Issue No.01 - January/February (2004 vol.24)
pp: 7, 71-73
Published by the IEEE Computer Society
In the May-June 2002 issue of IEEE Micro, part 1 of "Challenging search engines under copyright law" appeared with the promise of a part 2 if the US Court of Appeals on the West Coast ever came to its senses and revised its opinion in Kelly v. Arriba Soft. The opinion condemned as illegal what it called "framing" and seemingly condemned as well all hyperlinking to which the linked site did not consent. The losing party, a company operating an image search engine then called Arriba Soft, seconded by numerous amici curiae (friends of the court), immediately filed papers asking the court to reconsider and refrain from destroying the Internet. Nothing happened for 16 months as the court considered what to do next. But the shoe has now finally dropped. The court issued a revised opinion, dodging the issue entirely on procedural grounds.
The Arriba Soft image search engine was much like the present Google image search engine ( www.google.com). If a user entered a search word or phrase, say "Gold Rush," the engine would go to its index file and present an array of thumbnail images having some relation to the Gold Rush. (Thumbnails are small images in GIF or JPEG format that have much lower resolution than the originals.) The idea is that a user can readily scan a page of images for use much more quickly and easily than wading through verbal descriptions. For an example, see http://images.google.com/images?q=%22gold+rush%22&ie=UTF-8&oe=UTF-8&hl=en&btnG=Google+Search. (This will provide you with an array of reduced-size versions of images related to the Gold Rush, one of the first of which is a book cover, which will be used below as an illustrative example.)
One of the many Web pages that Arriba's spider program sought out for images was that of Leslie A. Kelly. Unlike many other Web page proprietors, Kelly objected violently to having users access his site via an image search engine. His grievance, as refined after several years of litigation, was that Arriba deep- linked to individual photographic images on his Web page, using a thumbnail image to identify Arriba's link. For example, at the preceding URL, if you click on the cover image for the book The Gold Rush, your browser will recognize it as a hyperlink to http://images.google.com/imgres?imgurl=www.lizaketchum.com/graphics/books/book-gold-rush.jpg&imgrefurl=http://www.lizaketchum.com/bookshelf-history.htm&h=220&w=168&prev=/images%3Fq%3D%2522gold%2Brush%2522%26svnum%3D100%26hl%3Den%26lr%3Dlang_en%7Clang_fr%26ie%3DUTF-8%26oe%3DUTF-8%26safe%3Doff%26sa%3DG and will immediately change your screen display to that associated with the listed site.
In the new display, the upper half of a split screen is a slightly enlarged version of the same image. That image serves in turn as a deep link to a full-size version of the same image. (A deep link sends the user's browser directly to the deep-linked image, bypassing any other material with which the Web site proprietor had associated the image.) The lower half of the foregoing split screen is a framed version of the actual Web page on which the image appears. The URL of that Web site is http://www.lizaketchum.com/bookshelf-history.htm, and the image and text referring to it are about halfway down the framed page.
By deep linking, the Arriba search engine caused Arriba's users to bypass Kelly's home page and any commercial messages located there. By simply presenting the image in a separate browser window all by itself (like a pop-up window), the Arriba search engine separated Kelly's images (as they appeared to Arriba's users) from any adjacent text or other material on an inner page of the Web site—from what Kelly referred to as the "context" in which he chose to present his image. Each of Kelly's images was of a copyright-protected photograph that he had taken and used promotionally in his business. Presumably, the context (which Arriba bypassed) had something to do with what Kelly wanted to promote.
What the Courts Did
Kelly lost in the trial court—a US District Court in California—and appealed to the Ninth Circuit, the federal appellate court for California and several other Western US states. The court of appeals upheld the part of the district court's ruling that held the thumbnails to be a fair use of Kelly's images. The use was fair, both courts felt, because making thumbnails from Kelly's images (and thousands of others) made it possible to provide users with a new and useful function that Arriba could not otherwise provide. (Although neither opinion discussed this point in detail, it is clear that using thumbnail images, rather than words to describe available images, is a much faster and more efficient way to communicate. The verbal alternative is probably hopelessly tedious.) In addition, the reduced size of the thumbnails makes them noncompetitive with the full-size images; therefore, the thumbnails do not divert users who want the images from going to Kelly or his site for them (where they could download them freely, anyway).
The district court had also rejected Kelly's claim of copyright infringement of the full-size images, based on the links to them, but the court of appeals reversed that ruling in its first opinion.
The first court of appeals opinion was what caused great alarm in cyberspace circles, and it was the subject of my prior column. Briefly, the court of appeals said that Arriba was engaged in unlawful framing. This was the court's misnomer for deep-linking, especially when the Web page opened the linked subject matter in a new browser window (like a pop-up). The appeals court bought the argument of Kelly's lawyer that presenting the images in a new window, separated from the entire context in which Kelly had intended to present the images, amounted to an unauthorized public display of the images.
The court's reason for trying to put Arriba's actions into the public-display pigeonhole was that the other usual pigeonhole—reproduction of the copyright-protected photograph—didn't fit the situation, for reasons explained below. Copyright law does not protect every conceivable use of a work. It protects only those few listed in the statute. As US Supreme Court Justice Potter Stewart famously observed in explaining this principle, he didn't commit copyright infringement when he sang a copyrighted song in his bathtub. The statute does not extend to that use of a copyright-protected work.
The usual main categories of copyright infringement are unauthorized reproductions of copies of a work, preparation of derivative works, and distribution of copies of the work. In addition, there are public performance and public display.
A hyperlink does not constitute a reproduction of a copy. It is simply giving a browser an address or location where it can find the referenced item. Perhaps a hyperlink is a little more than that, because a link interacts with a browser to permit it to automatically transition to the referenced URL. Links to URLs are often analogized to cards in a library card catalog, but a link is somewhat more than that. It operates like a library card catalog that can lead a library user to the shelf and say, "There is the book you wanted." Still, a link makes no reproduction of a copy.
One argument for infringement—though not a very good one—posits that the link creator causes the user who clicks on a hyperlink to reproduce a copy of the copyright-protected content. That is, the user does this by making the link creator's computer load the Web page's HTML code into the video RAM of the user's computer, thus causing the temporary display of the content on the user's screen. But this is a poor argument, first, because the legislative history of the statute expressly said that Congress did not intend that a computer's transitory memory storage and display screen usage should be considered production of copies. Second, by placing the copyright-protected material at his Web site for anyone to view, the Web page proprietor and copyright owner has invited such use. Thus by implication he "licensed" or "authorized" the use; these are fancy legal terms for "You asked for it, buddy."
For the same reason, a hyperlink is also not a distribution of a copy. The link creator does not have any copy that it does anything with. The only person doing any distribution of something is the Web site proprietor (copyright owner).
A link is also not a public display of the copyright-protected work, because the Web page proprietor is the one who publicly displayed the work. For example, if a statue is put up in a public place, and I say to you, "Hey, look at that statue," am I publicly displaying the statue? How is that different from a hyperlink to Kelly's pictures? Does it make any difference if I invite you to look through a telescope at the same statue?
The initial opinion of the court of appeals seemingly bought the argument that Arriba's divorcing Kelly's pictures from the context into which he had chosen to place them turned Kelly's public display into Arriba's unauthorized public display. According to the court, Arriba was publicly displaying Kelly's pictures by providing a link that "imported" the pictures to where Arriba had "framed" them (that is, put them in a pop-up window in which the picture appeared after a user's click).
What did the court mean? For example, where is the "where" of the statement that Arriba provided a link that "imported" the pictures to where Arriba had "framed" them? There is no "where"; it's in virtual space, or it's just a figure of speech. A hyperlink does not physically import anything anywhere. It simply provides code that permits a browser to shift its screen display from that on which the link appears to that referenced in the link. The code and what it causes the screen to display are as follows for a sample passage from IEEE Micro's Web page ( http://www.computer.org/micro/):
The screen display reads "…Our new online manuscript submission service, including answers to any questions you might have about the submission process, is now available."
The HTML code reads "…Our new <a href=" http://cs-ieee.manuscriptcentral.com/"> online manuscript submission service</a>, including answers to any questions you might have about the submission process, is now available." (For a link where you click on an image rather than a word or phrase, the code is a little more complicated.)
At the outset, forget the court's nonsense about linking as framing. The court was apparently telling you that providing a hyperlink to an image file (such as a jpeg, gif, tiff, or bmp) is a public display of the encoded image because the link provider moves the image from Kelly's Web page to its own. That isn't what happened. Nothing moved. Arriba just caused Kelly's jpeg to appear to users in juxtaposition with Arriba's image search engine material.
That's not a reproduction, distribution, or public display by Arriba. Is it any of the categories in the copyright act? Well, it's none of those that the court or parties mention. But, arguably, this use is a different category than what the copyright act recognizes or something close to it. But let's permit that to wait a minute.
After the court of appeals issued its first opinion, Arriba and many organizations told the court it didn't have any idea of how the Web works, and that it was wreaking havoc in cyberspace. They asked it to reconsider or refer the case to the whole court of appeals (not just the three judges who were on the initial panel and were responsible for the initial opinion). The court asked for more briefs. Numerous amici complained loudly. The court dithered for 16 months trying to figure out what to do and then decided to punt.
The court withdrew the part of its opinion addressing the full-size images. It said that the parties had asked the district court only for a ruling on the thumbnails and hadn't said anything about the full-size pictures. Therefore, the district court should have never considered that issue, nor should the court of appeals have done so. The court of appeals then dumped the whole case back in the lap of the district court with no further explanation. In other words: "Tag, you're it, and we're going home quickly." What the district court will do now is anybody's guess.
WHAT CAN EVERYONE LEARN FROM ALL THIS?
In one sense, there's nothing to learn. The court of appeals just marched up the hill, embarrassed itself by demonstrating its unfamiliarity with Internet technology, and then marched down again.
But something else went unmentioned. The copyright law (section 106(2)) also prohibits the unauthorized "preparation of a derivative work," that is, one based on a preexisting work. An example is a motion picture version of a novel or a condensation "or any other form in which a work can be recast, transformed, or adapted." An editorial revision of an earlier work is also a derivative work. There is no requirement of a physical copy; the mere preparation of an unauthorized derivative work is an infringement.
For copyright protection, a candidate for derivative-work status must have some "originality." Therefore, a protectable derivative work must embody more than mere trivial variations, such as a wholly uncreative rendition of Mickey Mouse into a plastic toy. Machine translations, such as those that come from Google, are probably not original. Others have posed similar questions about the products of routing programs for printed circuits or gate arrays.
It is unresolved whether the same originality requirement applies to alterations made to a product purchased from the copyright owner or her licensee. Say that I buy your studio's videotape and splice commercials into its beginning. Or I buy a book from your publisher and stick my advertising message on the flyleaf or inside covers. Is that an infringement in the form of an unauthorized preparation of a derivative work? Or, does it make any difference how trivial the alteration?
Some really trivial changes to legitimately purchased objects have been held de minimis and thus are not infringements. Examples are painting a hobbyhorse (protected as a sculptural work) a different color, and sewing together to form a handbag some towels bearing copyrighted pictorial designs—when the products are purchased from the copyright owner or a licensee. There is a division of authority over whether cutting or tearing pictures out of a book and pasting them onto a tile is an infringement. Thus, the weight of authority holds that a legitimate possessor of a copyright-protected item can make minor changes to the item without becoming a copyright infringer by doing so; the legitimate possessor has, in effect, a license to make small changes. This issue might have implications for framing Web site material, or putting pop-ups over it.
Anyway, in some circumstances you could consider putting a pop-up window on top of a Web page's screen display the preparation of a derivative work. For example, a company called Gator and another called UWhen are in the business of persuading computer users to download "free" software containing contextual advertising programs. The company makes a deal with, say, Amazon to cause the program to put an Amazon pop-up on the screen display when somebody who has downloaded the program surfs the Barnes and Noble Web site. The pop-up says something like, "Get books cheaper from Amazon," or "Here's a 10 percent off coupon on any book from Amazon." The theory of contextual advertising is that it is more effective because it targets only people who have an interest in the advertised subject matter. (Why would anyone surf the Barnes and Noble Web site unless they were interested in buying a book?)
So, is putting a pop-up ad on top of the screen display of somebody's Web page too trivial an addition to count as preparation of a derivative work, or does that even matter for liability? A federal district court in Alexandria, Va., recently ruled on the issue, holding that UWhen did not commit copyright infringement, but the case probably will be appealed to the federal court of appeals in Richmond, Va. Michigan and New York courts have also chimed in. How does that fit in with what happened in the Arriba Soft case?
I'll address these and other issues in part 3 in the next issue of IEEE Micro.