Architecture of consent: Internet protocols and their legal implications.

INTRODUCTION
I. A PROPERTY-BASED VIEW OF THE INTERNET
   A. A Brief Introduction to the Internet
   B. Articulating Our Intuitions: Notions of Property on the Internet
   C. Applying Our Intuitions: Governing the Internet as Property
     1. Internet common law: Trespass on the web
     2. The Internet and intellectual property
   D. Critiquing Our Intuitions: The War over Internet Property Rights
     1. Defending Internet property rights
     2. Attacking Internet property rights
II. A CONSENT-BASED VIEW OF THE INTERNET
   A. An Introduction to Consent
   B. Architecting Consent Through Code
   C. Implied Consent: The Low-level Protocols
     1. The link layer
     2. Internet protocol
   D. Express Consent: TCP Connections
     1. Transmission control protocol
     2. Legal implications
   E. Defining Expectations: The Application Layer
     1. HyperText transfer protocol and beyond
     2. Legal implications
III. COMBINING PROPERTY AND CONSENT: THE INTERNET AS A COMMUNICATIONS
NETWORK
   A. Debunking the Real Property Metaphor
   B. Consent as a Defense to Trespass to Chattels
     1. The server owner expressly notifies the crawler operator not
     to crawl the site
     2. The website contains human-readable terms expressing a crawler
     exclusion policy
     3. The website contains machine-readable terms expressing a
     crawler exclusion policy
     4. The crawler operator obfuscates the source IP address of its
     datagrams
     5. The crawler "exceeded the scope of any such consent'" it has
     been granted
   C. Consent and Copyright
   D. Boundaries of Consent
CONCLUSION

INTRODUCTION

As more of our societal interactions, both personal and professional, take place on the Internet, the pressure to apply a cogent legal scheme to govern this maturing technology grows increasingly urgent. Faced with a fundamentally new creation, courts have been struggling to formulate a comprehensive view of the Internet that will guide them towards the proper application of existing legal doctrines.

Recent decisions indicate that the courts are just now beginning to approach such a view. Influenced by popular intuitions of Internet architecture, they are starting to see the Internet from the perspective of property rights. They envision the Internet as a massive interconnection of individually owned chattels or plots of land, each containing the intellectual property of its owner. They therefore apply existing property law doctrines, both statutory and common law, to manage and protect these varied property interests.

But in adopting this view, courts make the crucial mistake of ignoring the fundamental purpose underlying the Internet's design: communication. An examination of the Internet's design principles reveals a network based on a norm of consent, not one of strong property rights. Manifestations of this consent occur both implicitly and expressly. The very lowest-level Internet protocols are built to operate via cooperation, and compliance with them should therefore imply consent to certain behaviors that would be prohibited under a purely property-based scheme. Higher-level protocols, such as those utilized in most web interactions, involve exchanges that should be considered express consent: the formation of a legally binding contract.

Current jurisprudence is unjustified in protecting Internet property rights so broadly. Although the property-based intuitions articulated by the courts aren't wrong, they only tell part of the story. Modern legal thought has retreated from the notion of complete and inviolable property interests; (1) the Internet should be no exception. By acknowledging the nature of the Internet as a communications network, we can properly identify the ways in which Internet property owners voluntarily circumscribe their property rights. We can apply the correct legal doctrines only by understanding the technology they govern.

This Note undertakes to examine Internet technology in some detail and discuss how its communicative nature creates a norm of consent that is fundamental to its legal framework. Part I discusses the current, property-based view of the Internet, focusing on how courts have applied it in recent cases. Part II introduces the technological underpinnings of the Internet and argues for a legally enforceable norm of both implicit and express consent based upon the protocols by which Internet-connected computers communicate. Part III then applies the ideas of Part II to the doctrines and cases from Part I, arguing that the Internet norms of implicit consent and express contract formation function to undermine the property-based arguments on which the courts currently rely. I conclude by arguing for a technically accurate version of Internet law that accounts for the freedom built into the Internet's protocols.

I. A PROPERTY-BASED VIEW OF THE INTERNET

A. A Brief Introduction to the Internet

Although Part II will go into much greater detail about the Internet's technical underpinnings, a brief overview here will help to ground the discussion and introduce some useful terminology. Although sometimes used to refer only to the World Wide Web, the term "Internet" broadly encompasses a global array of interconnected computers that communicate with one another. (2) This communication usually happens in one of two ways: "peer-to-peer architectures," in which computers communicate as coequals, (3) and "client/server architectures," in which one computer functions as the "server" and processes the requests of the other computer, the "client." (4)

One of the most popular forms of client/server architecture on the Internet is the "World Wide Web," a system of Internet servers that "serve" files to clients in a special format. (5) This format, called "HyperText Markup Language" (HTML), allows "links" to other HTML documents, as well as links to different types of files such as media content (graphics, audio, and video). (6) Using a computer program called a "browser," (7) a user can utilize her client machine to locate and display HTML "websites" (8) and the other files to which they link.

Conceptually, legal analysts have found it useful to view networks such as the Internet as having three distinct layers. (9) The top layer is the "content" layer, comprising the human-usable information that is exchanged, such as webpages and media content. This content layer utilizes a "code" layer of software that is designed to send and receive content as a distinctly non-human-readable sequence of bits and bytes. These bits and bytes, in turn, are transported along the "physical" layer, comprised of the machines themselves and the various direct connections such as phone lines, cables, and wireless connections.

B. Articulating Our Intuitions: Notions of Property on the Internet

In attempting to discern what law, if any, might govern the architecture of the Internet, both courts and commentators have thus far focused on applying a regime based upon property law. (10) The idea of applying property law to the Internet is an intuitively pleasing one, for three main reasons. First of all, the physical machines that together comprise the network are each individually the property of some entity. Although most computers spend some, if not all, of their lives as part of a larger network, (11) there are clearly significant personal property interests at stake in each node of this network. Secondly, the colloquial language of the Internet prompts us to think of the virtual world as a physical space and want to regulate it as such. (12) When I speak metaphorically of "visiting" a web-"site," it's almost as though I am physically traveling to someone else's land. (13) Finally, the growth of the Internet as one of the main media for the dissemination of ideas gives rise to significant interest in protecting the intellectual property it contains. (14) In this conception, a website is less like a location and more like a book(15)--a publicly available font of knowledge whose expression of various ideas needs to be protected.

Each of these three intuitions points towards a different scheme for Internet property governance. The physical makeup of the network suggests that the law of chattels ought to apply, with each node treated as an inviolable piece of tangible personal property. (16) The place metaphor provides compelling rhetorical support for a real property scheme where websites are treated like plots of land onto which end-user clients might enter. (17) Finally, the informational content of the Internet argues for the application of intellectual property law such as statutory copyright and trademark law, as well as common law unfair competition and misappropriation. (18)

C. Applying Our Intuitions: Governing the Internet as Property

1. Internet common law: Trespass on the web.

These conceptions of property on the Internet, and courts' views of them, become more concrete if we focus, as an example, on the paradigmatic case of a "crawler"--an automated program that serially visits, or "crawls," websites and keeps a log of what it finds. (19) Crawlers are useful for, among other things, indexing websites to create the databases used by search engines. (20)

What cause of action might a website proprietor have when an unwanted crawler combs its website? This was exactly the case before the court in eBay, Inc. v. Bidder's Edge, Inc. (21) Plaintiff eBay, a prominent online auction site, sued defendant Bidder's Edge, an auction aggregator whose business model was to crawl many auction sites, aggregate data about their auctions, and offer users an interface by which they might search this aggregate data without having to individually visit every site. (22) eBay sought a preliminary injunction to stop the Bidder's Edge crawler from visiting its site. (23)

eBay proffered a number of alternative legal bases to support its request, (24) highlighting each of the three property views discussed in Part I.B. As might be expected, eBay's pleadings exhausted the supply of intellectual property theories, arguing copyright infringement, (25) trademark dilution, misappropriation, and unfair competition. (26) Yet the court found it unnecessary to reach the merits of any of these claims, instead finding an injunction to be warranted because Bidder's Edge had likely committed a common law trespass. (27) In making this finding, the court considered separately the possibilities of real property trespass (28) and trespass to chattels. (29) As to the former, the court analogized the Bidder's Edge webpage visits to unwanted customers visiting a "brick-and-mortar" store. (30) While the court seemed receptive to the idea that this analogy could have legal salience, it was uncomfortable finding that virtual visits constituted the "irreparable harm" necessary for a preliminary injunction. (31) Instead, the court rested its ultimate grant of the injunction on trespass to chattels, finding that electronic impulses sent from Bidder's Edge computers constituted physical "interference" with eBay servers. (32) Because eBay was required to devote some of its bandwidth and processing power to deal with these unwanted intrusions, the court reasoned that they disrupted eBay's property interest in these resources. (33)

Though only a district court decision, eBay is representative of the modern legal doctrine. In addition to being directly followed by other district courts facing actions involving web crawlers, (34) it has been cited favorably to support the idea of a common law property interest in "computer equipment, bandwidth, and server capacity." (35) The California Supreme Court recently discussed eBay and other crawler cases at length, suggesting that an action for trespass would lie if a plaintiff could "demonstrate some measurable loss from the use of its computer system" such as a showing that the defendant "actually did, or threatened to, interfere with the intended functioning of the system, as by significantly reducing its available memory and processing power." (36)

Equally as important as eBay's holding is the way in which the opinion worked through the various intuitions of Internet property. First of all, eBay provides support for the idea that intellectual property rights and common law property rights can coexist as separate legal doctrines on the Internet. Considering whether statutory copyright law preempted a common law trespass action, (37) the opinion noted that "[t]he right to exclude others from using physical personal property is not equivalent to any rights protected by copyright" and thus a trespass claim was "qualitatively different from a copyright infringement claim." (38) eBay thus avoided copyright preemption of the trespass claim by considering the trespass a physical invasion of the server rather than of the content it contained, (39) drawing a clear legal line between the content and physical layers of the Internet. (40)

Secondly, though the court rested its ultimate holding on trespass to chattels, the opinion also suggested that real property trespass was a viable doctrine on the Internet. On a theoretical level, eBay was not unfriendly to a real property analogy, readily comparing unwanted web visits to unwanted physical visits. (41) However, when forced to make a choice between real property trespass and trespass to chattels, it favored the latter because this made it easier to show the irreparable harm necessary to grant the injunction. (42) Although the impact on eBay servers was slight (eBay alleged that Bidder's Edge was responsible for less than two percent of its traffic), (43) the court nevertheless found that irreparable harm would ensue if it failed to issue the injunction. The court reasoned that its abstention would encourage other aggregators to send their crawlers to eBay's site; their combined effect would severely inhibit eBay's ability to serve its regular customers. (44) The court was less sanguine about whether the harm under the real property analogy would justify an injunction, expressing skepticism about whether two out of every hundred customers to a brick-and-mortar store could really interfere with the overall shopping experience. (45)

But this analysis of irreparable harm for the purposes of the preliminary injunction was separate from the court's later consideration of whether there was sufficient damage to support the merits of a trespass action. (46) In assessing the merits of trespass to chattels, the court found damage had resulted from Bidder's Edge's unlicensed use of even a small percentage of eBay's bandwidth and computing power, reasoning that the rightful owner might have used these for its own purposes. (47) But in assessing the merits of an action for real property trespass, not even this minimal showing would necessarily be required. Since owners of real property have traditionally enjoyed exclusion rights regardless of harm, (48) plaintiff website operators in future actions that do not involve preliminary injunctions might still be tempted to advance a theory of real property trespass. Nothing in the eBay decision would preclude such an argument; its influential reasoning is thus indicative of an expansive view of strong Internet property rights.

2. The Internet and intellectual property.

The application of copyright law to the Internet is somewhat murkier than the application of trespass, largely because copyright infringement is a somewhat more nebulous doctrine. (49) To illustrate this, we can start by considering the threshold question of what constitutes a "copy." Under one view, most web-based actions--such as accessing a webpage, listening to a sound file, or viewing a video clip--should be considered making a copy, since these all involve loading the material in question into the client computer's random access memory (RAM). (50) This "RAM copy doctrine" (51) finds support in multiple appellate opinions expressing the view that a local RAM copy of a digital data source constitutes a "fixed copy" under the Copyright Act. (52) Courts and commentators are split over the RAM copy doctrine and how it might apply on the Internet. (53) At least one district court has applied it to find simple browsing of a copyrighted work to be infringement. (54) Other courts, however, have attempted to avoid the doctrine's implications. (55) Were one to fully accept it, one might argue that copyright law preempts all state law causes of action on the Internet--including trespass. (56)

Although this all-encompassing idea might at first seem somewhat overprotective of an owner's interests, the RAM copy doctrine, and copyright in general, might actually protect the interests of website visitors more than would common law property. Faced with a copyright action rather than a trespass, the court would necessarily focus on the nature of the acquired content and how it was put to use, rather than the method used to acquire it. Unlike common law property, copyright is a creature of statute that purports to strike a balance between the owner's interest in control and the public's interest in access. (57) Courts might therefore be inclined to allow a great deal of access, finding either that the use falls under the fair use doctrine (58) or that the information used is not sufficiently expressive to be protected under copyright law. (59) Hope for such a result is implicit in critical attacks on eBay that have castigated its common law creation of a "new intellectual property right" protecting "every piece of information on the plaintiff's Web site, whether or not it is copyrighted, patented, or secret." (60)

Recent Internet copyright cases illustrate how courts might utilize copyright's built-in exceptions to assert a more access-protective view than have the courts in the trespass cases. Ticketmaster v. Tickets.com, Inc. involved a situation quite similar to eBay: The defendant's crawler accessed price information about tickets being sold on the plaintiff's website. (61) Although it accepted the RAM copy doctrine, the court in Ticketmaster nevertheless found the crawler's action permissible because the defendant utilized only price information, which the court dee med to be nonexpressive content properly failing outside the plaintiffs copyright-protected aegis and into the public domain. (62)

In Kelly v. Arriba Soft Corp., plaintiff Kelly was a professional photographer who operated an Internet site exhibiting his photographs; the defendant Arriba ran a search engine that indexed images from other websites (including Kelly's) via small low-quality reproductions known as "thumbnails." (63) By clicking on one of the thumbnails on the Arriba site, a user could bring up a browser frame containing the full image imported from Kelly's server. (64) Although the court found that Kelly had established a prima facie case for infringement for both the thumbnails and the "framing," (65) it applied the fair use factors to hold that the thumbnails were permissible whereas the framing was not. (66) Though used for commercial purposes, the thumbnails were sufficiently transformative, used for a distinctly different purpose than the original, and did not have an adverse effect on the market for Kelly's work. (67)

Ticketmaster and Kelly, however, are not necessarily indicative of a trend toward expansive Internet access rights under copyright or the RAM copy doctrine. First of all, there are legislative notions of removing much of the meat of the nonexpressive content exception as it applies to the Internet. (68) Secondly, many courts are much more reluctant to apply fair use than was the court in Kelly. (69) The standard for fair use is "less than crystal clear," (70) and the exception forces defendants to work their way out from under a presumption of infringement. (71) Therefore, practitioners are wary of trusting themselves to the ad hoc application of the fair use doctrine. (72) Though theoretically more permissive of access than common law property, copyright still operates as a mechanism of control. (73)

D. Critiquing Our Intuitions: The War over Internet Property Rights

1. Defending Internet property rights.

Is it a good idea to apply common law property and intellectual property to govern the architecture of the Internet? As noted above, the idea is intuitively satisfying. (74) These intuitions can also be supported by theory: One can certainly argue that website owners should have a right to the fruits of their labor, (75) that they have a strong "personality interest" in their ownership, (76) and that they should have a right to control their physical property even when it is connected to a network. (77)

Furthermore, its proponents argue that Internet property is good from a policy perspective. First of all, they claim it promotes the expansion of the web. Property rights to created works spur creativity, (78) so the ability to control one's web presence may incentivize the creation of such a presence. (79) For example, had eBay not thought it could protect the contents of its virtual marketplace, it might not have gone into the online auction business in the first place. (80) Secondly, they argue that it helps to avoid a "tragedy of the commons" problem, where a public made up of greedy individuals overutilizes a resource until it is no longer of any use to anyone. (81) If one accepts the idea that network bandwidth and/or server processing power are limited resources, then a tragedy of the commons is possible if, for example, too many users try to simultaneously visit the same website. (82) Thirdly, they point out that even in the presence of exclusion rights, access can be purchased, and they argue that this develops a more beneficial market than would exist with a system of free access. (83) The existence of this market helps to ensure the maximization of social utility: If Bidder's Edge places a higher value on having eBay's pricing information than eBay does in protecting it, the two can negotiate a licensing fee. (84)

2. Attacking Internet property rights.

The recent trend towards strong Internet property rights has no shortage of critics. These critics do not necessarily oppose all property rights on the Internet; instead, they argue that these rights should be more circumscribed than they currently are. For example, they might acknowledge the necessity of copyright to promote innovation, but oppose extending its scope or duration. (85) They tend to argue against anything that threatens to upset the control/access balance.

Trespass is a case in point. Because they view the eBay doctrine as overprotective, many of these critics argue that the law of trespass is misapplied on the Internet. They argue that Internet trespass is an unprecedented amalgam of trespass to chattels and real property trespass that creates an unjustified level of property protection. (86) They particularly regret the apparent trivializing of trespass' harm element: As we have seen, minimal server usage seems to suffice as harm for a trespass to chattels, and real property trespass might lie without any showing of harm at all. (87)

The critics' main normative point is that the public interest in innovation is threatened by strong mechanisms of control. (88) Society as a whole benefits from innovation, and innovation is greatly facilitated by the ability to access, utilize, and improve upon the ideas and information of others. For example, Bidder's Edge's aggregation of pricing information from multiple auction websites may be more socially beneficial than having several segregated auction websites, because it gives consumers more information at once and makes it easier for them to comparison-shop. (89) Strong property rights, however, threaten such utility-maximizing developments: When control of information rests wholly with a particular individual entity, that entity will seek to maximize its personal profit, rather than the welfare of society as a whole. (90) Thus, while increased competition might be good for society overall, eBay might still choose not to license its pricing information to Bidder's Edge. (91)

II. A CONSENT-BASED VIEW OF THE INTERNET

Critics of the expansion of property rights on the Internet opine that they are losing the war. (92) The proliferation of such doctrines as RAM copy and common law trespass help to explain their pessimism. While there are normative arguments on both sides of the divide, the property proponents perhaps possess a trump card in the strength of our societal intuitions: For strong moral and normative economic reasons, we tend to favor property. In many ways, it makes sense to yield to these intuitions. Trying to have legal rules that map to intuitions and norms is generally a good thing. We want the law to make sense to us and reflect who we are.

There are, however, normative arguments on the other side as well. The critics argue that in addition to our societal property norms, we also have strong societal norms in favor of freedom. (93) They point to the halcyon days of the Internet's youth as embodying this freedom norm. They argue that the Internet's original goal was one of free exchange, rather than one of control, (94) and that it should remain this way. As Lawrence Lessig puts it:

   The protocols of the Net embedded principles in the Net that
   constructed an innovation commons at the code layer. Though
   running on other people's property, this commons invited anyone to
   innovate and provide content for this space. It was a common market
   of innovation, protected by an architecture that forbade
   discrimination. (95)

A. An Introduction to Consent

Critics of Internet trespass have appealed to these Internet norms by advancing "consent" as a defense to the trespass claim. They argue that a website "consents" to access by users, even competitors. (96) If a website really found access by a particular user undesirable, it would choose to exclude, (97) much as someone with caller ID might choose not to answer a phone call from a solicitor.

The idea of a computer "consenting" may seem counterintuitive. Computers, after all, are mindless automata; the term "consent," with its connotations of "voluntary involvement ... in a process," (98) perhaps appears inappropriate. However, it is precisely because they are automata that computers serve as perfect conduits for the consent of their owners. Every task a computer performs is something that a human programmer has taught and that a human operator has requested. The "consent" comes not from the machine itself, but from the operator. Every time we use computers to perform tasks, we blur the distinction between humans and machines. When I click my mouse to buy a book online, is the purchase any less valid because my computer transmits my credit card data as opposed to my signing a sales slip?

Nor is the concept of machine as consent conduit confined to situations where the human's action is contemporaneous with the machine's. Imagine an answering machine connected to caller ID and programmed to only take messages from certain callers. Would we not say that the owner has consented for the nonblocked callers to leave messages and that the actions of the machine merely manifest this consent? Further, imagine a day trader who runs a computer program that will electronically purchase a certain stock when it reaches a certain price. If the computer does so and the price later plummets, would we allow her to repudiate her purchase by asserting that she never consented to the action taken by her computer?

All of the courts in the trespass cases treated the consent argument seriously, rejecting it on legal-factual grounds rather than finding it invalid per se. In considering argument on its merits, they offered up a wide range of possible reasons why consent might not have existed under the particular circumstances of each case: (99) (1) express notification to the crawler operator not to access the site; (100) (2) human-readable terms on the website expressing a policy of excluding crawlers; (101) (3) machine-readable terms on the site expressing a policy excluding crawlers; (102) (4) the inability of the website to know who was requesting access because the crawler intentionally obfuscated its source; (103) and (5) the general notion that even if there had been some consent, the crawlers "exceeded the scope of any such consent." (104)

However persuasive these counterarguments might at first glance seem, the consent argument deserves further consideration. Rather than being simple nostalgia, Internet norms have legal implications. While the content of the Internet has changed--growing from a small network of researchers into a global amusement park, speakers' corner, and shopping mall--its core technical specifications have remained largely unchanged. (105) The original Internet still exists in the code layer.

If we peel back the superficial analogies at the surface of Internet discourse and explore the Internet's technical underpinnings, we uncover a vibrant technology based upon a norm of consent. The Internet is a communications network, and consent is fundamental to every level of its proper functioning.

This Part will examine the two ways in which this consent norm operates. First of all, the most basic functions of the Internet--sharing wires and sending and receiving messages--would be impossible without the cooperation of every single machine connected to the global network. This cooperation, contrary to judicial assumptions, requires implied consent to outside use of bandwidth and processing power. Secondly, more complicated Internet functions--such as the formation of the type of connection necessary for client/server Web interaction--involve an express waiver of a machine's exclusion rights. I will argue that these express waivers constitute legally binding contracts that directly override whatever exclusionary rights a server owner might possess by default.

B. Architecting Consent Through Code

Arguing that the design of the Internet functions to control how it is used and the norms associated with this use, Lawrence Lessig has noted that "[c]ode ... is a law of cyberspace." (106) In order to understand the consent norms of the Internet and their possible legal implications, it is necessary to delve somewhat deeper into its technical specifications.

Though quite useful for analytical purposes, the layered view of network design discussed earlier is an admitted oversimplification of the Internet. (107) In particular, what it dubs the "code layer" actually consists of numerous sublayers, together comprising what network engineers call the "protocol stack." (108) A "protocol" is a set of rules that a software program follows in exchanging messages, defining both the format of messages as well as when each type of message can or ought to be sent. (109) Different protocols combine to form a "stack" because they operate atop one another to implement the complete functionality of the Internet's code layer. (110)

The protocol stack functions through both implicit and explicit consent. The network engineers who created the Internet designed it this way for a reason: In their view, it was the best way to solve the various problems that came up in building a flexible, reliable system. (111) The easiest way to understand their choices, and thereby come to understand why consent is so integral to the functioning of the network, is to attempt to get into their heads and try to rebuild the Internet from scratch, focusing on how we would establish web connections. (112)

C. Implied Consent: The Low-level Protocols

1. The link layer.

To start, we simply want to construct a "local area network" (LAN) by connecting a few physically proximate machines to each other using some sort of physical connection. (113) For example, we might connect multiple machines to a shared wire. This is conceptually similar to being on a Navy submarine with a shared public address (PA) system. In both cases, there's only a single medium of communication, so any message will be heard by everyone. (114) In order to make sure the intended recipient listens to the message, we implement a protocol where every message (called a "frame") (115) has to have a "header" specifying the sender and receiver ("Joe, this is Fred ..."). (116) This is known as a "link layer" protocol, because it governs how multiple users share a single link. (117) On our LAN, we use unique identifiers embedded in the hardware of each computer for "names." (118) The details of the protocol depend on the type of link: Local wired connections typically use one protocol, (119) wireless access points typically use another, (120) and dial-up connections typically use yet another. (121)

Already we can begin to see how the design requires the implicit consent of the participants. Every time a message is broadcast over the local connection, every machine on the connection receives it. One effect of this is that multiple machines can't "talk over" one another, lest their messages become garbled so that the recipient can't understand them. (122) Another effect is that every machine on the wire needs to at least read the header of each message in order to determine whether it is the intended recipient, (123) just like roommates need to read all the envelopes in order to find their own mail. This means that each machine on the LAN devotes at least some of its processing power to header-reading, (124) much of which will simply result in the ignoring of the message in question.

Recall that in the trespass cases, unwanted use was sufficient to support the cause of action. Do I therefore have a cause of action against other machines on the LAN if they continue, despite my protestations, to send traffic intended for each other but not for me? Hopefully not--the norms of the protocol would seem to indicate that I implicitly consented to this kind of use when I chose to become part of the LAN. If each machine on the network analyzed its own prisoner's dilemma concerning local traffic and decided to act selfishly--talking over others' messages, for instance--the LAN simply could not function. By plugging into the system, I implicitly consent to reading headers I would rather not read and remaining silent while others speak. This consent is simply part of the cost of doing business.

2. Internet protocol.

Now suppose we want machines on our LAN to talk to a distant LAN, or, analogously, we want sailors on one submarine to be able to communicate with sailors on another. Obviously, we can't use our PA system for this--we need a radio. But we can't give every sailor his own radio--we don't want hundreds of personal antennas sticking out of our hull, and perhaps we want to implement some kind of consistent military radio policy. So we have a single radio operator to serve the entire submarine who knows how to send messages to certain other submarines. Suppose Alice, the commander of submarine A, wants to send a message to the commander of submarine B (whose name, unbeknownst to Alice, is Bob). She gets on the PA system and sends a message to Amy, her radio operator, saying "Amy, this is Alice. Message from Commander A to Commander B. Message is as follows...." The radio operator then sends this message to the radio operator on submarine B. "Hi, Operator B, I'm Operator A. Message from Commander A to Commander B. Message is...." Becky, the radio operator on submarine B, receives this message and sends out the following message on her PA system: "Hey, Bob, it's Becky. Message from Commander A to Commander B. Message is...."

Note how we "encapsulate" the message between the two commanders inside three different "hops" (Alice to Amy, Amy to Becky, Becky to Bob), each with its own slightly different protocol (note how the introduction varies for each hop). This makes our system flexible, because everyone's knowledge can be incredibly specialized. Alice doesn't have to know exactly how to reach Bob, or even how to use a radio. She just needs to know (a) how to use the PA and (b) that Amy is the person to go to if she wants to send a message to submarine B.

Internet Protocol (IP) works much like our submarines. We connect LANs to each other using "routers"--machines that, like our radio operators, interconnect different physical networks. (125) In addition to its hardware identifier, we give each machine a global "IP address" that everyone on the Internet potentially knows, just like having "Bob" known globally as "Commander B." (126) When a machine on the first LAN wants to send an IP message (called a "datagram") to a machine on the second, it encapsulates it (including a header specifying the source and destination IP address) inside a link-layer frame to a local router, just like Alice sends her messages to submarine B through Amy and then Becky. (127)

The IP layer widens the consent norm we first observed on the link layer. Not only do routers implicitly consent to give their best effort to properly forward messages, (128) but the link layer consent norms take on broader implications for all machines. Since not all messages on my LAN necessarily originate there, my obligations extend beyond my local geography. If someone on submarine B talks out of turn, they are not only disrupting someone else on submarine B, but quite possibly someone on submarine A as well. The proper routing of messages depends on the cooperation of many machines for the mutual benefit of all. (129)

D. Express Consent: TCP Connections

1. Transmission control protocol.

The link layer and IP have gotten our protocol stack started, but it is not yet complete--there are at least two important complications we haven't yet accounted for. First of all, sometimes messages get lost. Perhaps Amy's radio is temporarily on the fritz and the message never reaches Becky--Alice's vitally important message never reaches Bob, but nobody knows. Secondly, what happens if we want to send a really long message? We don't want to tie up the radio or PA system reading the complete works of William Shakespeare.

What our system lacks is a way to make different messages relate to one another. If we had this ability, it would be very easy to find out that a message had gotten lost, because Alice could simply ask Bob to let her know he received the message. ("Commander B, please acknowledge.") Also, long messages wouldn't be a problem, because we could break them into shorter chunks and allow the other side to string the chunks together. ("Here's part one:...." "Here's part two:....")

This is exactly what Transmission Control Protocol (TCP) does by introducing the concept of a "connection." (130) Again, we use encapsulation. Just like we encapsulate an IP datagram inside a link-layer frame, we encapsulate a TCP message inside an IP datagram (an envelope inside an envelope inside an envelope). Also, just like IP datagram and link-layer frames, a TCP message contains a header with important information that allows us to track what's going on. But rather than containing sender and receiver names, the TCP header contains a "port number," identifying the set of messages to which this one belongs, and a "sequence number," indicating where this particular message falls within that set. (131) The protocol also allows for special messages, called "ACKs," that let the sender know a particular message has been received. (132)

As we will see later in Part II.E, TCP connections are not the end of the protocol stack. (133) For now, it is sufficient to note that TCP connections are usually formed at the request of some application, such as a web browser. When a user running a browser clicks on a link to download a web resource, this is ultimately translated into a TCP request, which gets sent from the client to the server. If the connection is established, the server sends across the requested resource in pieces that the client's TCP layer then reassembles and delivers to the browser for display to the user.

To relate this back to our submarine analogy, let's say Alice wants to send Bob a book, but, because of radio constraints, Amy can only transmit a page at a time. So Alice reads the book page by page, indicating exactly what she's sending each time. ("Here's page 1 of Wonder Boys:...." "Here's page 2 of Wonder Boys:....") As Bob receives each page, he (a) writes it down and places it in order into a binder and (b) sends a short message back to Alice, letting her know he got the message. ("Got page 1 of Wonder Boys.") If enough time goes by without receiving acknowledgement from Bob, Alice resends the page.

Again, note how flexible this system is. For one thing, it doesn't matter if a page arrives out of order (possibly because it had to be resent), because all the pages are numbered and Bob can place the out-of-order arrival in the appropriate spot in his binder. For another, Bob can potentially receive several books at once as long as he's careful to put the pages into the correct binders.

Obviously, TCP connections require processing power from both sides. TCP is a peer-to-peer protocol, meaning that over the lifetime of the connection, each machine might wind up either sending or receiving data. (134) The sender has to compute or retrieve whatever data it wants to send, break the data down into correctly sized "segments," and send the segments one at a time. (135) In doing so, it must keep track of inflight segments and resend them if they get lost. (136) Meanwhile, the receiver must keep track of which connection is which, acknowledge which segments have arrived, and reassemble segments that arrive out of order for any reason. (137)

Because such coordination is required, TCP cannot simply rely on good faith: One machine doesn't just start throwing segment after segment of TCP data at another machine and hope the other machine responds appropriately. Instead, before any actual data is sent, the connection is established by a process known as the "three-way handshake": Machine X sends a message to machine Y requesting a connection, Y responds by sending the connection's initial sequence number, and X acknowledges the server's response. (138) The connection remains active until too much time elapses without any data going across it, or both the client and the server close the connection with an explicit set of messages. (139)

2. Legal implications.

TCP Connections as Contracts. This three-way handshake is not simply implied consent. It is, in fact, express consent: a contract. As the Restatement (Second) of Contracts states:

   [T]he formation of a contract requires a bargain in which there is a
   manifestation of mutual assent to the exchange and a consideration.
   (140)

   The manifestation of mutual assent to an exchange ordinarily takes
   the form of an offer or proposal by one party followed by an
   acceptance by the other party or parties. (141)

The three-way handshake falls within the Restatement's rubric of contract formation. X proposes a TCP connection to Y, Y accepts, and an exchange of data occurs. (142)

This interpretation must be defended against three possible objections. First, we might quibble with the idea that there was a "bargain," because there was no real negotiation. However, in the sense that the Restatement uses it, "bargain" does not mean "haggling." The Restatement defines the "two essential elements of a bargain" as "agreement and exchange," (143) and both of those elements are clearly present here. Were "bargain" interpreted to require negotiation, it would invalidate any agreement that had not been preceded by disagreement. The history of contract law does not support such an underinclusive view. (144)

Secondly, we might question whether there was consideration. The Restatement defines consideration as "a performance or a return promise [that] must be bargained for." (145) It goes on to define "bargained for" in this context as meaning that "a performance or return promise ... is sought by the promisor in exchange for his promise and is given by the promisee in exchange for that promise." (146) This language specifies that a reciprocal exchange of promises is sufficient to constitute consideration, and this is exactly what occurs in the three-way handshake. The sender promises to send the data beginning with a particular sequence number; the receiver agrees to send acknowledgments and reassemble the data into its original format for delivery to whichever application (for example, a browser) requested the connection. (147) The exchange is "bargained for" in the sense that it would not be occurring absent these assurances.

Perhaps we might think that there is no consideration because the receiver seems to always get the better deal. But recall that each machine may be both a sender and a receiver over the lifetime of the connection--in such a case, the benefits are clearly reciprocal. Moreover, even if the data flow is not bidirectional, the receiver's promise to properly reassemble the data and deliver it to the requesting application would itself constitute sufficient consideration for the sender. Contract law makes it clear that consideration need not benefit the promisee--it is sufficient simply that promises are exchanged. (148) Indeed, it is not even clear that the promisee receives no benefit here; the correct reassembly and delivery of the data might be very important in certain contexts. Imagine, for example, a website that makes its money from advertising revenue. The assurance that the advertisements will not be stripped from the data prior to delivery to the browser might be very important.

Thirdly, we might argue that machines have no ability to manifest assent, at least not in a legally binding sense. However, recognizing machines' capacity in this regard would not be a giant leap within legal thought: If we are not already there, we are quite close. Legal scholarship is coming to recognize the advent of contracts made by "electronic agents," defined as "automated means for making or performing contracts." (149) Several uniform codes have adopted this view. (150) Particularly on point is the Uniform Computer Information and Transactions Act (UCITA), which the National Conference of Commissioners on Uniform State Laws has proposed as "a commercial contract code for computer information transactions." (151) Though only Maryland and Virginia have adopted UCITA in any form, (152) and it remains very controversial, (153) a number of recent courts have nevertheless looked to it for guidance, (154) and it can therefore be quite instructive. The UCITA unambiguously makes it clear that:

   [a] person that uses an electronic agent that it [sic] has selected
   for making an authentication, performance, or agreement, including
   a manifestation of assent, is bound by the operations of the
   electronic agent, even if no individual was aware of or reviewed the
   agent's operations or the results of the operations. (155)

This language supports the idea that TCP connections are enforceable contracts. TCP software is certainly an electronic agent "selected for making an agreement:" the agreement to exchange data in a particular way. (156) The fact that human review is absent from these transactions is of no moment in UCITA, (157) nor should it be. As discussed previously, (158) computer programs, including TCP software, do exactly what they're told to do. (159) A website proprietor has the power to modify the default behavior, altering the server to refuse undesired connections: One technological mechanism for doing so is called a "firewall," a combination of hardware and software that isolates an organization's internal network from the Internet at large, allowing the organization to block datagrams from unwanted senders. (160) Thus, when the TCP software manifests assent by accepting a connection, this does, on some level, represent the conscious choice of the software owner, making it reasonable for the manifestation to be binding. (161)

Defining the Terms. An additional objection to the contract theory of TCP connections might be that the three-way handshake is too terse to supply sufficient terms by which we might discern a contract. Lack of detail, however, does not by itself invalidate an otherwise legal contract. Although the Restatement (Second) of Contracts requires a contract's terms to be "reasonably certain," (162) this requirement is not overly restrictive. It simply means that the terms must "provide a basis for determining the existence of a breach and for giving an appropriate remedy." (163) Terms may be reasonably certain even though they empower one or both parties to select other terms in the course of performance, (164) and the act of performance may itself remove uncertainty and establish that a contractual bargain has been made. (165)

When essential terms have been omitted, the Restatement allows a court to supply terms it deems reasonable under the circumstances. (166) It makes several suggestions for how the court may find such reasonable terms. Primary in this methodology is the principle that "technical terms and words of art are given their technical meaning when used in a transaction within their technical field." (167) When relevant, course of performance, course of dealing, and usage of trade may be factored into the interpretation of the agreement. (168)

The Restatement's view certainly supports at least a minimalist interpretation of TCP contracts as agreements to deliver and reassemble data for delivery to an application. The messages are "technical terms" or "words of art" whose understood trade usage is to agree to such exchanges. Course of performance (since the exchange does subsequently occur) and course of dealing (since the same two parties may make multiple TCP connections) further support the contract theory.

But these rules also warn us against going too far in our interpretation of TCP connections. TCP contracts are, in fact, quite minimal. There is no guarantee to deliver a particular piece of data, or any nonempty data at all. The two parties simply agree upon a mechanism for delivery, should such a delivery occur. Nothing in the protocol prevents either party from simply closing the connection right after it is opened. Also, it would go too far to read any warranties into the TCP contract: There is no guarantee that one of the machines involved won't immediately crash or be turned off.

One might be tempted to ask what the point of such a minimal contract would be. I believe there are two. First of all, it establishes express consent for the exchange of data, should such an exchange occur. I will examine the implications of this in Part III. Secondly, it serves as a standard template to which further terms might be added. How might one specify such terms? To see this, we must continue our journey up the protocol stack, looking at the meanings of other protocols we could encapsulate inside a TCP connection.

E. Defining Expectations: The Application Layer

1. HyperText transfer protocol and beyond.

As mentioned previously, the protocol stack does not end with TCP. Thus far, we have not defined the content of the data we might exchange. This is the job of the "application layer" of our protocol stack, which is populated by the computer programs that utilize TCP for data exchange. (169) A web browser is one such application; the protocol it uses to download a webpage is called HyperText Transfer Protocol (HTTP). (170)

HTTP is a client-server protocol whose messages are encapsulated inside a TCP connection. (171) The two machines first open up a TCP connection. Over this connection, the client sends a "request," asking to be sent a particular resource, such as an HTML webpage or some video content. (172) The particular resource being requested is specified by a Uniform Resource Locator (URL), such as "http://www.ebay.com." (173) The server then sends a "reply," which either contains the requested data or some other content (such as HTML data indicating that the requested file does not exist at the specified location). (174)

The protocol stack does not end with HTTP. First of all, there are horizontal alternatives to HTTP that also utilize TCP connections. For example, email upload and download represent a different way to utilize TCP connections for data transfer. (175) Secondly, there are protocols further up the protocol stack that may be themselves be encapsulated within HTTP. (176)

2. Legal implications.

The encapsulation of application layer protocols within TCP connections suggests how we can build upon our minimalist view of TCP contracts. Note how the technical terminology (in the form of requests and replies), usage of trade, and course of performance and dealing create a different set of expectations for HTTP than for bare-bones TCP. (177) No longer is it necessarily acceptable to send nothing at all. Interacting via requests and replies creates a whole new set of behavioral implications.

For example, there is an expectation that the server will make its best effort to retrieve the requested data and send back an error message if it is unable to do so. One could perhaps argue instead that to send an unwanted virus might be a breach of contract or a breach of the implied covenant of good faith and fair dealing. (178) For the client, there is an expectation that the data will be forwarded to some kind of application (such as a browser or video player) capable of displaying the content to the user. Again, one could perhaps argue that some kind of modified browser that strips ads would be a breach of contract or of the implied covenant of good faith and fair dealing.

III. COMBINING PROPERTY AND CONSENT: THE INTERNET AS A COMMUNICATIONS NETWORK

What does this technically focused, consent-based view of the Internet mean for the Internet as property? Certainly, the two views are not mutually exclusive. As in real space, both property and contract law can coexist. But the consent-based view forces a reexamination of what property rights someone actually has on the Internet, and what she might be giving up when she consents to participate in the network protocols.

A. Debunking the Real Property Metaphor

The clearest implication of the protocol stack is that it completely undermines the real property metaphor as applied to the Internet. The idea of "visiting" a web "site," while a colorful colloquial linguistic construction, doesn't describe the transaction well at all. At the HTTP level, most of the data is actually traveling in the other direction: The bulk of the data is sent from the server to the client. It's much more like watching television than it is like visiting a piece of land--the client is mainly a passive receptor of data.

While one might assert that an unwanted HTTP request is a real property trespass, in that one person caused a "thing" (electronic impulses) to enter another person's property (their machine), (179) this doesn't make sense when we realize that the HTTP request is encapsulated within a TCP connection. With the TCP connection, a bilateral contract was reached in which the server agreed to listen to what the client had to say. This is similar to my picking up the telephone to find that there's a solicitor on the other end: I didn't agree to do anything more than hang up, but I hardly have a cause of action for real property trespass if the solicitor starts to talk. (180) If the server didn't want the request, it shouldn't have made the TCP contract.

The only way to make the real property analogy work is to assert that the IP datagram or link-layer frame that attempted to start the three-way handshake was a trespass. But doing this undermines the entire nature of the Internet as a communications network. By simply plugging in to the network and running the appropriate software, the server agreed to deal with these kinds of messages. What's the point of a user having an Internet presence if she is not going to accept IP datagrams?

If we must place this implied consent into some kind of doctrinal box, perhaps we might find some kind of easement or reciprocal servitude based on the planned structure of the Internet. (181) But for this to even be necessary, one must assume that real property law applies in the first place. A technical understanding of the Internet's communication protocols shows how misplaced such an assumption would be.

B. Consent as a Defense to Trespass to Chattels

Many of the same consent-based arguments that apply to the real property view also apply to the chattels view. At each level of the protocol--link layer, IP, TCP, and application--the server has either implicitly or expressly consented to the use of its processing power for other than its own immediate purposes. And it has had good reasons for doing so. We have seen how the Internet would fail to function properly if there were not some sort of cooperation and consent among its participants.

Armed with this knowledge, we must reconsider the previously stated reasons why the courts in the crawler cases did not view consent as a defense to trespass. (182) Though this discussion will necessarily focus particularly on crawlers, it aims to reveal more generally applicable principles as well.

As a threshold matter, all of the arguments in favor of trespass presume some initial right of the server owner to exclude traffic it doesn't desire. The boldest argument in favor of the crawlers, then, is to simply challenge this presumption. Why presume a right to exclude? Why is property necessarily the scheme to which the law defaults? The communicative nature of the Internet makes a strong argument that consent ought to be the norm with property being the exception. This would thrust the burden of proof onto the server owner to justify her exclusion in the first place and argue why she has some special right to participate in the Internet yet ignore the right of others to at least offer to form TCP contracts.

Having made this initial observation, I will put it aside for the moment and retreat to the real world where property occupies an entrenched legal and normative position into which consent must make inroads. I will attempt to present a rebuttal to each of the courts' counterarguments against the consent doctrine in the crawler cases. My goal here is not to justify crawlers, or argue that the doctrines should unambiguously permit them. I understand that in certain situations, the arguments on the other side may be stronger. I merely seek to present a consent-based response to property-based views the court has presented on the topic and try to even out the rhetoric.

1. The server owner expressly notifies the crawler operator not to crawl the site.

One might argue that the server owner is being inconsistent here. She protests that she has no desire to exchange data with the crawler, then goes ahead and forms a contract agreeing to just such an exchange. Whose mental state ought to control here: the big sign on the door saying "No Shirt, No Shoes, No Service" or the employees' policy of allowing barefoot beachcombers? It seems unfairly anthropocentric to argue that prior intent of the owner ought to override the contemporaneous assent that she allows her machine to manifest. All the concerned parties know that the relevant transactions are taking place via electronic agents; they should thus bear some responsibility for at least attempting to assure that those agents act in accordance with their wishes. The crawler does nothing wrong in attempting to form a contract. The server, or its firewall, ought to have the good sense to say "no."

2. The website contains human-readable terms expressing a crawler exclusion policy.

In addition to the inconsistency issue expressed above, this strategy has the problem of presuming that an actual human will read the terms of the site before a crawler visits it. This is a very poor assumption. Google, one of the most popular Internet search engines, claims to index over three billion web pages. (183) Comprehensive web indices such as Google would be impossible if human intervention were required for every page; crawlers are necessary to accomplish the task. (184) Expecting a human to read the terms on every page is unrealistic. This is exactly the kind of scenario that justifies UCITA's concern that requiring human oversight in every transaction would be "inefficient." (185) Website proprietors are on notice that the relevant transactions will take place without human intervention. There is no reason to override the TCP software's manifestation of assent simply because the website contains natural language text that nobody will read.

3. The website contains machine-readable terms expressing a crawler exclusion policy.

Unlike the previous strategy, this one acknowledges the machine-based nature of the transaction and attempts to cope. The most common methodology here is to include a special file on one's server called "robots.txt," accessible via HTTP. (186) This file contains information about which areas of the website should not be available to unauthorized automatons. (187) The eBay court referred to this technique as "robot exclusion headers." (188) Essentially, one might view this as a standard running atop HTTP: The expectation is that the first HTTP request the crawler makes will be for this special file.

As a threshold matter, we should note that, unlike IP, TCP, or HTTP, robot exclusion headers are not an officially recognized standard. There is a process used by the Internet community for developing the standards governing the code layer. (189) IP, TCP, and HTTP have all taken significant steps to become recognized standards; (190) there is no similar consensus behind robot exclusion headers. The eBay court, perhaps unconsciously, notes the difference between protocols and simple "netiquette" when it refers to "programmers who wish to comply with the Robot Exclusion Standard," (191) acknowledging that the existence of programmers who do not wish to comply. The court would have been hard-pressed to find a similar group of TCP holdouts.

We cannot place enormous weight on compliance with contested "standards" that express the preferences of only one concerned party: server owners. Respecting them might be better viewed as a matter of politeness than of rigid protocol. Analogously, I'd much rather solicitors not call me at home during dinner--I tell them this every time they call--but, unfortunately, they seem not to respect this preference. Since the headers are not an uncontested Internet norm, it is unclear what legal significance they have.

Aside from their questionable status, there are still other problems with exclusion headers. One problem is temporal. Since HTTP must be used to access the robots.txt file, a TCP contract has already been formed, and an HTTP request and reply have already been exchanged by the time the crawler has a chance to read the file. The trespass to chattels theory is based upon the idea of unwanted use of server processing power. This seems somewhat inconsistent with a measure wherein the server accepts the TCP connection and uses it to send data informing the crawler to terminate the connection. If we truly believe in the inviolability of servers, shouldn't simply reading the robots.txt file be a trespass? If not, we must admit that consent exists, at least for the purposes of reading this file.

There is also an interesting form of the inconsistency problem that we saw in the context of express notification. Although the robots.txt file says "no," the HTTP software will nevertheless transmit the theoretically excluded files when the crawler requests them. The server owner might point out that, within the context of a single connection, she can't disambiguate between clients that are crawlers and those that are not: It only becomes obvious who's a crawler when one analyzes the network traffic as a whole to determine which machines might be the source of an unnatural number of requests. But if the server can't tell the difference between humans and crawlers within the context of the single connection it has agreed to, what's the harm (in a legal sense) in continuing that one, consenting connection? The exclusion headers attempt a transactional solution to a problem that only arises in a global context encompassing many transactions.

4. The crawler operator obfuscates the source IP address of its datagrams.

Recognizing the difficulty with solving a global problem in a transactional manner, a server might try to solve the problem globally, by implementing a firewall to block all messages from offending IP addresses. As I have noted previously, (192) firewalls are quite consistent with the consent-based view of the Internet. They allow the software to better reflect its owner's policy regarding when to manifest assent.

Unfortunately, there's a technological wrinkle here, as eBay illustrates. eBay appears to have attempted a firewall solution during its difficulties with Bidder's Edge. (193) However, Bidder's Edge was able to get past this firewall by utilizing third party "proxy servers," which had the effect of making requests from Bidder's Edge look the same as other web traffic. (194) As the court explained:

   Organizations often install "proxy server" software on their
   computers. Proxy server software acts as a focal point for outgoing
   Internet requests. Proxy servers conserve system resources by
   directing all outgoing and incoming data traffic through a
   centralized portal. Typically, organizations limit the use of their
   proxy servers to local users. However, some organizations, either as
   a public service or because of a failure to properly protect their
   proxy server through the use of a "firewall," allow their proxy
   servers to be accessed by remote users. Outgoing requests from
   remote users can be routed through such unprotected proxy servers
   and appear to originate from the proxy server. Incoming responses
   are then received by the proxy server and routed to the remote user.
   Information requests sent through such proxy servers cannot easily
   be traced back to the originating IP address and can be used to
   circumvent attempts to block queries from the originating IP
   address. Blocking queries from innocent third party proxy servers is
   both inefficient, because it creates an endless game of
   hide-and-seek, and potentially counterproductive, as it runs a
   substantial risk of blocking requests from legitimate, desirable
   users who use that proxy server. (195)

With Bidder's Edge using a proxy server for its requests, eBay could not disambiguate Bidder's Edge requests from non-Bidder's Edge requests, eBay was therefore faced with an undesirable choice: consent to the crawler transactions or refuse to consent to a lot of transactions that might or might not be from crawlers.

Bidder's Edge's actions might seem underhanded, but are they enough to overcome the TCP contract's express consent? One might be tempted to argue that a TCP contract obtained under such circumstances is not binding, perhaps applying the defense of misrepresentation (196) or that of unilateral mistake. (197) With the consent voided, the trespass claim would be restored.

But misrepresentation or mistake requires an assertion or belief "that is not in accord with the facts." (198) The use of a proxy server does not satisfy this criterion. A proxy server is simply a centralized portal that clients can utilize to make outgoing HTTP requests. The proxy server keeps a local "cache" of web-based resources that clients have recently requested from other servers. (199) The client establishes a TCP connection with the proxy server, and sends an HTTP request to the proxy server for the resource it wants from some other server (the origin server). (200) If the resource exists in the proxy server's local cache, the proxy server simply sends back an HTTP reply containing its local copy, without even accessing the origin server. (201) Otherwise, it makes a TCP connection with the origin server and uses HTTP to request the resource. (202) When it receives a response, it forwards this response back to the client, and keeps a copy in its cache for the next request. (203)

We can see how a third-party proxy server might be said to "conceal" the source of a request. Since the origin server sees the request as coming from the proxy server, it doesn't ever see the original client's IP address. But there's no misrepresentation or mistake--the origin server is knowingly making a TCP contract with a proxy server. What's happening is simply two separate transactions: The client transacts with the proxy server, and the proxy server transacts on the client's behalf with the origin server. There's no assertion or belief that isn't in accord with the facts.

Proxy servers aren't designed to conceal; they're designed to reduce Internet traffic. (204) Their existence means that origin servers will have to deal with fewer client requests, since many client requests will be handled by proxy servers distributed throughout the network. (205) It seems ironic to base a trespass to chattels claim for unwanted use of server resources on a scheme designed to conserve those same resources. If server owners like eBay really feel that the cost of not knowing the identity of the client outweighs the benefit of reduced overall traffic, they have a simple solution--refuse all requests from third-party proxy servers. (206) While it is true that this solution will likely exclude desirable clients as well as crawlers, it's hardly unfair to ask a business to perform this kind of cost/benefit analysis. Conducting business on the Internet means making decisions about how to deal with Internet conventions such as proxy servers. Maybe several owners have no unambiguously good choice here, but the courts have no business creating one for them.

Again, I think firewalls in general are a step in the right direction--servers ought to refuse connections they don't want to establish. But the server knows what it means to make a connection with a proxy server--it is consenting to use on behalf of an unknown client. Such consent negates a trespass claim. (207)

5. The crawler "'exceeded the scope of any such consent" it has been granted.

Finally, it makes no sense to say that a crawler "exceeded the scope of any such consent" it might have been granted. (208) The consent the server granted was a TCP contract to send and receive data; this is exactly what the crawler did. Even if one takes a global view encompassing all of the crawler's connections, it's impossible to get past the simple fact that each individual connection was explicitly permitted.

C. Consent and Copyright

The architecture of consent can also serve as a defense to certain types of intellectual property claims that arise in the context of Internet transactions. (209) Kelly provides an instructive example. (210) Recall Arriba's practice of "framing"--displaying one of Kelly's images in Arriba's own browser frame-and how the court found this to be a violation of Kelly's public display rights under copyright law. (211)

In order to be able to "frame" one of Kelly's images, Arriba had to obtain a copy of that image. One can imagine two ways in which they might have done this. The first way would be for Arriba to have previously requested the image via HTTP and to have saved a copy on its own server. Every time a user clicked on a thumbnail, Arriba would pop up a frame containing its local copy of the image. The second way would be to retrieve the image anew each time Arriba wanted to frame it. This would cause Arriba to initiate a new TCP connection and HTTP request each time a user wanted to frame the image. It is unclear from the text of the opinion which method Arriba employed. (212)

Under the consent-based model of Internet communications, the two methods have very different legal implications. Under the first, we would probably say the use exceeded the scope of the consent. While the TCP connection and HTTP request/response allowed Arriba's original download, nothing in the protocol added a term allowing Arriba to save the image and display it indiscriminately.

If, however, Arriba employed the second method, consent might arise as a defense to the copyright claim. When Kelly's server receives a request for one of its images, it is clearly within the scope of the transaction that the image would be displayed on the screen of some end-user--why else would someone request it? By agreeing to send the image, Kelly was agreeing to its limited display. By requesting the image anew each time a different end user wanted to see it, Arriba was making sure that Kelly consented to each viewing--clearly distinguishing this from the first method, in which no such verification was sought. While a single HTTP request is insufficient to indicate consent to indiscriminate display, it should be sufficient to indicate consent for a single display. Therefore, even if the framing did not fall under a fair use exception, it would nonetheless be explicitly permitted use.

Even if we were to accept the most comprehensive view of Internet copyright--the RAM copy doctrine consent still seems to provide a defense to certain limited uses. What is the point of responding to an HTTP request if not to allow a RAM copy to be made? Property proponents might argue that verbal notice or terms on the website negate the consent of the TCP contract, but there's no reason to be any more sanguine about those arguments than we were in the trespass cases. The inconsistency problem, the temporal problem, and the fact that the transactions are automated by machines (it's not as though some human operator manually requests the image every time a user clicks on the corresponding thumbnail) still exist in the copyright context.

While consent is only a limited defense in the context of intellectual property--it only allows use that might reasonably be inferred from the TCP contract (213)--it does provide ground on which to defend against unchecked expansion of Internet intellectual property rights.

D. Boundaries of Consent

Just as property rights such as trespass and RAM copy may seem to go too far in one direction, one might think that the architecture of consent swings the pendulum too far in the other direction. How might a server owner protect its legitimate property interests?

As the copyright example illustrates, it is important that we not take the consent-based view too far. By themselves, TCP contracts and their associated encapsulated protocols represent agreements circumscribed for the limited purposes of data exchange. They should not be read to completely usurp our notions of common law and intellectual property, which still have a place on the Internet. The architecture of consent supplements and informs these rights; it does not negate them.

It is true, however, that the boundaries of the consent doctrine might be greatly expanded by further development of Internet technology. Since the terms of TCP contracts are supplemented by the expectations of higher-level protocols, the true limits of consent exist only in the code itself. There is no precise legal terminus for the consent doctrine, no logical point at which we can say "no, she never agreed to that." In theory, we could imagine progressively more intricate protocols that might signal agreement to any number of things. Perhaps the idea that you can consent to a credit card purchase by encrypting your card number and sending it to the seller is only the beginning.

But the idea that the code itself defines consent can also cut the other way; technology can be developed to limit, as well as expand, the boundaries of consent. Nothing prevents server owners from adopting protocols that better express their optimal (and presumably lower) level of consent, eBay took a step in the right direction with its firewall. Perhaps it makes sense for eBay to stop transacting with third-party proxy servers. Servers might wish to begin requiring that users "authenticate" (214) their identities and use these verified credentials to "authorize" (215) only certain amounts of access. The simple act of requiring registration and passwords may be an effective way to limit many unwanted uses. (216)

The architecture of consent merely tells servers that they can't have their cake and eat it too---their choice to do business on the Internet has both positive and negative implications. Where it goes from here is largely up to them. They are free to propose and develop new protocols and see how well they catch on.

CONCLUSION

Choosing to "opt in" to the Internet as it now exists means agreeing to abide by certain norms of behavior. The architecture of the code layer enforces this norm legally by serving as a means for consensual communication. Although some form of property rights makes sense on the Internet, the architecture of consent shows us that either those rights don't extend as far as we might have thought, or that those rights are both implicitly and expressly waived by the use of Internet protocols.

I am not arguing that the Internet must be free, or even that it ought to be. I simply note that the current architecture of the Internet provides strong legal arguments favoring freedom. As the technical nature of Internet transactions shows, servers accept the implications of this architecture both implicitly and expressly. If change must come, it should not occur through the unthinking application of doctrines that happen to match our lay intuitions. If we are to apply doctrines such as trespass and copyright to the Internet, we must attempt a full understanding of the Internet as a technological phenomenon and make sure that we see all of its implications.

By focusing on imperfect analogies and only seeing the technical points that match our normative intuitions, we close our eyes to reality. The Internet is a dynamic, rapidly changing world, encompassing a wide set of relationships among numerous parties with varying interests. The relationships among the parties are just as important as their individual interests; it is important to understand both in order to make an informed decision on how we ought to apply the law.

(1.) See generally Thomas C. Grey, The Disintegration of Property, in NOMOS XXII: PROPERTY 69 (J. Roland Pennock & John W. Chapman eds., 1980) (discussing modern disaffection with the notion of absolute property and the implications of this doctrinal shift).

(2.) Webopedia.com, at http://www.webopedia.com/TERM/I/Internet.html (last visited Feb. 5, 2004) (listing results of searching the term "Internet").

(3.) Webopedia.com, at http://www.webopedia.com/TERM/P/peer_to_peer_architecture.html (last visited Feb. 5, 2004) (listing results of searching the term "peer to peer architecture").

(4.) Webopedia.com, at http://www.webopedia.com/TERM/C/client_server_architecture.html (last visited Feb. 5, 2004) (listing results of searching the term "client server architecture").

(5.) Webopedia.com, at http://www.webopedia.com/TERM/W/world_wide_web.html (last visited Feb. 5, 2004) (listing results of searching the term "World Wide Web").

(6.) Id.

(7.) Webopedia.com, at http://www.webopedia.com/TERM/B/browser.html (last visited Feb. 5, 2004) (listing results of searching the term "browser").

(8.) Webopedia.com, at http://www.webopedia.com/TERM/W/web_site.html (last visited Feb. 5, 2004) (listing results of searching the term "web site").

(9.) See Yochai Benkler, From Consumers to Users: Shifting the Deeper Structures of Regulation Toward Sustainable Commons and User Access, 52 FED. COMM. L.J. 561, 56263 (2000). Lawrence Lessig is probably most responsible for popularizing this conception. See LAWRENCE LESSIG, THE FUTURE OF IDEAS 23 (2001) [hereinafter LESSIG, THE FUTURE OF IDEAS]. He credits the idea to Benkler. Id. at 23 n.12.

(10.) See Maureen A. O'Rourke, Property Rights and Competition on the Internet." In Search of an Appropriate Analogy, 16 BERKELEY TECH. L.J. 561, 564 (2001).

(11.) As of August 2000, over eighty percent of households with a computer also had Internet access. NAT'L TELECOMM. & INFO. ADMIN., U.S. DEP'T OF COMMERCE, FALLING THROUGH THE NET: TOWARDS DIGITAL INCLUSION 2 (2000), available at http://search.ntia.doc.gov/pd/fttn00.pdf (last visited Feb. 5, 2004).

(12.) See O'Rourke, supra note 10, at 586-87; see also Alfred C. Yen, Western Frontier or Feudal Society?: Metaphors and Perceptions of Cyberspace, 17 BERKELEY TECH. L.J. 1207, 1215 (2002).

(13.) See O'Rourke, supra note 10, at 586-87. For an excellent discussion of the historical growth of the notion of the Internet as a "place," see Dan Hunter, Cyberspace as Place and the Tragedy of the Digital Anticommons, 91 CAL. L. REV. 439, 446-58 (2003).

(14.) There is something of a chicken-and-egg problem with listing this as an "intuitive" reason for governing the Internet through property, in that the rise of the Internet and associated technologies may have itself given rise to an increased interest in protecting intellectual property, rather than vice versa. Lawrence Lessig, among others, has shown how the panic among content owners at the web's increase in the unprotected exposure of their material has led to the creation of various legal and software-based protection mechanisms. See LESSIG, THE FUTURE OF IDEAS, supra note 9, at 177-217.

(15.) See O'Rourke, supra note 10, at 581-82.

(16.) A "chattel" is defined as "movable or transferable property; esp., personal property." BLACK'S LAW DICTIONARY 229 (7th ed. 1999). For a further discussion of the chattels view of the Internet, see R. Clifton Merrell, Trespass to Chattels in the Age of the Internet, 80 WASH. U. L.Q. 675, 679-84 (2002); see also Ticketmaster Corp. v. Tickets.com, Inc., No. CV99-7654-HLH, 2000 U.S. Dist. LEXIS 12987, at * 15-16 (C.D. Cal. Aug. 10, 2000) ("The computer is a piece of tangible personal property. It is operated by mysterious electronic impulses which did not exist when the law of trespass to chattels was developed, but the principles should not be too different."), aff'd, 2001 U.S. App. LEXIS 1454 (9th Cir. Jan. 22, 2001).

(17.) See, e.g., Trotter Hardy, The Ancient Doctrine of Trespass to Web Sites, 1996 J. ONLINE L., art. 7, [paragraph] [paragraph] 19-22 (1996) (discussing how the definition of trespass in the RESTATEMENT (SECOND) OF TORTS [section] 158 (1965) might apply in the Internet context); Hunter, supra note 13, at 446-58 (noting the intuitive strength of the place metaphor); id. at 472-500 (documenting ways in which the place metaphor has influenced the development of law on the Internet).

(18.) For a more complete discussion of these different areas of law and how they might apply to websites, see O'Rourke, supra note 10, at 580-95.

(19.) Crawlers are alternatively known as "spiders." For a definition, see Webopedia.com, at http://www.webopedia.com/TERM/C/crawler.html (last visited Feb. 5, 2004) (listing results of searching the term "crawler"); Webopedia.com, at http://www.webopedia.com/TERM/S/spider.html) (last visited Feb. 5, 2004) (listing results of searching the term "spider").

(20.) Webopedia.com, at http://www.webopedia.com/TERM/S/spider.html) (last visited Feb. 5, 2004) (listing results of searching the term "spider").

(21.) 100 F. Supp. 2d 1058 (N.D. Cal. 2000).

(22.) Id. at 1060-62.

(23.) Id. at 1063.

(24.) Id. at 1069.

(25.) Id. at 1072. This claim was dismissed, for reasons the record does not make clear. See id.

(26.) 100 F. Supp. 2d 1058, 1069 (N.D. Cal. 2000).

(27.) Id.

(28.) The traditional formulation of real property trespass states that:

   One is subject to liability to another for trespass, irrespective
   of whether he thereby causes harm to any legally protected interest
   of the other, if he intentionally (a) enters land in the possession
   of the other, or causes a thing or a third person to do so, or (b)
   remains on the land, or (c) fails to remove from the land a thing
   which he is under a duty to remove.

RESTATEMENT (SECOND) OF TORTS [section] 158 (1965).

(29.) "A trespass to a chattel may be committed by intentionally (a) dispossessing another of the chattel, or (b) using or intermeddling with a chattel in the possession of another." Id. [section] 217. See also id. [section] 218:

   One who commits a trespass to a chattel is subject to liability to
   the possessor of the chattel if, but only if, (a) he dispossesses
   the other of the chattel, or (b) the chattel is impaired as to its
   condition, quality, or value, or (c) the possessor is deprived of
   the use of the chattel for a substantial time, or (d) bodily harm is
   caused to the possessor, or harm is caused to some person or thing
   in which the possessor has a legally protected interest.

(30.) eBay, 100 F. Supp. 2d at 1065-66.

(31.) Id. at 1066; see also Hunter, supra note 13, at 484.

(32.) eBay, 100 F. Supp. 2d at 1069. The traditional formulation of trespass to chattels requires the element of physical contact. See Merrell, supra note 16, at 678. The court's finding in eBay that the electronic impulses inherent to any Internet activity constitute physical contact essentially renders this element a nullity in the context of internet transactions. See id. at 679-85; cf. eBay, 100 F. Supp. 2d at 1069 (citing Thrifty-Tel, Inc. v. Bezenek, 46 Cal. App. 4th 1559, 1566 n.6 (Cal. Ct. App. 1996) (noting that "the requirement of a tangible has been relaxed almost to the point of being discarded")). The court in eBay was therefore able to articulate a very broad view of what would constitute trespass to chattels on the Internet: "(1) defendant intentionally and without authorization interfered with plaintiff's possessory interest in the computer system; and (2) defendant's unauthorized use proximately resulted in damage to plaintiff." eBay, 100 F. Supp. 2d at 1069-70.

(33.) eBay, 100 F. Supp. 2d at 1069-72.

(34.) See Oyster Software, Inc. v. Forms Processing, Inc., No. C-00-0724JCS, 2001 U.S. Dist. LEXIS 22520, at * 36-41 (N.D. Cal. Dec. 6, 2001) (reading eBay to imply that any unwanted automated use of a website is enough to sustain a trespass action); Register.com, Inc. v. Verio, Inc., 126 F. Supp. 2d 238, 250-51 (S.D.N.Y. 2000) (applying eBay to support a trespass to chattels action for a crawler); Ticketmaster Corp. v. Tickets.com, Inc., No. CV997654-HLH, 2000 U.S. Dist. LEXIS 12987, at *14-18 (C.D. Cal. Aug. 10, 2000) (finding eBay analysis persuasive but holding that the crawler in the instant case did not cause sufficient harm to justify a trespass action), aff'd, 2001 U.S. App. LEXIS 1454 (9th Cir. Jan. 22, 2001). Even though it eventually held in favor of the defendant on the issue of harm in the trespass claim, the court in Ticketmaster was quite laudatory of eBay's "original and resourceful thinking" and "agree[d] with much of what Judge Whyte sa[id]." Id. at * 15.

(35.) See Specht v. Netscape Communications Corp., 306 F.3d 17, 47 n.19 (2d Cir. 2002).

(36.) See Intel Corp. v. Hamidi, 71 P.3d 296, 305-07 (Cal. 2003). This ease involved a trespass to chattels suit by Intel against a former employee for sending mass emails to its current employees. Id. at 299. Despite the court's favorable treatment of eBay, Hamidi was nevertheless able to avoid liability because "Intel does not claim the type of functional impact that spammers and robots have been alleged to cause." Id. at 306.

(37.)

   A state law cause of action is preempted by the Copyright Act if,
   (1) the rights asserted under state law are "equivalent" to those
   protected by the Copyright Act, and (2) the work involved falls
   within the "subject matter" of the Copyright Act as set forth in
   17 U.S.C. [subsections] 102 and 103.

eBay, 100 F. Supp. 2d at 1072 (citing Kodadek v. MTV Networks, Inc., 152 F.3d 1209, 1212 (9th Cir. 1998)); see also 17 U.S.C. [section] 301 (2003).

(38.) eBay, 100 F. Supp. 2d at 1072. Though Judge Whyte goes on to note disagreement with Ticketmaster's Judge Hupp, his arguments ultimately persuaded his colleague. See Ticketmaster, 2000 U.S. Dist. LEXIS 12987, at * 15.

(39.) See Merrell, supra note 16, at 696. Merrell argues that the court was incorrect to do this, since it effectively allowed protection of data that copyright had explicitly chosen not to cover, Id.

(40.) The chattels conception clearly corresponds to the physical layer of the Internet, and the intellectual property view clearly corresponds to the content layer. The real property metaphor arguably corresponds to the code layer, because that layer handles transportation. Since the court never had to address the issue of whether copyright would preempt a real property trespass claim, there is no way to know whether it would have seen as clear a dichotomy between the code and content layers as between the physical and content layers. Professor O'Rourke argues that preemption should apply in either case. See O'Rourke, supra note 10, at 590-95.

(41.) eBay, 100 F. Supp. 2d at 1065-66; see also Hunter, supra note 13, at 484 (observing that "[t]he court did not reject the extraordinary idea that Bidder's Edge's bots had invaded eBay's virtual space").

(42.) eBay, 100 F. Supp. 2d at 1065-66; see also Hunter, supra note 13, at 484.

(43.) eBay, 100 F. Supp. 2d at 1063.

(44.) Id. at 1066. But see Ticketmaster Corp. v. Tickets.com, Inc., No. CV99-7654-HLH, 2000 U.S. Dist. LEXIS 12987, at * 16-17 (C.D. Cal. Aug. 10, 2000) (finding that defendant crawler's comparatively small use of plaintiff's website did not constitute irreparable harm), aff'd, 2001 U.S. App. LEXIS 1454 (9th Cir. Jan. 22, 2001).

(45.) eBay, 100 F. Supp. 2d at 1066.

(46.) Compare id. at 1065-66 (analyzing irreparable harm), with id. at 1071 (analyzing the merits of the trespass action).

(47.) Id. The court's argument on this point has been read to support the proposition that any level of "use" by the defendant, no matter how minimal, is sufficient to support a trespass action. See Oyster Software, Inc. v. Forms Processing, Inc., No. C-00-0724JCS, 2001 U.S. Dist. LEXIS 22520, at * 40-41 (N.D. Cal. Dec. 6, 2001).

(48.) See RESTATEMENT (SECOND) OF TORTS [section] 158 (1965) ("One is subject to liability to another for [real property] trespass, irrespective of whether he thereby causes harm to any legally protected interest of the other....") (emphasis added); see also Merrell, supra note 16, at 689; O'Rourke, supra note 10, at 587.

(49.) For a fairly comprehensive view of copyright law applied to Internet linking, see Mark Sableman, Link Law Revisited: Internet Linking Law at Five Years, 16 BERKELEY TECH. L.J. 1273 (2001).

(50.) See Hardy, supra note 17, [paragraph] [paragraph] 7, 13-14; O'Rourke, supra note 10, at 590; R. Anthony Reese, The Public Display Right: The Copyright Act's Neglected Solution to the Controversy over RAM "Copies ", 2001 U. ILL. L. REV. 83, 138-43.

(51.) This nomenclature comes from Reese, supra note 50, at 84.

(52.) See Stenograph L.L.C. v. Bossard Assocs., Inc., 144 F.3d 96, 101-03 (D.C. Cir. 1998); MAI Sys. Corp. v. Peak Computer, Inc., 991 F.2d 511, 518 (9th Cir. 1993). For definitions of "fixed" and "copy," see 17 U.S.C. [section] 101 (2003).

(53.) See U.S. COPYRIGHT OFFICE, DMCA SECTION 104 REPORT 49-60 (2001), available at http://www.copyright.gov/reports/studies/dmca/sec-104-report-vol-l.pdf (last visited Feb. 5, 2004); Reese, supra note 50, at 139-40.

(54.) See Intellectual Reserve, Inc. v. Utah Lighthouse Ministry, Inc., 75 F. Supp. 2d 1290, 1294 (D. Utah 1999) (applying the RAM copy doctrine to find that a website owner might be liable for contributory infringement for linking to copyrighted works).

(55.) For a summary of several of these cases, see Perfect 10, Inc. v. Cybernet Ventures, Inc., 213 F. Supp. 2d 1146, 1167-68 (C.D. Cal. 2002).

(56.) See O'Rourke, supra note 10, at 590. But see Ticketmaster Corp. v. Tickets.com, Inc., No. CV99-7654-HLH, 2000 U.S. Dist. LEXIS 12987, at * 11-18 (C.D. Cal. Aug. 10, 2000) (accepting the RAM copy doctrine but still considering the viability of a trespass action), aff'd, 2001 U.S. App. LEXIS 1454 (9th Cir. Jan. 22, 2001).

(57.) See Stewart v. Abend, 495 U.S. 207, 228 (1990) (noting that "dissemination of creative works is a goal of the Copyright Act" and that "the Act creates a balance between the artist's right to control the work during the term of the copyright protection and the public's need for access to creative works").

(58.) See U.S. COPYRIGHT OFFICE, supra note 53, at 57-58. The fair use doctrine allows uses that might otherwise be prohibited if they are for "purposes such as criticism, comment, news reporting, teaching (including multiple copies for classroom use), scholarship, or research." 17 U.S.C. [section] 107 (2003). The court is asked to weigh several factors, including the nature of the use and the effect of the use on the market value of the work. Id.

(59.) See O'Rourke, supra note 10, at 584-85. As O'Rourke notes, the paradigm case here is Feist Publications, Inc. v. Rural Telephone Service Co., 499 U.S. 340 (1991). In Feist, the Court stated that "[c]opyright does not prevent subsequent users from copying from a prior author's work those constituent elements that are not original--for example ... facts, or materials in the public domain--as long as such use does not unfairly appropriate the author's original contributions." Id. at 350 (quoting Harper & Row Publishers, Inc. v. Nation Enters., 471 U.S. 539, 548 (1985)). The Court explicitly rejected the idea that property rights might derive simply from the "sweat of the brow" in assembling facts. 499 U.S. at 351-54.

(60.) Brief of Amici Curiae Mark A. Lemley and Twenty-Seven Other Professors of Law at 11-12, Bidder's Edge, Inc. v. eBay, Inc., 100 F. Supp. 2d 1058 (N.D. Cal. 2000) (No. 00-15995), available at http://www.law.berkeley.edu/institutes/bclt/pubs/lemley/bedgeami.pdf (last visited Oct. 5, 2003) (pleading by law professors in support of Bidder's Edge, Inc.'s appeal of the injunction) [hereinafter Law Professors' Brief]. Two weeks before Bidder's Edge's appeal would have been heard, Bidder's Edge went out of business and the case was settled. Interview with Mark. A. Lemley, Professor of Law, U.C. Berkeley, in Stanford, Cal. (Nov. 4, 2003).

(61.) Ticketmaster, 2000 U.S. Dist. LEXIS 12987, at *4-9.

(62.) Id. at * 12-14. The court went on to analyze the action under trespass. See supra note 34.

(63.) Kelly v. Arriba Soft Corp., 280 F.3d 934, 937-38 (9th Cir. 2002), withdrawn, 336 F.3d 811 (9th Cir. 2003). Since the court later withdrew the Kelly opinion, it has no precedential value. It is nevertheless instructive as to how a court might consider the issues presented.

(64.) Id. at 938.

(65.) Id. at 940 (finding the thumbnails constituted prima facie infringement); id. at 947 (finding that the framing infringed on Kelly's display rights).

(66.) Id. at 948.

(67.) Id. at 940-45. Unlike Ticketmaster and eBay, Kelly did not involve a trespass claim; there is no evidence that Kelly advanced such a theory, even at the district court level. See Kelly v. Arriba Soft Corp., 77 F. Supp. 2d 1116 (C.D. Cal. 1999), withdrawn, 336 F.3d 811 (9th Cir. 2003). Had he done so, he might have been successful under eBay. Even if Arriba were no longer crawling his site, it still was apparently accessing the site without permission when it framed his images. The court might, however, have failed to consider the trespass to show sufficient harm.

(68.) Congress has proposed bills that would create a sui generis protection of database information, which would by statute create the "sweat of the brow" right rejected in Feist. See Merrell, supra note 16, at 684-87.

(69.) See David N. Weiskopf, The Risks of Copyright Infringement on the Internet: A Practitioner's Guide, 33 U.S.F.L. REV. 1, 39-40 nn.199-203 (1998) (citing several cases where the fair use doctrine was held not to apply to online service providers and noting that the courts in those cases found every single fair use factor to weigh against the alleged infringer).

(70.) JOSEPH WILLIAM SINGER, PROPERTY LAW: RULES, POLICIES, AND PRACTICES 1244 (3d ed. 2002).

(71.) Cf. Kelly, 280 F.3d at 940.

(72.) See Wieskopf, supra note 69, at 41 ("Fair use does not offer clients guaranteed protection from claims of copyright infringement in any circumstances, let alone as applied to online service providers."); see also U.S. COPYRIGHT OFFICE, supra note 53, at 58.

(73.) Cf. LESSIG, THE FUTURE OF IDEAS, supra note 9, at 180-217 (analyzing the expansion of protections afforded content owners under intellectual property law).

(74.) See supra notes 10-18 and accompanying text.

(75.) See Hardy, supra note 17, [paragraph] [paragraph] 24-29.

(76.) See id. [paragraph] [paragraph] 49-52.

(77.) See eBay, Inc. v. Bidder's Edge, Inc., 100 F. Supp. 2d 1058, 1071 (N.D. Cal. 2000).

(78.) See Jane C. Ginsburg, Copyright and Control over New Technologies of Dissemination, 101 COLUM. L. REV. 1613, 1613 (2001); cf. U.S. CONST. art. I, [section] 8, cl. 8 (codifying the idea that it would "promote the Progress of Science" to give limited property rights to "Authors and Inventors").

(79.) See Hardy, supra note 17, [paragraph] [paragraph] 36-37.

(80.) See generally Richard Warner, Border Disputes." Trespass to Chattels on the Internet, 47 VILL. L. REV. 117, 136-37 (2002) (discussing incentives to enter a market absent exclusion rights).

(81.) The "tragedy of the commons" idea was originated by Garrett Hardin. Lawrence Lessig provides a one-page summary. See LESSIG, THE FUTURE OF IDEAS, supra note 9, at 22 (citing Garrett Hardin, The Tragedy of the Commons, SCIENCE 162, 1243-44 (1968)).

(82.) See Hardy, supra note 17, [paragraph] [paragraph] 38-48. This idea supports the court's finding of irreparable harm in eBay. See eBay, 100 F. Supp. 2d at 1066.

(83.) See Brief of Amici Curiae Reed Elsevier, Inc., the National Association of Realtors, and the e-commerce Coalition at 11-23, Bidder's Edge, Inc. v. eBay, Inc., 100 F. Supp. 2d 1058 (N.D. Cal. 2000) (No. 00-15995) (supporting eBay on Bidder's Edge's appeal of the injunction, authored by University of Chicago professor Richard Epstein).

(84.) See id. eBay claimed to have successfully negotiated such licenses with other commercial entities, Id. Indeed, it attempted to negotiate terms with Bidder's Edge as well; the breakdown of these negotiations started the sequence of events that resulted in the litigation. See eBay, 100 F. Supp. 2d at 1062.

(85.) See LESSIG, THE FUTURE OF IDEAS, supra note 9, at 105-10.

(86.) See Laura Quilter, The Continuing Expansion of Cyberspace Trespass to Chattels, 17 BERKELEY TECH. L.J. 421,438-39 (2002).

(87.) See, e.g., Dan L. Burk, The Trouble with Trespass, 4 J. SMALL & EMERGING BUS. L. 27, 32-38 (2000); Quilter, supra note 86, at 435-43.

(88.) See R. Polk Wagner, Information Wants to Be Free: Intellectual Property and the Mythologies of Control, 103 COLUM. L. REV. 995, 995 n.2 (2003) (collecting sources). Professor Wagner then proceeds to list numerous examples, Id.

(89.) See Law Professors' Brief, supra note 60, at 4.

(90.) See LESSIG, THE FUTURE OF IDEAS, supra note 9, at 146.

(91.) See Law Professors' Brief, supra note 60, at 6-8.

(92.) See, e.g., LESSIG, THE FUTURE OF IDEAS, supra note 9, at 15 ("Fueled by a bias in favor of control ... our social and political institutions are ratifying changes in the Internet that will ... reduce innovation on the Internet and in society generally.").

(93.) For example, property critics' slogan "Free the Mouse" seems targeted towards such a norm. See THE ERIC ELDRED ACT (2003), at http://eldred.cc (last visited Feb. 5, 2004).

(94.) See LESSIG, THE FUTURE OF IDEAS, supra note 9, at 35-36 (arguing that the free exchange of data, absent any centralized control, was one of the original goals of the Internet).

(95.) Id. at 85.

(96.) See O'Rourke, supra note 10, at 590; see also Burk, supra note 87, at 38-39 (discussing consent in the context of trespass actions for bulk emails); Merrell, supra note 16, at 693-94 (arguing that consent should be an allowable defense to trespass claims).

(97.) See O'Rourke, supra note 10, at 590.

(98.) Margaret Jane Radin, Humans, Computers, and Binding Commitment, 75 IND. L.J. 1125, 1141-42 (2000) (emphasis added).

(99.) Although not all of these arguments wound up being accepted by the courts, they demonstrate the range of arguments the courts considered as having possible bearing on consent.

(100.) See Register.com, Inc. v. Verio, Inc., 126 F. Supp. 2d 238, 249 (S.D.N.Y. 2000); eBay, Inc. v. Bidder's Edge, Inc., 100 F. Supp. 2d 1058, 1070 (N.D. Cal. 2000).

(101.) See Verio, 126 F. Supp. 2d at 249. Though the court did not find that Register.com's terms actually prohibited crawlers, had they done so, the court would have found Verio's crawler to be in breach of contract. Id.

(102.) See eBay, 100 F. Supp. 2d at 1061. The court found that these "robot exclusion headers" expressly prohibited automated uses of the site. Id. at 1070.

(103.) Id. at 1062-63.

(104.) Id. at 1070.

(105.) For instance, the specifications for TCP and IP were introduced in 1981. See INFO. SCIENCES INST., UNIV. OF S. CAL., INTERNET PROTOCOL: DARPA INTERNET PROGRAM PROTOCOL SPECIFICATION, RFC 791 (Jon Postel ed., 1981), available at http://www.faqs.org/rfcs/rfc791.html (last visited Feb. 5, 2004) (specifying IP) [hereinafter RFC 791]; INFO. SCIENCES INST., UNIV. OF S. CAL., TRANSMISSION CONTROL PROTOCOL: DARPA INTERNET PROGRAM PROTOCOL SPECIFICATION, RFC 793 (Jon Postel ed., 1981), available at http://www.faqs.org/rfcs/rfc793.html (last visited Feb. 5, 2004) (specifying TCP) [hereinafter RFC 793]. These protocols still form the backbone of the Internet today. See infra notes 125-39 and accompanying text.

(106.) LESSIG, THE FUTURE OF IDEAS, supra note 9, at 35 (emphasis omitted). For a more expansive argument of this viewpoint, see generally LAWRENCE LESSIG, CODE AND OTHER LAWS OF CYBERSPACE (1999).

(107.) See LESSIG, THE FUTURE OF IDEAS, supra note 9, at 23 n.13.

(108.) Webopedia.com, at http://www.webopedia.com/TERM/P/protocol_stack.html (last visited Feb. 5, 2004) (listing results of searching the term "protocol stack").

(109.) DOUGLAS E. COMER, THE INTERNET BOOK: EVERYTHING YOU NEED TO KNOW ABOUT COMPUTER NETWORKING AND HOW THE INTERNET WORKS 314 (2d ed. 1997) [hereinafter COMER, THE INTERNET BOOK].

(110.) Webopedia.com, at http://www.webopedia.com/TERM/P/protocol_stack.html (last visited Feb. 5, 2004) (listing results of searching the term "protocol stack"). For a more detailed description of the protocol stack and its different layers, see DOUGLAS E. COMER, INTERNETWORKING WITH TCP/IP VOLUME I: PRINCIPLES, PROTOCOLS, AND ARCHITECTURES 177-95 (4th ed. 2000) [hereinafter COMER, INTERNETWORKING].

(111.) This is not to imply that some group of network engineers sat around and planned out the Internet as we know it today on a whiteboard. The history of the Internet's development has been, and continues to be, quite organic. See generally Barry M. Leiner, Vinton G. Cerf, David D. Clark, Robert E. Kahn, Leonard Kleinrock, Daniel C. Lynch, Jon Postel, Larry G. Roberts & Stephen Wolff, A Brief History of the Internet, at http://www.isoc.org/internet/history/brief.shtml (Aug. 4, 2000) (last visited Feb. 5, 2004) (providing a short history of how the Internet was developed). For further information about how an Internet standard gets adopted, see S. Bradner, The Internet Standards Process--Revision 3, RFC 2026 (Oct. 1996), available at http://www.ietf.org/rfc/rfc2026.txt (last visited Feb. 5, 2004) [hereinafter The Internet Standards Process].

(112.) This is necessarily an abbreviated introduction to the protocol stack. In particular, it may make it seem like there's only one choice of protocol at each layer of the stack--this is not true. For example, there are protocols we might run instead of TCP were our goals somewhat different. See infra note 130.

(113.) For an introduction to local area networks, see JAMES F. KUROSE & KEITH W. ROSS, COMPUTER NETWORKING: A Top-DOWN APPROACH FEATURING THE INTERNET 392 (2001).

(114.) Kurose and Ross refer to such links as "broadcast links." Id. at 391-94.

(115.) Id. at 380.

(116.) W. RICHARD STEVENS, TCP/IP ILLUSTRATED, VOLUME I: THE PROTOCOLS 21-23 (1994) (describing RFC 894 encapsulation).

(117.) KUROSE & ROSS, supra note 113, at 380.

(118.) Id. at 409-11.

(119.) See id. at 415- 27 (describing the Ethernet protocol).

(120.) See id. at 441-47 (describing 802.11 LAN Architecture).

(121.) See id. at 447-53 (describing Point-to-Point Protocol (PPP)).

(122.) See id. at 391-94.

(123.) See id. at 410-11.

(124.) While this does not use the computer's central processing unit (CPU), it does use processing power on the computer's network adapter. See id. at 411.

(125.) See COMER, INTERNETWORKING, supra note 110, at 56-57. In the submarine example, each radio operator was a router connecting her submarine's PA network to the larger radio network.

(126.) See KUROSE & ROSS, supra note 113, at 300-10 (discussing IP addresses).

(127.) See COMER, INTERNETWORKING, supra note 110, at 118 (describing the difference between direct delivery at the link layer and indirect delivery using IP); id. at 119-21 (detailing "next-hop" indirect routing in IP).

(128.) Also, nonrouters implicitly consent not to pretend that they are routers, lest many messages get lost.

(129.) See COMER, INTERNETWORKING, supra note 110, at 121 ("We need to ensure that routers cooperate to guarantee that two-way communication is always possible.").

(130.) TCP is only one of several alternative protocols that utilize IP. For example, User Datagram Protocol (UDP) is a "connectionless" protocol that also runs on top of IP, utilized by programs that don't care too much if a datagram or two gets lost. See, e.g., UYLESS BLACK, INTERNET ARCHITECTURE: AN INTRODUCTION TO 1P PROTOCOLS 219-20 (2000). TCP is presented here because it is the protocol utilized in many web transactions. See infra notes 166-71 and accompanying text.

(131.) KUROSE & ROSS, supra note 113, at 210-11.

(132.) COMER, INTERNETWORKING, supra note 110, at 211-13.

(133.) See infra notes 169-74 and accompanying text.

(134.) Thus, my future references to a "sender" and a "receiver" within a TCP connection are contextual, rather than referring to fixed roles.

(135.) See KUROSE & ROSS, supra note 113, at 209-10.

(136.) See id. at 215-21.

(137.) COMER, INTERNETWORKING, supra note 110, at 210-13.

(138.) See id. at 237-38; KUROSE & ROSS, supra note 113, at 208, 226-28.

(139.) See COMER, INTERNETWORKING, supra note 110, at 239-40; KUROSE & ROSS, supra note 113, at 228-29.

(140.) RESTATEMENT (SECOND) OF CONTRACTS [section] 17(1) (1981).

(141.) Id. [section] 22(1).

(142.) One might also try to view the server's first reply as a counteroffer and the client's acknowledgement as acceptance. While this, too, is a workable view of the bargain, I think it best to view the client's final reply as simply acknowledging receipt of the acceptance. The server's "counteroffer" does not materially alter the details of the bargain--it simply establishes the proper sequencing for the reassembly of the datagrams. See supra text accompanying note 138.

(143.) RESTATEMENT (SECOND) OF CONTRACTS [section] 17 cmt. b (1981).

(144.) See, e.g., Carlill v. Carbolic Smoke Ball Co., 1 Q.B. 256 (C.A. 1893) (finding that a contract existed when the plaintiff performed on an offer laid out in the defendant's advertisement despite not having contacted the defendant). Since the two parties in Carlill never even met prior to contract formation, it is clear that haggling has not historically been a requirement for a contract's enforceability.

(145.) RESTATEMENT (SECOND) OF CONTRACTS [section] 71(1) (1981).

(146.) Id. [section] 71(2).

(147.) One might view the third message of the three-way handshake, where the client accepts the server's proposal of an initial sequence number, as reassuring the server that the acknowledgement and reassembly will indeed happen. Cf. KUROSE & ROSS, supra note 113, at 226-28.

(148.) "If the requirement of consideration is met, there is no additional requirement of (a) a gain, advantage, or benefit to the promisor, or a loss, disadvantage, or detriment to the promisee; or (b) equivalence in the values exchanged; or (c) 'mutuality of obligation'." RESTATEMENT (SECOND) OF CONTRACTS [section] 79 (1981); see also Hamer v. Sidway, 27 N.E. 256, 257 (N.Y. 1891) ("Courts will not ask whether the thing which forms the consideration does in fact benefit the promisee or a third party, or is of any substantial value to anyone.") (internal citations omitted).

(149.) This is the term adopted by the National Conference of Commissioners on Uniform State Laws. Uniform Commercial Information and Transactions Act [section] 102 cmt. 23 (Approved Official Draft 2002), available at http://www.law.upenn.edu/bll/ulc/ucita/ 2002final.pdf (last visited Feb. 5, 2004) [hereinafter UCITA]. There are several excellent articles dealing with this topic in more detail than there is space to present here. See Radin, supra note 98, at 1130-31, 1136-37, 1142-45; see also John P. Fischer, Computers as Agents: A Proposed Approach to Revised U.C.C. Article 2, 72 IND. L.J. 545 (1997); Donnie L. Kidd, Jr. & William H. Daughtrey, Jr., Adapting Contract Law to Accommodate Electronic Contracts." Overview and Suggestions, 26 RUTGERS COMPUTER & TECH. L.J. 215, 238-76 (2000); Raymond T. Nimmer, Electronic Contracting: Legal Issues, 14 J. MARSHALL J. COMPUTER & INFO. L. 211 (1996).

(150.) The UCITA, discussed infra, is not alone among uniform codes in expressing this view of electronic agents. The latest draft of the Uniform Commercial Code, Article 2 (covering the sale of goods) states "[a] contract may be formed by the interaction of electronic agents of the parties, even if no individual was aware of or reviewed the electronic agents' actions or the resulting terms and agreements." U.C.C. [section] 2-204(4)(a) (Draft of Proposed Amendments 2002) (emphasis added), available at http://www.law.upenn.edu/bll/ ulc/ucc2/annual2002.pdf (last visited Feb. 5, 2004). A comment makes clear the drafters' intent that "[t]his subsection is intended to negate any claim that lack of human intent, at the time of contract formation, prevents contract formation. When machines are involved, the requisite intention to contract flows from the programming and use of the machine." Id. [section] 2-204 cmt. 4 (emphasis added). This language was adopted from the Uniform Electronic Transactions Act (UETA). Id. Though Article 2 of the UCC governs only the sale of goods and thus, unlike the UCITA, would not apply directly to data exchanges, it is nevertheless further indication of the pervasiveness of the idea that an electronic transaction might be legally binding.

(151.) UCITA, supra note 149, Prefatory Note.

(152.) Grant Gross, UCITA Hits Snag with Lawyer Group, INFOWORLD, at www.infoworld.com/article/03/02/11/HNucita fails_1.html?standards (last visited Feb. 5, 2004).

(153.) See Specht v. Netscape Communications Corp., 306 F.3d 17, 34 n.17 (2d Cir. 2002) (noting controversy among scholars as to UCITA's characterization of consent).

(154.) See id. ("UCITA's provisions offer insight into the evolving online 'circumstances.'"); Rhone Poulenc Agro, S.A.v. DeKalb Genetics Corp., 284 F.3d 1323, 1331 (Fed. Cir. 2002) ("UCITA (pertaining to the licensing of intangible property) provides guidance on the U.C.C.'s view of the common law."), cert. denied, 123 S.Ct. 2668 (2003); AGT Int'l, Inc. v. Level 3 Communications, L.L.C., No. 02-CV684, 2002 U.S. Dist. LEXIS 21536, at *15 (S.D. Ohio July 29, 2002) (using UCITA to support a contention about industry custom); M.A. Mortenson Co. v. Timberline Software Corp., 998 P.2d 305, 313 n. 10 (Wash. 2000).

(155.) UCITA, supra note 149, [section] 107(d).

(156.) This statement is clearly true under a broad reading of the UCITA. There are, however, a couple of narrower alternative readings that deserve consideration.

The first alternative reading is that an agent has not been "selected for making an agreement" unless the human is on constructive notice that the actions of the agent will be considered legally binding. This reading cautions against any retroactive application of TCP-based consent to harm server owners who could not have known the legal implications of the actions taken by their TCP software before the new consent-based rule was announced. However, this reading would not require a court to grant a prospective injunction in favor of eBay. At most, it might justify a very short temporary injunction to give eBay time to get its affairs in order. After that time, the legal wrinkle presented by this reading would disappear since eBay (and other server owners) would now be on constructive notice regarding the legal implications of their TCP software.

The second alternative reading is that an agent has not been "selected for making an agreement" unless the human affirmatively wants the agent to form legally binding agreements. However, I think that such a reading would be unworkable. It presents at least two major problems: (a) The agent on the other side of the transaction has no way to distinguish between agents of humans who really desire contracts and those who do not, and (b) this is a difficult standard to administer after the fact because it requires evidence of subjective intent. ("Oh sure, I had the program go buy that stock, but I didn't want to be legally bound by that transaction.") For similar reasons, the traditional law of contract formation relies on objective manifestations of assent rather than the subjective intent of the parties. See 17A AM. JUR. 2D Contracts [section] 26 (2003) ("[I]t is not the subjective thing known as meeting of the minds, but the objective thing, manifestation of mutual assent, which is essential to the making of a contract."). The UCITA should be no exception to this. Assuming the human operator is on constructive notice regarding the legal implications of her software, it would be confusing, inconsistent, and unworkable to allow her to repudiate those legal implications based solely on her subjective intent.

(157.) "In electronic transactions, automated systems can send and react to messages without human intervention. A rule that demands human assent would add an inefficient and error prone element or inappropriately cede control to one party." UCITA, supra note 149, [section] 214 cmt. 2.

(158.) See supra text accompanying note 98.

(159.) Of course, programs may have unintended errors. The UCITA would apply the common law doctrine of mistake to cover this possibility. UCITA, supra note 149, [section] 111 cmt. 3. TCP connections, however, represent programmed behavior.

(160.) KUROSE & ROSS, supra note 113, at 653. Blocking datagrams originating from unwanted IP addresses would prevent these datagrams from reaching the TCP layer in the first place to form connections. See BLACK, supra note 130, at 288-91; KUROSE & ROSS, supra note 113, at 653-57.

(161.) This would be true even if the software owner does not have personal knowledge of how TCP connections, firewalls, and other Internet devices work. In such a case, the owner (who will either be an individual or a corporate entity) would have had to rely on an agent to set up the server, and it would be reasonable to impute the agent's constructive knowledge of these matters to the owner. Cf. RESTATEMENT (THIRD) OF AGENCY [section] 5.03 (Tentative Draft No. 4, 2003) ("For purposes of determining a principal's legal relations with third parties, notice of a fact that an agent knows or has reason to know is imputed to the principal if knowledge of the fact is material to the agent's duties to the principal, unless the agent (a) acts adversely to the principal as stated in [section] 5.04, or (b) is subject to a duty to another not to disclose the fact to the principal.").

(162.) RESTATEMENT (SECOND) Or CONTRACTS [section] 33(1) (1981).

(163.) Id. [section] 33(2).

(164.) Id. [section] 34(1).

(165.) Id. [section] 34(2).

(166.) Id. [section] 204.

(167.) Id. [section] 202(3).

(168.) Id. [section] 202(5).

(169.) See generally KUROSE & ROSS, supra note 113, at 72-83 (providing an overview of the application layer).

(170.) See id. at 85.

(171.) Id. at 85-87. There are two interesting details to note here. First of all, as we shall see, both the client and the server act as a sender and as a receiver within the TCP connection that encapsulates the HTTP transaction. Secondly, what I refer to here as an HTTP "message" might actually be comprised of multiple IP datagrams.

(172.) See id. at 87-88.

(173.) Id. at 85. For a definition of URLs, see Webopedia.com, at http://www.webopedia.com/TERM/U/URL.html (last visited Feb. 5, 2004) (listing results of searching the term "URL").

(174.) See KUROSE & ROSS, supra note 113, at 87-88, 92-94.

(175.) See generally id. at 106-24 (describing email protocols).

(176.) See, e.g., DON Box, DAVID EHNEBUSKE, GOPAL KAKIVAYA, ANDREW LAYMAN, NOAH MENDELSOHN, HENRIK FRYSTYK NIELSEN, SATISH THATTE & DAVE WINER, SIMPLE OBJECT ACCESS PROTOCOL (SOAP) 1.1 (2000), available at http://www.w3.org/TR/SOAP/ (describing "a lightweight protocol for exchange of information in a decentralized, distributed environment" that utilizes HTTP).

(177.) It might be possible to envision the HTTP request and reply as themselves establishing a contract: The client asks for data, implicitly agreeing to send it on to the requesting application, and the server accepts by performance. I would not necessarily object to such a view. However, I think conceiving of contract formation at the TCP layer is preferable for two reasons. First, it is the lowest layer at which we can identify express consent. It therefore functions as a useful construct without even requiring the existence of a higher-level protocol, by establishing the client's right to utilize server resources in the first place. A TCP contract thus serves as a useful template for a contractual relationship even when the higher level protocol is not HTTP and lacks anything resembling mutual assent. Secondly, the same TCP connection may be utilized for many HTTP requests and replies. See KUROSE & ROSS, supra note 113, at 89-90. It may be easier to view this as longer-term performance of a single contract.

(178.) "Every contract imposes upon each party a duty of good faith and fair dealing in its performance and its enforcement." RESTATEMENT (SECOND) OF CONTRACTS [section] 205 (1981).

(179.) See RESTATEMENT (SECOND) OF TORTS [section] 158 (1965) (defining the action of real property trespass).

(180.) Cf. Quilter, supra note 86, at 440-41 & n.145 (noting similar observations about telephone calls in the context of trespass to chattels, citing Burk, supra note 87, at 34, 35-36 (discussing undesirable implications of the trespass claim to other telecommunications technologies)).

(181.) See generally RESTATEMENT (THIRD) OF PROPERTY (SERVITUDES) [section] 2.1(2) (2000) (detailing various ways that a servitude might arise by operation of law).

(182.) Supra notes 99-104 and accompanying text.

(183.) As of March 26, 2003, the actual number was 3,083,324,652. Google Homepage, at http://www.google.com (last visited Mar. 26, 2003).

(184.) For example, the Yahoo! Directory is a human-built index of web pages that allows users to browse pages by subject. It does not, however, attempt to be comprehensive. See Yahoo! Help-Directory, at http://help.yahoo.com/help/us/dir/dir-03.html (last visited Feb. 5, 2004).

(185.) See UCITA, supra note 149, [section] 214 cmt. 2.

(186.) See MARTJIN KOSTER, A STANDARD FOR ROBOT EXCLUSION (1994), available at http://www.robotstxt.org/wc/norobots.html (last visited Feb. 5, 2004).

(187.) Id.

(188.) eBay, Inc. v. Bidder's Edge, Inc., 100 F. Supp. 2d 1058, 1061 (N.D. Cal. 2000).

(189.) See The Internet Standards Process, supra note 111.

(190.) See RFC 791, supra note 105 (specifiying IP); RFC 793, supra note 105 (specifying TCP); R. Fielding, J. Gettys, J. Mogul, H. Frystyk, L. Masinter, P. Leach & T. Berners-Lee, Hypertext Transfer Protocol--HTTP/1.1, RFC 2616 (June 1999), available at http://www.w3.org/Protocols/rfc2616/rfc2616.html (last visited Feb. 5, 2004) (specifying HTTP).

(191.) eBay, 100 F. Supp. 2d at 1061 (emphasis added). The court does, however, go on to accept eBay's assertion that many of the large search engines do respect the headers. Id. at 1063. Bidder's Edge, however, disputed this point. Id. at 1063 n.5.

(192.) See supra note 161 and accompanying text.

(193.) See eBay, 100 F. Supp. 2d at 1062-63.

(194.) Id. at 1061, 1062-63.

(195.) Id. at 1061 (citations omitted).

(196.) See RESTATEMENT (SECOND) OF CONTRACTS, [subsections] 159, 162, 164, 167 (1981).

(197.) See Id. [subsections] 151, 153.

(198.) Id. [subsections] 151, 159.

(199.) See KUROSE & ROSS, supra note 113, at 97.

(200.) Id. at 98.

(201.) Id.

(202.) Id.

(203.) Id.

(204.) Id. at 99.

(205.) Id.

(206.) It is clear from the eBay opinion that this was one of eBay's choices. See eBay, Inc. v. Bidder's Edge, Inc., 100 F. Supp. 2d 1058, 1061 (N.D. Cal. 2000).

(207.) Indeed, even without the consent doctrine, the use of proxy servers still undermines the basis of a trespass action. The electronic impulses that theoretically provide the basis for the trespass action come from the proxy server rather than the client.

(208.) eBay, 100 F. Supp. 2d at 1071.

(209.) For an analysis of traditional defenses to linking claims, see Sableman, supra note 49, at 1328-36.

(210.) See supra notes 63-67 and accompanying text.

(211.) Kelly v. Arriba Soft Corp., 280 F.3d 934, 938, 947 (9th Cir. 2002), withdrawn, 336 F.3d 811 (9th Cir. 2003).

(212.) The opinion simply notes that the image was "imported directly from the originating website." Id. at 934. While this seems to imply the use of the second method, I am reluctant to assume the court intended its language to be technically precise rather than simply rhetorically effective at distinguishing the impermissible framing from the permissible thumbnails.

(213.) It is important to note that consent would provide no defense for a contributory infringement claim against a site that linked to a second site that itself infringed on a third party's copyright, because the second site would have no authority to give the first site permission to use the content. For a discussion of such cases, see Sableman, supra note 49, at 1316-27.

(214.) Webopedia.com, at http://www.webopedia.com/TERM/A/authentication.html (last visited Feb. 5, 2004) (listing results of searching the term "authentication").

(215.) Webopedia.com, at http://www.webopedia.com/TERM/A/anthorization.html (last visited Feb. 5, 2004) (listing results of searching the term "authorization").

(216.) Indeed, Yahoo! now requires registrants to prove that they are human beings rather than crawlers. See Clive Thompson, The Third Annual Year in Ideas: Proving You're Human, N.Y. TIMES MAGAZINE, Dec. 14, 2003, at 89. This is accomplished by presenting the registrant with a "small picture of a distorted word, something that looks as if it had been written on cellophane and stretched out of shape" and requiring her to type the word correctly into a response box. Id. The idea behind this test is that while even small children can recognize distorted words, the most sophisticated computers still lack this ability. Id.

Eric J. Feigin, J.D. Candidate, Stanford Law School, 2005. There are many people whom the author would like to thank for their support and assistance. Margaret Jane Radin provided insightful comments on both the original idea and an early draft. Judy, David, and Matthew Feigin supplied copious amounts of both encouragement and feedback. The staff of the Stanford Law Review worked tirelessly to get this Note into shape. Last but certainly not least, this Note would not have been possible without the invaluable and innumerable contributions of the author's wife, Becky Weiss.