David M. Levy and Catherine C. Marshall
Xerox Palo Alto Research Center, 3333 Coyote Hill Road, Palo Alto, CA 94304,
Are digital libraries necessarily archival collections of digital materials to be browsed and queried by information-seeking individuals? While this seems to be the prevailing view, we argue instead that the design of digital libraries must take into account a broader range of materials, technologies, and practices -- transient as well as permanent documents, fluid as well as fixed materials, paper as well as digital technologies, and collaborative as well as individual practices. To illustrate, we discuss the work practices of one particular community, information analysts, and conclude by exploring some implications for technology development raised by this broadened perspective.
Keywords: Digital library, documents, work practice, paper, collaboration.
What is a digital library? The answer to this question seems in many respects as self-evident as the answer to the riddle "what color was George Washington's white horse?" Digital libraries, it seems obvious, are collections of digital material. Or are they? What will happen to the masses of paper already in existence? Will they be converted to digital form, and, if not, how will newer digital materials be integrated with paper archives? And if paper is not about to disappear, how will its ongoing use be coordinated with the use of digital materials? These are just some of the questions that ought to be raised, we think, as a community forms to do research and development around digital libraries.
Our purpose is to make explicit, and to question, certain assumptions that underlie current digital library efforts. We will focus in particular on assumptions about the nature of collections digital libraries will house, the kinds of technologies that will be involved, and the nature of the work digital library users will perform. To do this, we start by providing a simple framework and some further distinctions, which will allow us to locate current digital library efforts and to indicate the larger territory beyond these efforts, which we believe is deserving of attention. We then illustrate these points with observations of the current work practices of one particular community of information consumers, a community of policy analysts. We conclude by discussing some of the implications of our analysis for future work on digital libraries.
2. A Framework
While there are many types of libraries -- including national, corporate, community and school libraries -- as well as significant individual differences among libraries of different types, all libraries are at minimum repositories for, or collections of, documents. We use the term "documents" in the broadest sense to denote all enduring communicative records, including paper materials, electronic files, videotapes and audiotapes. As a broad framework for approaching documents, we find it useful, following Yates , to distinguish three perspectives from which documents can be viewed: as artifacts, in relation to technology, and as embedded in work.
* Artifacts: This perspective focuses on documents themselves -- as physical and social artifacts.
* Technology: A second perspective is concerned with the technologies by which documents are created, manipulated, distributed, and so on. Pencil and paper, typewriters, staples, workstations, printers, text editing and graphics software, networks -- these are all technologies for creating and working with documents.
* Work: The third perspective brings people and the work they are doing into the picture. Because documents are ultimately social artifacts, they must be understood with respect to their use -- in relation to the human practices and institutions in which they are embedded.
One further crucial notion is that of genre [4, 13, 17] which helps to integrate these three perspectives. The central observation is that documents come to us not as isolated artifacts but as instances of recognizable social types or genres -- e.g. as novels, packing receipts, shopping lists, journal articles, and so on. These types, typically arising out of a particular constellation of technologies (e.g. the newspaper from press and paper), classify documents primarily according to their use. That is, each genre brings together a particular form with a particular set of functions or roles; indeed it allows us, to a large measure, to recognize the intended use and institutional role of a document from the form alone and thereby gives us a context for interpretation.
A library, then, is (among other things) a repository for certain genres of documents aimed at serving a range of practices. For any library, the genres available and the selection of items in that genre will be largely determined by the needs and practices of the constituency that library is serving. Compare the differences in orientation and collection between an elementary school library and a corporate library. The available technologies that support the collection, while of course critical, are in some sense secondary to the work they are enabling. Yet in talking about digital libraries, it is the technologies that are being considered first. Indeed, much of the current discussion is focussed on issues of media, protocols, and architectures -- on the technologies of creation, access, and distribution. Less attention has been given to considering the genres of documents that will be included in digital libraries. And even less still has been given to questions of work, to the uses to which library-based documents and technologies will be put.
3. Further Distinctions
Let's next consider four distinctions that are important to libraries. The first two, fixed/fluid and permanent/ transient, are concerned with documents; the third, paper/digital, is concerned with technology; and the fourth, individual/group, is concerned with people and their work practices.
3.1. Documents: fixed/fluid
Documents are those artifacts by which we fix or stabilize communicative intent. Through the use of certain "technologies of fixity"  we are able to inscribe meaningful marks in a surface in such a way that they can travel through space and time, allowing multiple people to see "the same thing." But to say that documents are fixed is not to say that they don't also change; all documents are fixed for certain periods of time and changing (fluid) at other times. Indeed, different document genres have different rhythms of fixity and fluidity (e.g. book vs. memo) -- different rates and patterns of change. Broadly speaking, documents need to change for one of two reasons: because the information they contain is out of date (e.g. flight departure times for the month of March have changed); or because their context of use or interpretation has changed (e.g. the service manual will be used in France, requiring translation).
3.2. Documents: permanent/transient
Documents also last, and are useful, for different lengths of time. At one end are documents we might think of as "permanent" because they surpass human lifetime (the Bible); while at the other end are transient or ephemeral documents (shopping lists) whose useful lifetime is small in relation to ours. Most documents fall somewhere in between.
These first two distinctions are often conflated: we take "fixed" and "permanent" to be synonymous. But there is a sense in which these distinctions represent independent dimensions. The fixed/fluid distinctions is about the pattern or rate of change of a document over some interval. How much change is it undergoing? For what periods of time is it fixed? When and why has it changed? The permanent/transient distinction is about the lifetime of a document -- about the length of interval over which we take it to exist. Does it endure for a short or a long period of time?
It is easy to see the sense in which these two dimensions are independent of each other. A document of brief existence or utility (a Post-it attached to someone's door) may be unchanged during its working lifetime; while a long-lasting document may undergo much change (a copy of Shakespeare annotated and passed on from generation to generation). Rate of change, in other words, is not the same thing as duration.
But there is also a sense in which rate of change and duration are dependent. When a document is changing over some period of time (a sequence of drafts of a paper, for example), we actually have a choice: we can take the document to have a fixed identity, in which case the document remains stable across changes and time; or we can choose to see those changes as constituting the creation of a new document, rather than a new version of the old. In the first case, we see a single document, changing but enduring over time t; in the second, we see two documents, enduring over time t1 and t2, where t = t1 + t2. Whether or not we take changes to a document to affect its identity will determine how we assess its duration; it is the interplay between documents and work practice that guides this interpretation.
The confusion that occurs when "fixed" and "permanent" are conflated comes about when "fixed" is used to refer to a document's identity rather than to its properties. Thus a "permanent" document, in this sense, is a document whose identity is maintained (fixed) for a long time. But for such a document, many of its (non-constitutive) properties may still change, and in this sense the document (its identity) is fixed while some of its properties are fluid.
3.3. Technology: paper/digital
A third critical distinction for libraries is that between paper and digital technologies. We can understand "paper" as shorthand for a constellation of technologies centered around crushed plant fibers as a markable surface. This constellation includes pens, typewriters, printing presses, photocopiers, paper clips, and vertical files. We can understand "digital" as shorthand for the document technologies the computer has enabled: text and graphics editors, raster screens, mice, spelling correctors, etc. Beyond this, there are also hybrid technologies, such as scanners and printers, which allow us to transmute material from one constellation to the other.
Today it is quite common for all three distinctions to be aligned according to the following equations (see for example ):
paper = fixed + permanent
digital = fluid + transient
That is, it is assumed that paper technologies produce documents that are fixed and permanent while digital technologies produce documents that are fluid (changing) and transient. But this is simply wrong, since all documents, regardless of medium, are fixed and fluid, as already noted.
A further assumption is that we are moving (or will eventually move) to a completely digital world, with little or no use for paper. While no one can actually predict future paper use, it seems unlikely that this material, so flexible, portable, and easily annotated, will simply go away. At minimum, paper will remain the medium of choice for reading so long as screen technologies don't adequately support this activity. Moreover, so long as paper continues to have certain advantages over digital materials, we can expect to see hybrid uses, mixing paper and digital forms as appropriate.
3.4. Work practice: individual/group
While it is possible to think of library use as primarily solitary work -- the lone individual searching for, then reading, a single document -- this idealization is at odds with observations of work practice (e.g. [7, 8, 12, 14, 15]). Libraries are meeting places, where collaborations can and do happen. And even information-seeking -- whether located in a physical library setting or conducted over the Internet -- is as often as not collaborative rather than solitary. People seek information by communicating with members of their communities -- they not only look for materials and specific answers, but for corroboration, new interpretations, and new methods of finding information. This suggests, among other things, that support for communication and collaboration is as important as support for information-seeking activities, and that, indeed, support for the former is needed to support the latter.
4. Inside (and Outside) the Digital Library
What do we mean by a digital library? This would seem to be a critical question for a community now forming around this topic. One way to get at it is to ask, what is (or will be) inside a digital library, and, by implication, what will remain outside? It is our sense that digital libraries are currently being viewed primarily as repositories or collections of relatively fixed, relatively permanent, digital documents to be browsed or searched by individuals. In saying this, we know that any such blanket statement is bound to be incomplete, and will inaccurately characterize some efforts, but we intend it as a starting point for further discussion within the community. In the rest of this section, we examine these assumptions, and in the next, we present an alternative, use-oriented perspective.
4.1. Fixed and Permanent?
Libraries, traditionally, have been viewed as the repositories of our sacred heritage as preserved and transmitted primarily in books. Bolter , for example, talks about the close, mutually supportive relationship between libraries and the canon of great books. Books are by-and-large one of the more fixed (least quickly changing), more permanent (long enduring) types of documents. It is not surprising, then, that notions of fixity and permanence should be associated with libraries and their collections.
This assumption seems to have been carried fairly directly into considerations of the digital library. Many proposed, or actual, projects are oriented toward the management of relatively fixed, relatively permanent collections (e.g. Cornell's Class project  or ARPA's CSTR program ). Yet this assumption deserves to be questioned. Certainly, there is nothing in the nature of digital technologies which argues that only rarely changing, long-lasting documents should be candidates for insider status in digital libraries. On the contrary, the ease of modification afforded by digital technologies means that there are, and will be, whole classes of digital document genres that do not fit the traditional profile for library inclusion. To what extent do we want to consider collections of listserv messages, wire service articles, preprints, and other quickly changing and/or ephemeral documents as appropriate materials for digital libraries?
What color was George Washington's white horse? It seems tautological that digital libraries are about collections of digital documents. In its strongest form, this means that the digital library community is only interested in, and will focus solely on, digital documents. A weaker interpretation is that the primary focus will be on digital forms; that paper may be a transitional medium needing temporary support, or that it may continue to exist into the indefinite future in library contexts but in a marginal and uninteresting way. All versions of this assumption are deserving of further consideration. As noted above, we don't believe that paper is on the verge of being eliminated, or even marginalized, and if this is true, it raises serious questions about the feasibility of taking an all digital approach. One might agree with this, however, and still believe that a segregationist approach would work, imagining digital libraries for our digital needs and traditional libraries for our paper needs. A more appropriate approach would involve a fuller integration of materials, including those outside computational reach.
Either implicitly or explicitly, much of the current work on digital libraries assumes a simple model of use: the lone researcher sitting at a workstation, browsing, scanning, searching and retrieving. But this is at odds with our observations of work practice, to be discussed in the next section. For all work, including library use, takes place within communities brought together by shared interests and culture, whether they are traditional or more ephemeral, distributed, networked, digital communities.
5. A Use-Oriented Perspective on Digital Libraries
There are two ways to depict the expanded conception of digital libraries we are arguing for; these two pictures differ in where the boundary is drawn between inside and outside. If we think of digital libraries, in the narrow sense, as repositories of relatively fixed, relatively permanent documents of broad applicability, then we are arguing that the design of such libraries must explicitly take into account the creation and use of additional, short-lived, locally-available documents that supplement a community's use of library materials. If, however, we take a broader view of the library, one that includes collections of ephemeral, locally-produced and maintained documents, then we are suggesting that the design of digital collections must explicitly take into account a broader range of library materials, including non-digital materials. The first account treats the use of additional materials as a ring of activity just outside the digital library, the second places it within the inner circle.
Our understanding of the need for such a broadened approach comes from our work in studying the practices of information analysts [10, 11], as well as that of others [2, 5, 16]. Information analysts are people whose job it is to make sense of complex situations in the world (e.g. the development of a National Information Infrastructure) and to write descriptions, explanations, predictions, and prescriptions as a means of communicating their understanding to others . Analysts are the perfect foil for our examination of the assumptions underlying digital libraries because they are the perfect information consumers: their work is research-intensive; they must comb many sources to get the desired degree of coverage; they are required to have both a broad, contextual understanding of a subject area, and a narrow, very focused view of the characteristics of a particular problem; and they are always pragmatic users of a variety of distributed information resources, including reports, information services, newswires, journals, books, and each other's expertise.
We take three different perspectives on the analysts' work: (1) the artifacts they use and produce -- how analysts acquire the background and details they need to address seemingly ill-defined reporting requirements and how their reports feed back into the community's pool of shared resources; (2) the interpretations they note in the course of analysis -- how analysts bridge the gap between reading source materials and writing reports; and (3) the collaborations that occur -- how a task that is often perceived as solo is really an informal group effort, dependent on a complex social structure.
5.1. Sources and reports: the artifactual bases for analysis
Analysts are voracious, pragmatic readers; they are as a rule concerned with keeping up on the literature in their areas of specialization. In addition to using traditional library materials like books and journals in their work, analysts also access commercial on-line information services like DIALOG and NEXIS; they receive a continuous flow of institutional electronic mail, cable traffic, newswires, and internally published journals made up of their colleagues' reports. Reference resources may also be drawn from other analysts' extensive files; these resources, while they may appropriately be thought of as comparable to library resources, are personal and local. Thus an analyst's sources may range from public archival materials, through institutional publications and shared collections, to personal files, collected, organized, and annotated for individual use.
Reading rooms provide an interesting window onto the artifactual bases for analysis. Reading rooms hold collections of archival reference materials that act as a resource for a group of analysts and are often given a different status and organization than one would find in a library, digital or otherwise. Like long term files, reading rooms are a locally constructed, locally available, locally controlled and maintained resource; like libraries, reading rooms contain archival materials (which are, of course, limited in scope and more highly tailored to the tasks of their immediate constituency). It is through reading rooms that we see the most direct connection between products and sources -- analysts' reports are made formally available and can become sources for other analysts.
Some portion of the information provided by institutional sources and commercial services has a transient quality to it, especially when it is brought to the task of analysis: the sources are most valuable when they are current, and much new material is added daily. This holds true for some of the reports that the analysts produce themselves as well -- they have a transience related to the perception of timeliness, a perception that is generally considered to conflict with the archival nature of library materials. On the other hand, some of the materials published within the analysts' organization -- journals and longer reports, in particular -- are often accorded the authority and archival properties of library materials, in spite of their less-than-public scope. The transience of materials exists independently from their fixity; it is from the fixity of sources (rather than their permanence) and the analysts' ability to trace conclusions back to the particular motivating sources that analytic work in part derives its authority.
More recent discussions with Congressional Research Service analysts (CRS is the research arm of the Library of Congress) also bring to light the difference between the fixity of components and the fixity of composite document structures in which these component documents may participate. CRS analysts produce report packs, collections of documents pulled together for a specific current need (although the original documents may have been produced for a variety of requirements). Thus while the reports themselves might be considered fixed for one period of time, the report pack itself is fixed for a separate, shorter duration. It is a composite document that derives its status as a single entity through an imposed structure (physically through temporary, ad hoc binding into a folder). The life of the composite is considerably shorter than the life of its components.
Recovery of documents from the multiplicity of available sources is guided by pragmatic constraints rather than those suggested by the artificial metrics of precision and recall. Analysts seldom ask, "Do I have everything that's relevant to this topic?" but instead ask, "Do I have time to read all this material?" or "Do I have what I need?" Retrieval and filtering then are guided by a desire to get the right number of documents or a representative set of documents rather than all the right documents. Frustration at content redundancy sometimes overshadows concerns with content relevance; because so many information services offer coverage of the same events, analysts frequently retrieve materials that are identified as unique, but that cover essentially the same ground.
5.2. Interpretation: a culture of annotation
The exigencies of analysis foster a culture of annotation. Analysts generally don't take notes by writing their observations down on a separate sheet of paper or in a text editor (although under pressure they may combine the activity of reading source materials with the concomitant activity of writing a new report). Instead, they mark on the documents themselves. In the case of books, photographic imagery, and paper archival materials, they do this marking in a non-destructive way -- they use post-its and stick-on signals (little colored dots). In the case of digitally-delivered documents, they print copies, and use these to contextualize notes. They highlight segments of text (sometimes whole paragraphs) and they scribble marginalia, sometimes noting where what they've seen in the text differs from what they'd expect to see ("Not true!"). They also print automatically marked text, documents retrieved from databases that have the keywords that triggered retrieval or filtering explicitly marked (usually underlined). These marking practices increase the value of the documents to the analysts and form the basis for their personal and shared files.
In spite of an institutional imperative to make all sources available through electronic means and all composition and final production digital, analysts make extensive use of paper as the principle interpretive medium. As we have pointed out, it is a valuable medium for recording many types of annotations not readily recorded in a digital medium. It is also manipulable in a manner that is not afforded by digital documents -- analysts can express nuances of meaning by simply juxtaposing paper documents on their desks; it is common for analysts to spread out their working papers over every available surface and to shuffle them around to reflect various alternative organizations.
Analysts' personal and workgroup (paper) files are used in a similar way -- materials can be extracted (for example, torn from a personal copy of a journal), grouped together, and categorized into very lightweight, flexible structures that reflect the elements of their specialty that they wish to be perspicuous. It is easy for analysts to visually inspect their files and know at a glance what kinds of materials are locally available on a topic, how much they've collected, and how much a given file is used.
The analysts' use of paper demonstrates its unique affordances for marking, annotation, and expressing interpretive structures, and suggests that a complete shift to digital technologies for analysis is unlikely.
5.3. Collaboration: the unseen practice
Practitioners of information-intensive intellectual work mostly will, if asked, assert that they work alone, discounting both their own reliance on their peers and the contributions they've made to the work of others. Yet a closer examination of their work shows us many kinds of collaborative practices, most of them informal and most of them institutionally unrewarded. For example, analysts working in different media (one in the scientific literature, another in satellite imagery) might confer over the phone, looking for corroboration of an interpretation, or to ask "what do you make of this?" Occasionally these informal collaborations grow into institutionally acknowledged co-authorships, in which the analysts write a report together, usually by passing drafts back and forth. But more commonly, informal collaborations remain just that, and are a deeply underappreciated part of intellectual work.
Analysts also share interpretive structures and partial interpretations of documents through mutual access to a set of files; one person's files are another analyst's well-tended and well-shepherded reference library. Occasionally, a group of analysts covering a shared topic or area will gather materials into a structured database or shared filing cabinet that a whole group refers to; more often, the materials they share are just those they've pulled out of an information resource and filed individually (sometimes in a container as informal as a shoebox). They rely on each other to act as librarians in this situation, evaluating the authority of and providing access to collected materials. Analysts usually are aware of the kinds of materials they can rely on each other to collect.
Other collaborations cut across types of work. Analysts often rely on the assistance of professional librarians or information specialists, people whose job it is to understand the art of query formulation against various databases or to find information in a traditional library setting. These specialists often work closely with analysts, first to pin down the analyst's particular information needs (since they often approach the specialists with a vaguely formed topic or a very general idea of what they need), then to identify possible places that might be rich sources of this type of information, and finally to refine queries so they return an "acceptable" number of documents (for, as we asserted earlier, it is seldom questions of precision and recall that guide the actual information retrieval process, but rather pragmatic constraints dictated by the length of time an analyst has to complete an assignment and the patience she has for reading semi-redundant documents).
The work of information analysts thus clearly shows how a broad range of materials (fluid as well as fixed, transient as well as permanent, paper as well as digital) is used in a collaborative fashion. For this class of users, a digital library, narrowly construed, would be highly inadequate if it was the sole, or the primary, information source. Faced with this evidence, one might still argue that this an aberrant or a minor case; we don't think so.
6. Summary and Implications
The term "digital library," it seems, is now taken to be synonymous with "the library of the future." For this to be true, however, digital libraries will have to encompass a broader range of materials, technologies, and uses than the current normative conception allows. Clearly, a digital library community is now forming. What better time to discuss the breadth of the subject matter, as well as the skills and perspectives needed to tackle it?
Since we are technologists, we close with three implications for technology development that follow from our observations to this point.
* Media integration. If paper is not about to disappear, the technological challenge will be to effect a rich integration of digital and paper forms. How must our protocols, our naming schemes, our search procedures be broadened if some of the referents are not -- and will never be -- in digital form? How must our document architectures be modified to accommodate hybrid documents, parts of which are in digital form while others are on paper?
* Versioning. Today's commercially available digital versioning schemes are inadequate for many document-intensive tasks. Current computer operating systems provide little or no help in keeping track of versions of files or documents. In at least one community, the hypertext/hypermedia community, this has become the subject of considerable research and discussion. So long as digital libraries are taken to involve relatively fixed, relatively permanent material, versioning will be considered to be a minor concern. But if future collections do include more fluid and transient material, then richer schemes for naming, finding, and manipulating versions will be critical -- all the more so if collections include hybrids of digital and paper forms.
* Tools for collaboration and communication. If library use is highly collaborative, then the tools we provide must not make unwarranted assumptions about single-user browsing and access. Tools will need not only to support a range of collaborative activities, such as shared annotation and the maintenance of local subcollections of materials, but also the communicative behaviors that underlie all work practice.
 Anderson, W. L., and Crocca, W. T. Engineering Practice and Codevelopment of Product Prototypes. 1993. Communications of the ACM 36, 4, 49-56.
 Blomberg, J., Giacomi, J., Mosher, A., & Swenton-Wall, P. 1993. Ethnographic field methods and their relation to design. In Participatory Design: Perspectives on Systems Design., D. Schuler & A. Namioka, Eds. Hillsdale, New Jersey: Lawrence Erlbaum Associates, 123-154.
 Bolter, J.D. 1991 Writing Space: The Computer, Hypertext, and the History of Writing. Hillsdale, New Jersey: Lawrence Erlbaum Associates.
 Brown, J.S. and Duguid, P. 1994. Borderline Issues: Social and Material Aspects of Design. Human-Computer Interaction 9, 1, 3-36.
 Brun-Cottan, F., Forbes, K., Goodwin, C., Goodwin, M., Jordan, B., Suchman, L., and Trigg, R. 1991. The Workplace Project: Designing for diversity and change (Videotape). Xerox Palo Alto Research Center, 3333 Coyote Hill Road, Palo Alto, CA 94304.
 Chan, S. and Sylvan, D. A. 1984. Foreign Policy Decision Making: An Overview. In Foreign Policy Decision Making, Sylvan and Chan Eds. New York: Prager, 1-19.
 Ehrlich, K. and Cash, D. 1994. Turning Information into Knowledge: Information Finding as a Collaborative Activity. In Proceedings of Digital Libraries '94, College Station, Texas, June (this volume).
 King, G., Kung, H., Grosz, B., Verba, S., Flecker, D., Kahin, B. 1994. The Harvard Self-Enriching Library Facilities (SELF) Project. In Proceedings of Digital Libraries '94, College Station, Texas, June (this volume).
 Levy, D. M. 1994. Fixed or Fluid? Document Stability and New Media. Submitted to ECHT '94, Edinburgh, September 18-23.
 Marshall, C.C. 1990. Work Practice Study: Analysts and Notetaking. Xerox Palo Alto Research Center, 3333 Coyote Hill Road, Palo Alto, CA 94304.
 Marshall, C.C. 1990. A Multi-Tiered Approach to Hypertext Integration: Negotiating Standards for a Heterogeneous Application Environment. In Proceedings of the NIST Workshop on Hypertext Standardization, Gaithersburg, MD, January.
 Marshall, C., Shipman, F., McCall, R. 1994. Putting Digital Libraries to Work: Issues from Experience with Community Memories. In Proceedings of Digital Libraries '94, College Station, Texas, June (this volume).
 Nunberg. G. 1993. The Places of Book in the Age of Electronic Reproduction. Representations, 42, 13-37.
 Schnase, J., Leggett, J., Metcalfe, E., Morin, N., Cunnius, E., Turner, J., Furuta, R., Ellis, L., Pilant, M., Ewing, R., Hassan, S., and Frisse, M. 1994. The CoLib Project: Enabling Digital Botany for the 21st Century. In Proceedings of Digital Libraries '94, College Station, Texas, June (this volume).
 Stotts, D., Smith J., Dewan, P., Jeffay, K., Smith, F. D., Smith D., Weiss, S., Coggins, J., and Oliver, W. 1994. A Patterned Injury Digital Library for Collaborative Forensic Medicine. In Proceedings of Digital Libraries '94, College Station, Texas, June (this volume).
 Suchman, L. A. 1993. Centers of Coordination: A Case and Some Themes. Presented at the NATO Advanced Research Workshop on Discourse, Tools and Reasoning, Lucca, Italy, November 2-7.
 Yates, J. 1989. Control Through Communication. Baltimore: The Johns Hopkins University Press.
 ARPA CSTR Project.