his own would usually cost more than obtaining one from the publisher), he would then be making a reproduction of the document in apparent violation of the copyright owner's exclusive right to "reproduce the copyrighted work in copies" (Act of 1976, section 106 (1)).

If a machine-readable copy is not made available by the publisher of a copyrighted document, an indexer would appear to be unable to use a computer in indexing that document unless he obtained permission from the publisher to make and use a machine-readable copy. To seek permission from a large number of individual publishers could be a very time-consuming and costly procedure, so much so perhaps as to discourage computer indexing of any large number of documents. Some persons interested in fostering the development and use of computers have suggested that in this situation, the making of a machine-readable copy and its input into the computer for the sole purpose of preparing an index should not be regarded as an infringement but should be treated as a fair use. They argue that, as long as the publisher does not offer such copies, making one for a use which is not itself an infringement would not injure the copyright owner in any way and would not displace the potential sale of a copy of the work. In fact, they say, the inclusion of the work in the index would create some demand for copies. Alternatively, some of the same persons suggest, the statute should provide for a compulsory license to make and use a machinereadable copy in situations of this character.

A.4.4.2.2 Abstracts in Data Bases. Bibliographic data bases may include, in addition to index headings and citations, abstracts of the contents of the cited documents. These abstracts aid the researcher in determining more precisely the relevance to his subject of the documents cited in connection with the pertinent index headings. Typically, the abstracts in a data base are similar to a table of contents in that they are brief identifying statements of the subjects dealt with in the document. Such abstracts of copyrighted works do not reproduce the substance of the work and would not be a substitute for the work in conveying the essential information to be derived from reading the document itself. Accordingly, it would seem that such abstracts, like indexes, may be made freely without regard to the copyright in the work.

On the other hand there are so-called "abstracts" that are really synopses or digests of the substance of the document, conveying that substance so fully that a researcher's need for the information in the document might be satisfied by his reading of the "abstract" alone. This kind of synoptic abstract would seem to constitute a derivative work under the definition in section 101 of the Act of 1976 reading in part:

"A 'derivative work' is a work based upon one or more preexisting works, such as (an abridgement (or) condensation ...

...

A person who makes an "abstract" amounting to a condensation of a copyrighted work infringes upon the exclusive right of the copyrightowner to "prepare derivative works based upon the copyrighted work" (Act of 1976, section 106 (2)).

It is evident that there will be difficulty in some borderline cases in determining whether a particular abstract would be considered a mere non-infringing identifier of the subjects covered in a document, or an infringing condensation of the document.

The author abstracts accompanying many copyrighted articles are often sufficiently full in themselves to be protected as a copyrighted component of the work, so that their unauthorized reproduction would infringe the copyright.

In sum, the compiler of a data base would risk being charged with copyright infringement if his data base included abstracts prepared by him that could be considered condensations of copyrighted works, or included author abstracts of some length.

A.4.4.3 Putting Copyrighted Data Bases into Computer Systems

A.4.4.3.1 Where Publishers Offer to Supply Machine-Readable Copies. As shown by the preceding examination of the operation of existing computerized information systems, machine-readable data bases are being produced by many of the publishers of the compiled indexes and abstracts making up the content of those data bases, and the computer systems obtain their data bases from the publishers. Under this established business practice, the rights of the system to use the data bases and supply information extracted from them to their subscribers, and the compensation to be paid to the publishers, are settled by the contracts between the parties. As such contracts become common, a standard pattern of terms and conditions, shaped by the industry needs and experience, can be expected to evolve. The recognized copyright problems that would otherwise be involved in the use of copyrighted data bases in computerized systems would generally be resolved by such contracts. Nor would these copyright problems arise in those instances where the computer systems are operated by the publishers themselves.

To be most effective, a bibliographic data base system should cover the literature in any particular field of information as comprehensively as possible. The rapid expansion of published information has been, and no doubt will continue to be, accompanied by a corresponding expansion in compiled indexes and abstracts. As computerized data base systems become more highly developed and more commonly used, the publishers of more of the printed compilations of bibliographic data will no doubt make them available in machine-readable form to meet the demand for their use in computerized systems. To the extent that this occurs, the copyright problems pertaining to the use of data bases in such systems will continue to be settled by contractual arrangements.

Where the publisher offers to supply a machine-readable copy of a copyrighted data base wanted by an operator for inclusion in his system, we suggest that the operator should be expected to obtain it from the publisher. For the operator to make his own machine-readable copy in that situation should constitute an infringement.

A.4.4.3.2 Where Publishers Do Not Offer Machine-Readable Copies. It may be supposed that instances will arise in the future when a large computerized information system, seeking comprehensive coverage of some field, will wish to include in its data bases certain copyrighted compilations of bibliographic data that have been published only in printed copies. No more than a few publishers would be involved at any particular time and the system operator could identify them readily. It would therefore seem reasonable in such cases to expect the system operator to deal directly with the individual publishers. The operator could ask the publisher to make and supply a machinereadable copy of the compilation for the operator's use under a contract, or, as an alternative, to grant permission to the system operator to make his own machine-readable copy for such use.

probable that one or the other of such requests would be acceded to by the publisher upon terms mutually agreed to.

But suppose further that the publisher refuses to accede to either request, or simply fails to respond to the system operator's inquiry. In light of the value for research of having comprehensive coverage in data base systems, there would seem to be a valid argument in favor of providing some kind of compulsory license to permit a system operator to make and use a machine-readable copy of a copyrighted compilation of data where the publisher refuses or fails to provide such a copy or to grant permission to the operator to make one for his own use, within a reasonable period of time after being requested to do so. Under the compulsory license, of course, the system operator would be required to pay equitable compensation to the publisher.

A.4.4.3.3 Where Third Persons Offer to Supply Machine-Readable Copies. A machine-readable copy of a copyrighted data base is not Tikely to be available to the operator of a computerized system from a source other than the publisher (or his agent). Publishers who supply machine-readable copies for use in such systems will normally not sell a copy to a system operator so as to give him ownership of it, but will lease it to him under an arrangement which expressly confines its use to that system and precludes its being made available to anyone else. This practice is necessary because of the so-called "first sale doctrine" which is well established in the copyright law. Under that doctrine, the copyright owner's control over the distribution of copies of his work ends, with respect to any particular copy, when he makes the first sale of that copy. The doctrine is reflected in section 109 (a) of the Copyright Act of 1976 which reads:

"... the owner of a particular copy or phonorecord lawfully
made under this title, or any person authorized by such
owner, is entitled, without the authority of the copyright
owner, to sell or otherwise dispose of the possession of that
copy of phonorecord."

How the "first sale doctrine" operates is best illustrated in the familiar setting of the sale of a copy of a book by the copyright owner. The purchaser of that copy becomes its owner. He is precluded by the copyright law from reproducing the work in other copies (either in its original form or in a derivative form) and from performing or displaying the work publicly (except as specially permitted by the copyright statute); but as the owner of the particular copy purchased, he is free to sell, lend, destroy, or otherwise dispose of that particular copy as he sees fit.

Machine-readable data bases have no use other than in computerized information systems, and the number of prospective customers for copies is limited. The publisher must therefore seek to prevent the system operator to whom he supplies a machine-readable copy from passing that copy on to another system operator. This is done by leasing copies under specified restrictions against allowing others to use them.

If leasing copies in this manner, rather than selling them, is known to be the usual practice, a system operator who is offered a machinereadable copy of a data base by another system operator, or by anyone other than the publisher, would have reason to be suspicious of the legitimacy of such offer. He would therefore be required to investigate the offeror's right to claim lawful ownership of the copy and to dispose of it, and he would subject himself to liability if he obtained the copy from an offeror who was acting in violation of the rights of the copyright owner.

Even assuming that a system operator could lawfully obtain a machinereadable data base for use in his system from someone other than the publisher, he would probably have little or nothing to gain from doing He would still need to input the data base into his system and to provide the output of material from the data base to the users of his system. It seems virtually certain that at some stage during these operations he would have to deal with the publisher to obtain a license for these uses of the data base. The terms of the license might well be much the same as if he had leased the data itself from the publisher.

A.4.4.3.4 Input of Data Base as Use Subject to Copyright. As we have observed above, in the usual case where the operator of a computerized information system obtains a machine-readable data base from the publisher, the copyright license he might need to use the data base in his system would no doubt be included in his lease agreement with the publisher. This would apparently be true also in the situation mentioned above where a system operator arranges with the publisher of a printed compilation of data to make his own machinereadable copy for use in his system.

There may be some special circumstances in which a system operator acquires a machine-readable copy of a copyrighted data base without having obtained a license for its use in his system. As an example of this unusual situation, we have mentioned above the possibility of an operator's acquiring a machine-readable data base from a person other than the publisher. The question would then arise as to whether the system operator should be required to obtain a license from the publisher before he puts the data base into his system or need only arrange thereafter to pay the publisher for output.

In the extended discussion of a similar question heretofore (in relation to full-text input of documents), it has generally been agreed that the copyright owner of works placed in and retrieved from computer systems should be entitled to compensaion for such use of his works. Differing views have been expressed, however, as to whether the copyright owner should be entitled to payment for input or only for output. The arguments advanced in the past discussion for free input have been concerned largely with the input of documents for experimental purposes during the developmental stages of computer systems, or for non-infringing purposes such as analyzing or indexing a work which do not entail any reproductive output of the work.