpose of § 1400(b)). See also Part IV (showing that a robust place requirement would address this concern).
• 150For example, even if the end user is Google’s agent, the place of business—presumably the user’s computer—must still be the place of Google. See Google III, 949 F3d at 1340 (approvingly reciting In re Cray’s venue requirements, including the requirement that the place be “the place of the defendant”). Presumably, an end user’s own computer would not satisfy this requirement. But what if a plaintiff pointed to a user who used a Google-made device—perhaps a Pixel phone—that the user had bought from Google but not fully paid off yet? See Rick Osterloh, Google Hardware. Designed to Work Better Together. (Google, Oct 9, 2018), archived at https://perma.cc/DX9A-URCY (describing Google’s hardware, including the Pixel). Is such a phone a place of Google? Perhaps even on a lax reading of “place,” a phone wouldn’t make the cut. But maybe a larger and more permanent device, like a Google Home Hub, could. See id (describing the Google Home Hub). Either way, the complexity of this inquiry speaks for itself: a court that determines an end user can count as an agent has invited a wild-goose chase for an end user who satisfies the rest of the venue test.
• 151Even if a district decides that end users can never count as Google’s agents, a plaintiff might be able to find a non–end user and argue that this user is Google’s agent. For example, there might be a business in the district that places advertisements on Google. If so, perhaps the employees who design the Google-bound ads count as Google’s agents.
• 152Google III, 949 F3d at 1348 (Wallach concurring).
• 153Note that rejecting the human-agent requirement and reading “place” more restrictively are analytically distinct. For example, the Supreme Court could overrule the Federal Circuit just on the human-agent requirement question and leave the question of defining “place” to the Federal Circuit. In my view, this outcome would make some sense: the human-agent requirement is a clear-cut matter of statutory right and wrong, while the place requirement is on a spectrum.
• 154For a summary of the Federal Circuit’s rundown of these requirements, see Part I.C.
• 155In re Cray, 871 F3d at 1362.
• 156Id (quotation marks omitted).
• 157Id.
• 158Id (quotation marks and alteration omitted).
• 159Because the rule is really just a preliminary qualification before courts can get to the standard, I generally refer to the place requirement as a standard. When I’m referring to the rule part of the analysis, I’ll say so.
• 160Rensselaer, 2019 WL 3755446 at *10–11.
• 161Id at *10.
• 162Id at *10–11.
• 163Id at *10 (quotation marks omitted).
• 164Rensselaer, 2019 WL 3755446 at *10, quoting In re Cray, 871 F3d at 1362. The word “substantial” is intentionally vague; because “place” is a standard rather than a rule, there simply isn’t a more precise way to put it.
• 165Rensselaer, 2019 WL 3755446 at *10. Among others, the court cited CDx Diagnostic, Inc v United States Endoscopy Group, Inc, 2018 WL 23888534 (SDNY), for the (in my view, dubious; see Part IV.B) proposition that some storage units “likely” counted as a place. See Rensselaer, 2019 WL 3755446 at *10, citing CDx Diagnostic, 2018 WL 2388534 at *3.
• 166Rensselaer, 2019 WL 3755446 at *10–11.
• 167Id at *11.
• 168Id.
• 169See Google III, 949 F3d at 1343.
• 170See id (“Google’s petition suggests that a court’s inquiry into whether the defendant has a physical ‘place of business’ should focus on whether the defendant has real property ownership or a leasehold interest in real property. We hold that a ‘place’ need not have such attributes.”). Instead of real property ownership or a leasehold interest in real property, the defendant just has to “possess or control” the property. Id (alteration omitted).
• 171Id at 1343–44. In addition to the two justifications I’m about to discuss, the court pointed out that the servers were “physically located in the district in a fixed, geographic location.” Id at 1343. See also id, citing In re Cray, 871 F3d at 1362 (“While the place need not be a fixed physical presence in the sense of a formal office or store, there must still be a physical, geographical location in the district from which the business of the defendant is carried out.”) (citation and quotation marks omitted). This claim is true, but trivially so. Any physical thing that doesn’t move, like a wall clock or a ceiling fan, is “in a fixed, geographic location.” Google III, 949 F3d at 1343. Yet those are obviously not places. Contrast a server, which is in a physical location, with the In re Cray court’s example of something that would satisfy the place requirement—a “home”—which itself is a physical location rather than just being in a location. All this is to say that a thing does not become a place merely by staying put.
• 172Google III, 949 F3d at 1343.
• 173Id at 1343–44, citing Tinnus Enterprises, LLC v Telebrands Corp, 2018 WL 4560742, *4 (ED Tex), and Peerless Network, 2018 WL 1478047 at *3.
• 1742018 WL 4560742 (ED Tex).
• 175Id at *4. Perhaps because of this substantial presence, the defendants in Tinnus Enterprises didn’t even dispute the place requirement, which led to a dearth of analysis on it in the opinion. Id. Granted, the court did characterize leased shelf space as the relevant place of business—but only under the “place of the defendant” analysis, which considers the defendant’s level of control. Id (emphasis added). By the time the court reached this part of the analysis, it already considered place satisfied because the defendants had conceded that point upfront.
• 176Peerless Network, 2018 WL 1478047 at *3 (alteration in original) (“The shelf is a ‘physical place in the district’ insofar as it is ‘[a] building or a part of a building set apart for any purpose.’”), quoting In re Cray, 871 F3d at 1362.
• 177In SEVEN Networks, the court analyzed quite similar servers. In fact, they may be some of the very same units. Both cases featured servers hosted by Suddenlink. Compare Google III, 949 F3d at 1340–41, with SEVEN Networks, 315 F Supp 3d at 950. The SEVEN Networks court’s opinion included pictures of a set of servers, which appear to have taken up about half of one server rack. See SEVEN Networks, 315 F Supp 3d at 950.
• 178Google III, 949 F3d at 1340–41.
• 179Id at 1340 n 1.
• 180Peerless Network, 2018 WL 1478047 at *1.
• 181Indeed, perhaps the Peerless Network court let its intuitions show when, even after holding the place requirement satisfied, it stated the following: “Under a faithful reading of the statute, the Court must conclude that whatever a ‘place of business’ is, it is not a shelf.” Id at *5.
• 182280 F Supp 3d 922 (ED Tex 2017).
• 183See SEVEN Networks, 315 F Supp 3d at 950–54. The place analysis in SEVEN Networks focuses almost exclusively on Google’s level of control over the servers in question. I am not sure of the relationship between control and whether the servers are a place in the first, well, place. In fact, In re Cray already had a prong that considers control-of-place: the place must be “of the defendant.” In re Cray, 871 F3d at 1364–65. The SEVEN Networks court may have been mixing the two analyses. See also Tinnus Enterprises, 2018 WL 4560742 at *4.
• 184Personal Audio, 280 F Supp 3d at 934.
• 185Id. Even Google III’s analysis wouldn’t contest this point; Google III didn’t have to do with the server rooms, but instead with the server shelves.
• 186Id.
• 187Id.
• 188Or, as an amici brief in the SEVEN Networks appeal put it, “A Thing Isn’t A Place.” Brief of Etsy, HP, Netflix, Ringcentral, Red Hat, Salesforce, SAP America, Twitter, and the High Tech Inventors Alliance as Amici Curiae in Support of Petitioner, In re Google LLC (Google I), No 18-152, *4 (Fed Cir filed Aug 29, 2018) (Etsy Brief) (going on to argue that, had Congress wanted to refer to objects in the patent venue statute, it could have used the word “property” instead, as it did in the general civil venue provision). See also 28 USC § 1391(b)(2).
• 189With the qualification, of course, that this is very unclear and ultimately depends on how the human-agent requirement gets implemented in any given district. See Part III.B.
• 190It’s slightly more of a tragedy that these behemoths could have forum shopping leveraged against them. Forum shopping is not a good thing. But again, the solution to this isn’t to strip away venue from districts to which defendants really do have a significant physical connection. Similarly, it doesn’t seem too problematic that the owner of cell towers could be haled into court in any district where it has a tower and commits acts of infringement. See In re Google LLC, 914 F3d 1377 1380 (Fed Cir 2019) (Google II) (Reyna dissenting from denial of rehearing en banc) (seeming to suggest that venue in such a case would be a bad outcome). This seems no more problematic than the notion that Apple can be haled into court in any district where it has an Apple Store (and has committed acts of infringement, of course). See 28 USC § 1400(b).
• 191William Shakespeare, The Tragedy of Julius Caesar Act I, Scene II at 142–43 (Folger Shakespeare Library) (Barbara A. Mowat and Paul Werstine, eds).
• 192See Etsy Brief at *4–5 (cited in note 188) (urging the Federal Circuit to overturn the district court’s decision because “[t]he district court’s decision rests on the notion that a thing (here, a computer server) occupies space and is therefore a ‘place’ [yet m]erely because an object occupies space in the real world doesn’t mean that it is a place”).
• 193One could also argue that, because the human-agent requirement limits venue more than “place” does on its own, the former fulfills the aim of the Act better than the latter. But obviously the Act didn’t aim for maximal venue limitation; if it did, it would have declared that venue was proper for patent infringement suits in only one judicial district, specified in one way or another. Further, the Act aimed to clarify venue as well as limiting it. Thus, even if the human-agent requirement were marginally better at fulfilling the venue-limiting function, this advantage would be outweighed by the drastic increase in uncertainty (and intercourt disagreement) the human-agent requirement will cause.
• 194See Part III.A (arguing the human-agent requirement is based on a misreading of § 1400(b)).
• 195See Part III.B (describing the judicial administrability problems of the human-agent requirement).
• 196A possible objection is that Peerless Network really did replace the place requirement with the human-agent requirement; that’s why the court’s analysis of “place” was so short. But this is a huge stretch. It only takes a little reflection to see that, no matter one’s view on agency, courts are obliged by § 1400(b)’s text to do the “place” analysis. There’s no getting around that. Peerless Network’s neglect of “place” is, therefore, best viewed as nothing more than a compliance error.