The practice of "caching" files in conjunction with currently popular software for World Wide Web applications on the Internet illustrates the difficulties both of applying current copyright concepts to electronic transmissions and of crafting a privilege under copyright law that exempts legitimate caching from prima facie direct copyright infringement. [NOTE 4]
Uncertainty surrounding the scope of permissible file copying threatens the continued development of a unique communications resource, and exposes existing content and service providers to liability for acts that virtually all participants would, ex ante, deem legitimate. That is, even if one believed that caching involves a prima facie infringement of the Web page owner's exclusive right to produce "copies" of the Web page in question, it seems reasonable and appropriate to describe basic forms of caching (such as those described in Appendix A) as impliedly authorized by the content provider, inasmuch as they merely facilitate access to, and utilization of, the information placed on the Web by the content provider -- to assert, in other words, that the context in which these transactions take place, and the "custom and usage" of the relevant community, provide sufficient authorization for basic caching transactions.
Difficulties arise, however, in trying to distinguish privileged caching of this kind from acts of copying less clearly authorized, and in determining whether -- or to what extent -- the owner of a web page may expressly override such implied authorization and impose restrictions and conditions on such caching. For example, the principle that implied authorization is limited to ephemeral copying, i.e., transitory storage in RAM, is likely to prove too narrow a privilege, because many web browsers (including the current industry standard, Netscape) store copies of cached material even after the session is terminated and the browser is exited (so that the next time the browser is run, the cached files are available to it), and we see no principled reason to distinguish between caching procedures on this basis. [NOTE 5] On the other hand, simply extending the privilege to all cached copies made merely to facilitate retrieval of the same material in the future is likely to prove overbroad, because such a privilege could frustrate a content supplier's legitimate interest in restricting secondary copying in some ways. [NOTE 6] Ideally, the uncertainty regarding the respective rights of web page owners and those engaged in caching would be resolved by agreement. The capabilities of the technology are evolving so rapidly that no static description of rights and duties would do justice to the potential for the relationship between those who upload web pages and those who make them available to users to evolve. The most serious problems arise, however, when the owner of a web page seeks to impose the terms of an "agreement," or "license" to cache, including specific restrictions that might take many forms, by means of placing the terms and conditions of this "license" on the face of the page or otherwise seeking to communicate conditions by means of English (or other) text. The problems stem from the fact that caching results from automated algorithms and occurs at a scale that renders attention to idiosyncratic terms and conditions infeasible.
One way to avoid the problem of attempting to comply with terms and conditions on caching imposed by the owners of web pages, purportedly based on their ownership of the copyrighted status of the web pages coupled with the claim that caching constitutes the making of an infringing copy (absent an express or implied license), would be to argue that the owners of web pages do not have a right to impose such terms -- because (1) caching does not produce an infringing copy, (2) the imposition of such terms and conditions is an abuse of copyright or an impermissible restraint on alienation or antitrust offense, or (3) the posting of a web page constitutes the implied grant of an irrevocable license to employ caching for the purpose of enabling users to browse the page. Such arguments may be raised in defense against any claim of infringement. But the better way to avoid the problem would be to foster widespread agreement on acceptable practices or, better yet, encourage those posting web pages to EXPRESSLY assert any conditions they wish to impose on caching, using the technical means provided by standard protocols (and processable by standard server software).
In short, the ideal situation would be one in which the owner of a web page is permitted to impose restrictions on caching to the extent, but only to the extent, that such restrictions are capable of being communicated and implemented by means of the standard technical protocols that control automatic caching algorithms. New technology will likely emerge to allow page authors to set an expiration date for the page, and to allow server software to determine whether the original web page has been updated since the cached copy was made. Web page owners may well have an interest in conditioning caching, for example, on the ability to receive information regarding the number of hits to the cached page, or on the continued ability to implement password protections or to charge fees, or on the receipt of demographic information regarding users of the page. We are conducting inquiries to determine the extent of other means available to web page owners to use html to embed instructions regarding caching in web pages, which instructions might then be carried out by servers using the http protocol. As the protocols develop, there is every reason to believe that the community of web page authors and publishers, Internet service providers and users will be able to reach agreement on standard protocols that support reasonable restrictions desired by publishers but that do not interfere unduly with the ability of service providers to render efficient access to web pages to their users.
As these technical means develop, the remaining risk will be that individual web page owners who do not agree to use technical means to communicate and implement desired restrictions on caching will use litigation, and claims of copyright infringement, to hold up system operators. Resolving this problem, given the rather unique context of law-making in electronic communities, calls for some innovative action. Re-assessment and revision of the U.S. Copyright Act, whether via common-law or statutory development, is not likely to be a sufficient response in light of both the multi-jurisdictional characteristics of many of the affected transactions and the generally rather glacial pace of common-law and statutory revision. Particular cases could be litigated, but that takes time and has uncertain significance in an international environment. We could attempt to impose a requirement for express agreement to applicable terms and conditions on the part of all who post web pages, by means of agreements imposed on service providers as a condition of obtaining an Internet addresses -- but that seems infeasible at this time. We could try to establish a new doctrine of international law or foster a treaty regarding the intellectual property rights involved -- but that would be extremely difficult. And we could try to develop general agreement among responsible netizens regarding customary practices -- but that would not necessarily operate to override the express reservation of rights (assuming such rights exist) by web page owners.
Accordingly, the very best means that now appear available to reduce uncertainty regarding rights and duties applicable to web caching appears to be the strategy of building the technical means to assert and comply with any desired restrictions into standard protocols and server software. The easier it is for a web page owner to express desired restrictions in machine-readable format, as part of standard protocols that the server software is readily capable of interpreting and implementing, the harder it will be for (1) a web page owner to assert an additional right to impose restrictions by merely asserting such rights in any other way, and (2) for a system operator to claim a right not to comply with the desired restrictions. In turn, the availability of such means to implement reasonable restrictions will make it easier to assert that the "default" condition, in the absence of the utilization of these technical procedures, is that caching can be freely utilized. For example, if a web page owner could easily set a bit in the html source code to instruct the server not to cache the web page, then (1) the server could comply with that instruction (not likely to be sent by most page owners, since it would slow user access to the page), and (2) the web page owner can reasonably be required to use that technical means to impose the restriction and not to use the mere assertion of a copyright and/or precatory language on the face of the page as the means of asserting such a right.
As technical means develop to allow expression of and compliance with licensing restrictions, we can act to build consensus that all parties should be required to use such technical means, to the exclusion of other forms of asserting alleged legal rights. We will talk with leading copyright experts as well as the technical community to seek to develop consensus on that proposition. [NOTE 7]
***********************************************
Note 2. See MAI Sys. Corp., 991 F.2d 511 (9th Cir. 1993) and Advanced Computer Services, 845 F.Supp. 356 (E.D. Va. 1994).
Note 3. See Information Infrastructure Task Force, Intellectual Property and the National Information Infrastructure: A Preliminary Draft of the Report of the Working Group on Intellectual Property Rights (July 1994) (the "Green Paper") at 36-37.
Note 4. Caching, for this purpose, is defined to mean the copying of a web page, made incidental to the first access to the page, and storage of that copy for the purpose of speeding subsequent access; a brief overview of caching usage is set forth in Appendix A.
Note 5. That is, we are not persuaded that the distinction, central to U.S. copyright law, between "transitory" and "non-transitory" fixation is a useful guidepost in this context.
Note 6. For example, if a content provider were to attempt to impose restrictions on the content published on the Web -- charging a fee, for example, for access to certain material -- potential users might find it convenient to access the material on the intermediary's host computer (see Appendix A, Example 3), thus avoiding the fee. Or, Consumer might pay once for an access to Content Server, but then obtain another copy from cached version on the provider's host and make prohibited copies of that second cached copy, arguing that the restrictions applicable to the original accessed material do not apply to material accessed from the provider's host. Or caching might be used to circumvent any system on Content Server that records the number of "hits" from Consumers.
Note 7. With regard to mirroring, the situation is somewhat different. Because mirroring involves FTP sites, it typically applies to files that can be password protected -- so technical means already exist for the owner of the file to impose any desired restrictions. Having not done so, the owner can be said to have granted an implied license for copying without restriction. Some owners may argue that they did not contemplate wholesale copying for the purpose of providing a site from which files can be downloaded for a fee. But that argument is somewhat undermined by (1) the fact that the owner did not attempt to charge a fee and therefore loses no sought-after economic advantage, and (2) the likely fact that the ultimate number of end users who get access to the files is not significantly affected (or, at least, not necessarily affected in a manner adverse to that contemplate by the act of posting the file to an open FTP site) by the number of intermediate points from which such files may be downloaded. For the moment, we conclude that no new technical developments would increase the degree of certainty regarding the entitlement of a service provider to engage in mirroring and that the best arguments favor continuation of that practice.
Example 1. The simplest World Wide Web transaction involves a consumer (using some Web browser client software) and a Web server to which the consumer has access. The server may provide "pointers" to other servers (in which case we can refer to the server as a "Pointer Server"), or it may provide content that can be retrieved by the consumer ("Content Server") -- or, of course, it may be both a Pointer Server and a Content Server. For simplicity, assume that the World Wide Web consists of a single Pointer Server and a (large) number of Content Servers. Consumer initially establishes a connection with Pointer Server to locate material of interest, finds a pointer to an item on a particular Content Server, and clicks on this pointer. Typically, Pointer Server then transmits the selected pointer, in the form of a URL, to Consumer's computer, and the web browser software running on Consumer's computer establishes a connection with Content Server and retrieves the web file referenced by the URL. Consumer's web browser then caches the file thus retrieved; as a result, if Consumer requests it again, the cache copy can be loaded directly from RAM rather than requiring another transmission over the Internet.
Example 2. As a second example, suppose that Consumer's computer is connected through a Local Area Network that in turn is connected to the Internet through a firewall computer. While there are a variety of firewall configurations, one common one would cause the firewall to serve as a proxy for Consumer; most popular web browser applications have a proxy option to accommodate such a firewall role. The proxy function works as follows: when Consumer attempts to activate a pointer on Pointer Server, the request is initially transmitted to Firewall Computer, which duplicates the request and sends it to Pointer Server. Pointer Server returns the requested information (the URL pointer) to Firewall Computer, which then caches it and sends a copy along to Consumer. After Consumer's computer establishes the connection to Content Server, Content Server returns the requested file to Firewall Computer; Firewall Computer caches the file and sends a copy to Consumer's computer, which also caches it (as in example 1 above).
Example 3. As a third example, consider an information service provider such as Compuserve, Delphi, Prodigy, or America Online. The service provider might want to provide its subscribers with the most rapid access possible to World Wide Web documents. It would be natural for the service provider to design its client software to include a web browser function that would work somewhat differently than the web browser software described in Examples 1 and 2. The service provider's World Wide Web access system may provide for the caching of material retrieved through the web on the service provider's host, so that if another subscriber wish to retrieve the same material, the second subscriber could do so directly from the service provider's host computer, without the host having to go out and fetch the material again across the Internet.